1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
2 // expected-no-diagnostics
3 #ifndef HEADER
4 #define HEADER
5
6 // Test host codegen.
7 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
8 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
9 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK2
10 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
11 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
12 // RUN: %clang_cc1 -DCK1 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK4
13
14 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK5
15 // RUN: %clang_cc1 -DCK1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
16 // RUN: %clang_cc1 -DCK1 -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK6
17 // RUN: %clang_cc1 -DCK1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK7
18 // RUN: %clang_cc1 -DCK1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
19 // RUN: %clang_cc1 -DCK1 -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK8
20
21 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
22 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
23 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
24 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
25 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
26 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
27 #ifdef CK1
28
29 template <typename T, int X, long long Y>
30 struct SS{
31 T a[X];
32 float b;
fooSS33 int foo(void) {
34
35 #pragma omp target
36 #pragma omp teams distribute parallel for
37 for(int i = 0; i < X; i++) {
38 a[i] = (T)0;
39 }
40 #pragma omp target
41 #pragma omp teams distribute parallel for schedule(static)
42 for(int i = 0; i < X; i++) {
43 a[i] = (T)0;
44 }
45 #pragma omp target
46 #pragma omp teams distribute parallel for schedule(static, X/2)
47 for(int i = 0; i < X; i++) {
48 a[i] = (T)0;
49 }
50
51 #pragma omp target
52 #pragma omp teams distribute parallel for schedule(dynamic)
53 for(int i = 0; i < X; i++) {
54 a[i] = (T)0;
55 }
56
57 #pragma omp target
58 #pragma omp teams distribute parallel for schedule(dynamic, X/2)
59 for(int i = 0; i < X; i++) {
60 a[i] = (T)0;
61 }
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78 return a[0];
79 }
80 };
81
teams_template_struct(void)82 int teams_template_struct(void) {
83 SS<int, 123, 456> V;
84 return V.foo();
85
86 }
87 #endif // CK1
88
89 // Test host codegen.
90 // RUN: %clang_cc1 -DCK2 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK13
91 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
92 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK14
93 // RUN: %clang_cc1 -DCK2 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK15
94 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
95 // RUN: %clang_cc1 -DCK2 -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK16
96
97 // RUN: %clang_cc1 -DCK2 -verify -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK17
98 // RUN: %clang_cc1 -DCK2 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
99 // RUN: %clang_cc1 -DCK2 -fopenmp -fopenmp-version=45 -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK18
100 // RUN: %clang_cc1 -DCK2 -verify -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK19
101 // RUN: %clang_cc1 -DCK2 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
102 // RUN: %clang_cc1 -DCK2 -fopenmp -fopenmp-version=45 -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK20
103
104 // RUN: %clang_cc1 -DCK2 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
105 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
106 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
107 // RUN: %clang_cc1 -DCK2 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
108 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
109 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
110 #ifdef CK2
111
112 template <typename T, int n>
tmain(T argc)113 int tmain(T argc) {
114 T a[n];
115 int m = 10;
116 #pragma omp target
117 #pragma omp teams distribute parallel for
118 for(int i = 0; i < n; i++) {
119 a[i] = (T)0;
120 }
121 #pragma omp target
122 #pragma omp teams distribute parallel for schedule(static)
123 for(int i = 0; i < n; i++) {
124 a[i] = (T)0;
125 }
126 #pragma omp target
127 #pragma omp teams distribute parallel for schedule(static, m)
128 for(int i = 0; i < n; i++) {
129 a[i] = (T)0;
130 }
131 #pragma omp target
132 #pragma omp teams distribute parallel for schedule(dynamic)
133 for(int i = 0; i < n; i++) {
134 a[i] = (T)0;
135 }
136 #pragma omp target
137 #pragma omp teams distribute parallel for schedule(dynamic, m)
138 for(int i = 0; i < n; i++) {
139 a[i] = (T)0;
140 }
141 return 0;
142 }
143
main(int argc,char ** argv)144 int main (int argc, char **argv) {
145 int n = 100;
146 int a[n];
147 int m = 10;
148 #pragma omp target
149 #pragma omp teams distribute parallel for
150 for(int i = 0; i < n; i++) {
151 a[i] = 0;
152 }
153 #pragma omp target
154 #pragma omp teams distribute parallel for dist_schedule(static)
155 for(int i = 0; i < n; i++) {
156 a[i] = 0;
157 }
158 #pragma omp target
159 #pragma omp teams distribute parallel for dist_schedule(static, m)
160 for(int i = 0; i < n; i++) {
161 a[i] = 0;
162 }
163 #pragma omp target
164 #pragma omp teams distribute parallel for schedule(dynamic)
165 for(int i = 0; i < n; i++) {
166 a[i] = 0;
167 }
168 #pragma omp target
169 #pragma omp teams distribute parallel for schedule(dynamic, m)
170 for(int i = 0; i < n; i++) {
171 a[i] = 0;
172 }
173 return tmain<int, 10>(argc);
174 }
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209 #endif // CK2
210 #endif // #ifndef HEADER
211 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
212 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
213 // CHECK1-NEXT: entry:
214 // CHECK1-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
215 // CHECK1-NEXT: [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
216 // CHECK1-NEXT: ret i32 [[CALL]]
217 //
218 //
219 // CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
220 // CHECK1-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
221 // CHECK1-NEXT: entry:
222 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
223 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
224 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
225 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
226 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
227 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
228 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
229 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
230 // CHECK1-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
231 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 8
232 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 8
233 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 8
234 // CHECK1-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
235 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 8
236 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 8
237 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 8
238 // CHECK1-NEXT: [[_TMP20:%.*]] = alloca i32, align 4
239 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 8
240 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 8
241 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 8
242 // CHECK1-NEXT: [[_TMP27:%.*]] = alloca i32, align 4
243 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
244 // CHECK1-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
245 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
246 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
247 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
248 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
249 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
250 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
251 // CHECK1-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
252 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
253 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
254 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
255 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
256 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
257 // CHECK1-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
258 // CHECK1-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
259 // CHECK1-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
260 // CHECK1: omp_offload.failed:
261 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
262 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
263 // CHECK1: omp_offload.cont:
264 // CHECK1-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
265 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
266 // CHECK1-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
267 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 8
268 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
269 // CHECK1-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
270 // CHECK1-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 8
271 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
272 // CHECK1-NEXT: store i8* null, i8** [[TMP13]], align 8
273 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
274 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
275 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
276 // CHECK1-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
277 // CHECK1-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
278 // CHECK1-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
279 // CHECK1: omp_offload.failed7:
280 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40(%struct.SS* [[THIS1]]) #[[ATTR2]]
281 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT8]]
282 // CHECK1: omp_offload.cont8:
283 // CHECK1-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
284 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
285 // CHECK1-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
286 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
287 // CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
288 // CHECK1-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
289 // CHECK1-NEXT: store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 8
290 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 0
291 // CHECK1-NEXT: store i8* null, i8** [[TMP22]], align 8
292 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
293 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
294 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
295 // CHECK1-NEXT: [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
296 // CHECK1-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
297 // CHECK1-NEXT: br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
298 // CHECK1: omp_offload.failed14:
299 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45(%struct.SS* [[THIS1]]) #[[ATTR2]]
300 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT15]]
301 // CHECK1: omp_offload.cont15:
302 // CHECK1-NEXT: [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
303 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
304 // CHECK1-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
305 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 8
306 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
307 // CHECK1-NEXT: [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
308 // CHECK1-NEXT: store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 8
309 // CHECK1-NEXT: [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i64 0, i64 0
310 // CHECK1-NEXT: store i8* null, i8** [[TMP31]], align 8
311 // CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
312 // CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
313 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
314 // CHECK1-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
315 // CHECK1-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
316 // CHECK1-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
317 // CHECK1: omp_offload.failed21:
318 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51(%struct.SS* [[THIS1]]) #[[ATTR2]]
319 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT22]]
320 // CHECK1: omp_offload.cont22:
321 // CHECK1-NEXT: [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
322 // CHECK1-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
323 // CHECK1-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
324 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
325 // CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
326 // CHECK1-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
327 // CHECK1-NEXT: store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 8
328 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 0
329 // CHECK1-NEXT: store i8* null, i8** [[TMP40]], align 8
330 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
331 // CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
332 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
333 // CHECK1-NEXT: [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
334 // CHECK1-NEXT: [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
335 // CHECK1-NEXT: br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
336 // CHECK1: omp_offload.failed28:
337 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57(%struct.SS* [[THIS1]]) #[[ATTR2]]
338 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT29]]
339 // CHECK1: omp_offload.cont29:
340 // CHECK1-NEXT: [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
341 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 0
342 // CHECK1-NEXT: [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
343 // CHECK1-NEXT: ret i32 [[TMP45]]
344 //
345 //
346 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35
347 // CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
348 // CHECK1-NEXT: entry:
349 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
350 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
351 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
352 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
353 // CHECK1-NEXT: ret void
354 //
355 //
356 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
357 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
358 // CHECK1-NEXT: entry:
359 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
360 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
361 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
362 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
363 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
364 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
365 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
366 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
367 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
368 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
369 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
370 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
371 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
372 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
373 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
374 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
375 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
376 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
377 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
378 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
379 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
380 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
381 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
382 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
383 // CHECK1: cond.true:
384 // CHECK1-NEXT: br label [[COND_END:%.*]]
385 // CHECK1: cond.false:
386 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
387 // CHECK1-NEXT: br label [[COND_END]]
388 // CHECK1: cond.end:
389 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
390 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
391 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
392 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
393 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
394 // CHECK1: omp.inner.for.cond:
395 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
396 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
397 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
398 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
399 // CHECK1: omp.inner.for.body:
400 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
401 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
402 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
403 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
404 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
405 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
406 // CHECK1: omp.inner.for.inc:
407 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
408 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
409 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
410 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
411 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
412 // CHECK1: omp.inner.for.end:
413 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
414 // CHECK1: omp.loop.exit:
415 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
416 // CHECK1-NEXT: ret void
417 //
418 //
419 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
420 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
421 // CHECK1-NEXT: entry:
422 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
423 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
424 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
425 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
426 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
427 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
428 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
429 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
430 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
431 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
432 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
433 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
434 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
435 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
436 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
437 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
438 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
439 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
440 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
441 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
442 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
443 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
444 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
445 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
446 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
447 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
448 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
449 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
450 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
451 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
452 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
453 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
454 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
455 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
456 // CHECK1: cond.true:
457 // CHECK1-NEXT: br label [[COND_END:%.*]]
458 // CHECK1: cond.false:
459 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
460 // CHECK1-NEXT: br label [[COND_END]]
461 // CHECK1: cond.end:
462 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
463 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
464 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
465 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
466 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
467 // CHECK1: omp.inner.for.cond:
468 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
469 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
470 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
471 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
472 // CHECK1: omp.inner.for.body:
473 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
474 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
475 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
476 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
477 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
478 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
479 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
480 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
481 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
482 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
483 // CHECK1: omp.body.continue:
484 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
485 // CHECK1: omp.inner.for.inc:
486 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
487 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
488 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
489 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
490 // CHECK1: omp.inner.for.end:
491 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
492 // CHECK1: omp.loop.exit:
493 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
494 // CHECK1-NEXT: ret void
495 //
496 //
497 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40
498 // CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
499 // CHECK1-NEXT: entry:
500 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
501 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
502 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
503 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
504 // CHECK1-NEXT: ret void
505 //
506 //
507 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
508 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
509 // CHECK1-NEXT: entry:
510 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
511 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
512 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
513 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
514 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
515 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
516 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
517 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
518 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
519 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
520 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
521 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
522 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
523 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
524 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
525 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
526 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
527 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
528 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
529 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
530 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
531 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
532 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
533 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
534 // CHECK1: cond.true:
535 // CHECK1-NEXT: br label [[COND_END:%.*]]
536 // CHECK1: cond.false:
537 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
538 // CHECK1-NEXT: br label [[COND_END]]
539 // CHECK1: cond.end:
540 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
541 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
542 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
543 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
544 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
545 // CHECK1: omp.inner.for.cond:
546 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
547 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
548 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
549 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
550 // CHECK1: omp.inner.for.body:
551 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
552 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
553 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
554 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
555 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
556 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
557 // CHECK1: omp.inner.for.inc:
558 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
559 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
560 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
561 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
562 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
563 // CHECK1: omp.inner.for.end:
564 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
565 // CHECK1: omp.loop.exit:
566 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
567 // CHECK1-NEXT: ret void
568 //
569 //
570 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
571 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
572 // CHECK1-NEXT: entry:
573 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
574 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
575 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
576 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
577 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
578 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
579 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
580 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
581 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
582 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
583 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
584 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
585 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
586 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
587 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
588 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
589 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
590 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
591 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
592 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
593 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
594 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
595 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
596 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
597 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
598 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
599 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
600 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
601 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
602 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
603 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
604 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
605 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
606 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
607 // CHECK1: cond.true:
608 // CHECK1-NEXT: br label [[COND_END:%.*]]
609 // CHECK1: cond.false:
610 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
611 // CHECK1-NEXT: br label [[COND_END]]
612 // CHECK1: cond.end:
613 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
614 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
615 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
616 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
617 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
618 // CHECK1: omp.inner.for.cond:
619 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
620 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
621 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
622 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
623 // CHECK1: omp.inner.for.body:
624 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
625 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
626 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
627 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
628 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
629 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
630 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
631 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
632 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
633 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
634 // CHECK1: omp.body.continue:
635 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
636 // CHECK1: omp.inner.for.inc:
637 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
638 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
639 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
640 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
641 // CHECK1: omp.inner.for.end:
642 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
643 // CHECK1: omp.loop.exit:
644 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
645 // CHECK1-NEXT: ret void
646 //
647 //
648 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45
649 // CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
650 // CHECK1-NEXT: entry:
651 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
652 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
653 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
654 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
655 // CHECK1-NEXT: ret void
656 //
657 //
658 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
659 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
660 // CHECK1-NEXT: entry:
661 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
662 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
663 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
664 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
665 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
666 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
667 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
668 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
669 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
670 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
671 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
672 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
673 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
674 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
675 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
676 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
677 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
678 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
679 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
680 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
681 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
682 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
683 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
684 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
685 // CHECK1: cond.true:
686 // CHECK1-NEXT: br label [[COND_END:%.*]]
687 // CHECK1: cond.false:
688 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
689 // CHECK1-NEXT: br label [[COND_END]]
690 // CHECK1: cond.end:
691 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
692 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
693 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
694 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
695 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
696 // CHECK1: omp.inner.for.cond:
697 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
698 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
699 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
700 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
701 // CHECK1: omp.inner.for.body:
702 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
703 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
704 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
705 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
706 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
707 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
708 // CHECK1: omp.inner.for.inc:
709 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
710 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
711 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
712 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
713 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
714 // CHECK1: omp.inner.for.end:
715 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
716 // CHECK1: omp.loop.exit:
717 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
718 // CHECK1-NEXT: ret void
719 //
720 //
721 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
722 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
723 // CHECK1-NEXT: entry:
724 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
725 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
726 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
727 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
728 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
729 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
730 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
731 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
732 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
733 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
734 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
735 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
736 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
737 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
738 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
739 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
740 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
741 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
742 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
743 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
744 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
745 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
746 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
747 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
748 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
749 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
750 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
751 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
752 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
753 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
754 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
755 // CHECK1-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
756 // CHECK1: omp.dispatch.cond:
757 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
758 // CHECK1-NEXT: [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
759 // CHECK1-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
760 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
761 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
762 // CHECK1: cond.true:
763 // CHECK1-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
764 // CHECK1-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
765 // CHECK1-NEXT: br label [[COND_END:%.*]]
766 // CHECK1: cond.false:
767 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
768 // CHECK1-NEXT: br label [[COND_END]]
769 // CHECK1: cond.end:
770 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
771 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
772 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
773 // CHECK1-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
774 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
775 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
776 // CHECK1-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
777 // CHECK1-NEXT: br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
778 // CHECK1: omp.dispatch.body:
779 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
780 // CHECK1: omp.inner.for.cond:
781 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
782 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
783 // CHECK1-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
784 // CHECK1-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
785 // CHECK1: omp.inner.for.body:
786 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
787 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
788 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
789 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
790 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
791 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[I]], align 4
792 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
793 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
794 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
795 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
796 // CHECK1: omp.body.continue:
797 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
798 // CHECK1: omp.inner.for.inc:
799 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
800 // CHECK1-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
801 // CHECK1-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
802 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
803 // CHECK1: omp.inner.for.end:
804 // CHECK1-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
805 // CHECK1: omp.dispatch.inc:
806 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
807 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
808 // CHECK1-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
809 // CHECK1-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
810 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
811 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
812 // CHECK1-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
813 // CHECK1-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
814 // CHECK1-NEXT: br label [[OMP_DISPATCH_COND]]
815 // CHECK1: omp.dispatch.end:
816 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
817 // CHECK1-NEXT: ret void
818 //
819 //
820 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
821 // CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
822 // CHECK1-NEXT: entry:
823 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
824 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
825 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
826 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
827 // CHECK1-NEXT: ret void
828 //
829 //
830 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..10
831 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
832 // CHECK1-NEXT: entry:
833 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
834 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
835 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
836 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
837 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
838 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
839 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
840 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
841 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
842 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
843 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
844 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
845 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
846 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
847 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
848 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
849 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
850 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
851 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
852 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
853 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
854 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
855 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
856 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
857 // CHECK1: cond.true:
858 // CHECK1-NEXT: br label [[COND_END:%.*]]
859 // CHECK1: cond.false:
860 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
861 // CHECK1-NEXT: br label [[COND_END]]
862 // CHECK1: cond.end:
863 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
864 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
865 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
866 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
867 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
868 // CHECK1: omp.inner.for.cond:
869 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
870 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
871 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
872 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
873 // CHECK1: omp.inner.for.body:
874 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
875 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
876 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
877 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
878 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
879 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
880 // CHECK1: omp.inner.for.inc:
881 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
882 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
883 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
884 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
885 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
886 // CHECK1: omp.inner.for.end:
887 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
888 // CHECK1: omp.loop.exit:
889 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
890 // CHECK1-NEXT: ret void
891 //
892 //
893 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
894 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
895 // CHECK1-NEXT: entry:
896 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
897 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
898 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
899 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
900 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
901 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
902 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
903 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
904 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
905 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
906 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
907 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
908 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
909 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
910 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
911 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
912 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
913 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
914 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
915 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
916 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
917 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
918 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
919 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
920 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
921 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
922 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
923 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
924 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
925 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
926 // CHECK1-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
927 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
928 // CHECK1-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
929 // CHECK1-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
930 // CHECK1: omp.dispatch.cond:
931 // CHECK1-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
932 // CHECK1-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
933 // CHECK1-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
934 // CHECK1: omp.dispatch.body:
935 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
936 // CHECK1-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
937 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
938 // CHECK1: omp.inner.for.cond:
939 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
940 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !10
941 // CHECK1-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
942 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
943 // CHECK1: omp.inner.for.body:
944 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
945 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
946 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
947 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !10
948 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
949 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !10
950 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
951 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
952 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !10
953 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
954 // CHECK1: omp.body.continue:
955 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
956 // CHECK1: omp.inner.for.inc:
957 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
958 // CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
959 // CHECK1-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
960 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
961 // CHECK1: omp.inner.for.end:
962 // CHECK1-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
963 // CHECK1: omp.dispatch.inc:
964 // CHECK1-NEXT: br label [[OMP_DISPATCH_COND]]
965 // CHECK1: omp.dispatch.end:
966 // CHECK1-NEXT: ret void
967 //
968 //
969 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57
970 // CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
971 // CHECK1-NEXT: entry:
972 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
973 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
974 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
975 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
976 // CHECK1-NEXT: ret void
977 //
978 //
979 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..14
980 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
981 // CHECK1-NEXT: entry:
982 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
983 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
984 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
985 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
986 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
987 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
988 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
989 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
990 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
991 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
992 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
993 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
994 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
995 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
996 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
997 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
998 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
999 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1000 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1001 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1002 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1003 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1004 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1005 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1006 // CHECK1: cond.true:
1007 // CHECK1-NEXT: br label [[COND_END:%.*]]
1008 // CHECK1: cond.false:
1009 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1010 // CHECK1-NEXT: br label [[COND_END]]
1011 // CHECK1: cond.end:
1012 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1013 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1014 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1015 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1016 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1017 // CHECK1: omp.inner.for.cond:
1018 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1019 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1020 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1021 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1022 // CHECK1: omp.inner.for.body:
1023 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1024 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
1025 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1026 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
1027 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
1028 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1029 // CHECK1: omp.inner.for.inc:
1030 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1031 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1032 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1033 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1034 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
1035 // CHECK1: omp.inner.for.end:
1036 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1037 // CHECK1: omp.loop.exit:
1038 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1039 // CHECK1-NEXT: ret void
1040 //
1041 //
1042 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..15
1043 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1044 // CHECK1-NEXT: entry:
1045 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1046 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1047 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1048 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1049 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1050 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1051 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
1052 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1053 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1054 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1055 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1056 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
1057 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1058 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1059 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1060 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1061 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1062 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1063 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1064 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1065 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1066 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
1067 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1068 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
1069 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1070 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1071 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1072 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1073 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1074 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1075 // CHECK1-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1076 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
1077 // CHECK1-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
1078 // CHECK1-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
1079 // CHECK1: omp.dispatch.cond:
1080 // CHECK1-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
1081 // CHECK1-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
1082 // CHECK1-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1083 // CHECK1: omp.dispatch.body:
1084 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1085 // CHECK1-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
1086 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1087 // CHECK1: omp.inner.for.cond:
1088 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
1089 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
1090 // CHECK1-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
1091 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1092 // CHECK1: omp.inner.for.body:
1093 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
1094 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
1095 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1096 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
1097 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1098 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
1099 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
1100 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
1101 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
1102 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1103 // CHECK1: omp.body.continue:
1104 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1105 // CHECK1: omp.inner.for.inc:
1106 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
1107 // CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
1108 // CHECK1-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
1109 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
1110 // CHECK1: omp.inner.for.end:
1111 // CHECK1-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
1112 // CHECK1: omp.dispatch.inc:
1113 // CHECK1-NEXT: br label [[OMP_DISPATCH_COND]]
1114 // CHECK1: omp.dispatch.end:
1115 // CHECK1-NEXT: ret void
1116 //
1117 //
1118 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1119 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
1120 // CHECK1-NEXT: entry:
1121 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
1122 // CHECK1-NEXT: ret void
1123 //
1124 //
1125 // CHECK2-LABEL: define {{[^@]+}}@_Z21teams_template_structv
1126 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
1127 // CHECK2-NEXT: entry:
1128 // CHECK2-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
1129 // CHECK2-NEXT: [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
1130 // CHECK2-NEXT: ret i32 [[CALL]]
1131 //
1132 //
1133 // CHECK2-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
1134 // CHECK2-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
1135 // CHECK2-NEXT: entry:
1136 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1137 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
1138 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
1139 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
1140 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1141 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
1142 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
1143 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
1144 // CHECK2-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
1145 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 8
1146 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 8
1147 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 8
1148 // CHECK2-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
1149 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 8
1150 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 8
1151 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 8
1152 // CHECK2-NEXT: [[_TMP20:%.*]] = alloca i32, align 4
1153 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 8
1154 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 8
1155 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 8
1156 // CHECK2-NEXT: [[_TMP27:%.*]] = alloca i32, align 4
1157 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1158 // CHECK2-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1159 // CHECK2-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
1160 // CHECK2-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1161 // CHECK2-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
1162 // CHECK2-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
1163 // CHECK2-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1164 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
1165 // CHECK2-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
1166 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1167 // CHECK2-NEXT: store i8* null, i8** [[TMP4]], align 8
1168 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1169 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1170 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
1171 // CHECK2-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1172 // CHECK2-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
1173 // CHECK2-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1174 // CHECK2: omp_offload.failed:
1175 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
1176 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT]]
1177 // CHECK2: omp_offload.cont:
1178 // CHECK2-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1179 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
1180 // CHECK2-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
1181 // CHECK2-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 8
1182 // CHECK2-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
1183 // CHECK2-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
1184 // CHECK2-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 8
1185 // CHECK2-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
1186 // CHECK2-NEXT: store i8* null, i8** [[TMP13]], align 8
1187 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
1188 // CHECK2-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
1189 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
1190 // CHECK2-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1191 // CHECK2-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
1192 // CHECK2-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
1193 // CHECK2: omp_offload.failed7:
1194 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40(%struct.SS* [[THIS1]]) #[[ATTR2]]
1195 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT8]]
1196 // CHECK2: omp_offload.cont8:
1197 // CHECK2-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1198 // CHECK2-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
1199 // CHECK2-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
1200 // CHECK2-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
1201 // CHECK2-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
1202 // CHECK2-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
1203 // CHECK2-NEXT: store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 8
1204 // CHECK2-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 0
1205 // CHECK2-NEXT: store i8* null, i8** [[TMP22]], align 8
1206 // CHECK2-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
1207 // CHECK2-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
1208 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
1209 // CHECK2-NEXT: [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1210 // CHECK2-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
1211 // CHECK2-NEXT: br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
1212 // CHECK2: omp_offload.failed14:
1213 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45(%struct.SS* [[THIS1]]) #[[ATTR2]]
1214 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT15]]
1215 // CHECK2: omp_offload.cont15:
1216 // CHECK2-NEXT: [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1217 // CHECK2-NEXT: [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
1218 // CHECK2-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
1219 // CHECK2-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 8
1220 // CHECK2-NEXT: [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
1221 // CHECK2-NEXT: [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
1222 // CHECK2-NEXT: store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 8
1223 // CHECK2-NEXT: [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i64 0, i64 0
1224 // CHECK2-NEXT: store i8* null, i8** [[TMP31]], align 8
1225 // CHECK2-NEXT: [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
1226 // CHECK2-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
1227 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
1228 // CHECK2-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1229 // CHECK2-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
1230 // CHECK2-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
1231 // CHECK2: omp_offload.failed21:
1232 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51(%struct.SS* [[THIS1]]) #[[ATTR2]]
1233 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT22]]
1234 // CHECK2: omp_offload.cont22:
1235 // CHECK2-NEXT: [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1236 // CHECK2-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
1237 // CHECK2-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
1238 // CHECK2-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
1239 // CHECK2-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
1240 // CHECK2-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
1241 // CHECK2-NEXT: store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 8
1242 // CHECK2-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 0
1243 // CHECK2-NEXT: store i8* null, i8** [[TMP40]], align 8
1244 // CHECK2-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
1245 // CHECK2-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
1246 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
1247 // CHECK2-NEXT: [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1248 // CHECK2-NEXT: [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
1249 // CHECK2-NEXT: br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
1250 // CHECK2: omp_offload.failed28:
1251 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57(%struct.SS* [[THIS1]]) #[[ATTR2]]
1252 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT29]]
1253 // CHECK2: omp_offload.cont29:
1254 // CHECK2-NEXT: [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1255 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 0
1256 // CHECK2-NEXT: [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1257 // CHECK2-NEXT: ret i32 [[TMP45]]
1258 //
1259 //
1260 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35
1261 // CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
1262 // CHECK2-NEXT: entry:
1263 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1264 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1265 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1266 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
1267 // CHECK2-NEXT: ret void
1268 //
1269 //
1270 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
1271 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1272 // CHECK2-NEXT: entry:
1273 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1274 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1275 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1276 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1277 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1278 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1279 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1280 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1281 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1282 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1283 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1284 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1285 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1286 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1287 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1288 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1289 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1290 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1291 // CHECK2-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1292 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1293 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1294 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1295 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1296 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1297 // CHECK2: cond.true:
1298 // CHECK2-NEXT: br label [[COND_END:%.*]]
1299 // CHECK2: cond.false:
1300 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1301 // CHECK2-NEXT: br label [[COND_END]]
1302 // CHECK2: cond.end:
1303 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1304 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1305 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1306 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1307 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1308 // CHECK2: omp.inner.for.cond:
1309 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1310 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1311 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1312 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1313 // CHECK2: omp.inner.for.body:
1314 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1315 // CHECK2-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
1316 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1317 // CHECK2-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
1318 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
1319 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1320 // CHECK2: omp.inner.for.inc:
1321 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1322 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1323 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1324 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1325 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1326 // CHECK2: omp.inner.for.end:
1327 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1328 // CHECK2: omp.loop.exit:
1329 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1330 // CHECK2-NEXT: ret void
1331 //
1332 //
1333 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
1334 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1335 // CHECK2-NEXT: entry:
1336 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1337 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1338 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1339 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1340 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1341 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1342 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1343 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1344 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1345 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1346 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1347 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1348 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1349 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1350 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1351 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1352 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1353 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1354 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1355 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1356 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1357 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
1358 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1359 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
1360 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1361 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1362 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1363 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1364 // CHECK2-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1365 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1366 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1367 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1368 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1369 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1370 // CHECK2: cond.true:
1371 // CHECK2-NEXT: br label [[COND_END:%.*]]
1372 // CHECK2: cond.false:
1373 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1374 // CHECK2-NEXT: br label [[COND_END]]
1375 // CHECK2: cond.end:
1376 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1377 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1378 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1379 // CHECK2-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1380 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1381 // CHECK2: omp.inner.for.cond:
1382 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1383 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1384 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1385 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1386 // CHECK2: omp.inner.for.body:
1387 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1388 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1389 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1390 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1391 // CHECK2-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1392 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
1393 // CHECK2-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
1394 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
1395 // CHECK2-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1396 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1397 // CHECK2: omp.body.continue:
1398 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1399 // CHECK2: omp.inner.for.inc:
1400 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1401 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
1402 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1403 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1404 // CHECK2: omp.inner.for.end:
1405 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1406 // CHECK2: omp.loop.exit:
1407 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1408 // CHECK2-NEXT: ret void
1409 //
1410 //
1411 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40
1412 // CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1413 // CHECK2-NEXT: entry:
1414 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1415 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1416 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1417 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
1418 // CHECK2-NEXT: ret void
1419 //
1420 //
1421 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
1422 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1423 // CHECK2-NEXT: entry:
1424 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1425 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1426 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1427 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1428 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1429 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1430 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1431 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1432 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1433 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1434 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1435 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1436 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1437 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1438 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1439 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1440 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1441 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1442 // CHECK2-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1443 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1444 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1445 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1446 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1447 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1448 // CHECK2: cond.true:
1449 // CHECK2-NEXT: br label [[COND_END:%.*]]
1450 // CHECK2: cond.false:
1451 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1452 // CHECK2-NEXT: br label [[COND_END]]
1453 // CHECK2: cond.end:
1454 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1455 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1456 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1457 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1458 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1459 // CHECK2: omp.inner.for.cond:
1460 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1461 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1462 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1463 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1464 // CHECK2: omp.inner.for.body:
1465 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1466 // CHECK2-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
1467 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1468 // CHECK2-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
1469 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
1470 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1471 // CHECK2: omp.inner.for.inc:
1472 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1473 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1474 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1475 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1476 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1477 // CHECK2: omp.inner.for.end:
1478 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1479 // CHECK2: omp.loop.exit:
1480 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1481 // CHECK2-NEXT: ret void
1482 //
1483 //
1484 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
1485 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1486 // CHECK2-NEXT: entry:
1487 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1488 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1489 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1490 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1491 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1492 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1493 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1494 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1495 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1496 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1497 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1498 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1499 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1500 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1501 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1502 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1503 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1504 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1505 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1506 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1507 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1508 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
1509 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1510 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
1511 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1512 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1513 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1514 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1515 // CHECK2-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1516 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1517 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1518 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1519 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1520 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1521 // CHECK2: cond.true:
1522 // CHECK2-NEXT: br label [[COND_END:%.*]]
1523 // CHECK2: cond.false:
1524 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1525 // CHECK2-NEXT: br label [[COND_END]]
1526 // CHECK2: cond.end:
1527 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1528 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1529 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1530 // CHECK2-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1531 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1532 // CHECK2: omp.inner.for.cond:
1533 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1534 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1535 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1536 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1537 // CHECK2: omp.inner.for.body:
1538 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1539 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1540 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1541 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1542 // CHECK2-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1543 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
1544 // CHECK2-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
1545 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
1546 // CHECK2-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1547 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1548 // CHECK2: omp.body.continue:
1549 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1550 // CHECK2: omp.inner.for.inc:
1551 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1552 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
1553 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1554 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1555 // CHECK2: omp.inner.for.end:
1556 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1557 // CHECK2: omp.loop.exit:
1558 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1559 // CHECK2-NEXT: ret void
1560 //
1561 //
1562 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45
1563 // CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1564 // CHECK2-NEXT: entry:
1565 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1566 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1567 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1568 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
1569 // CHECK2-NEXT: ret void
1570 //
1571 //
1572 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..6
1573 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1574 // CHECK2-NEXT: entry:
1575 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1576 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1577 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1578 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1579 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1580 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1581 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1582 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1583 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1584 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1585 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1586 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1587 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1588 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1589 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1590 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1591 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1592 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1593 // CHECK2-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1594 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1595 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1596 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1597 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1598 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1599 // CHECK2: cond.true:
1600 // CHECK2-NEXT: br label [[COND_END:%.*]]
1601 // CHECK2: cond.false:
1602 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1603 // CHECK2-NEXT: br label [[COND_END]]
1604 // CHECK2: cond.end:
1605 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1606 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1607 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1608 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1609 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1610 // CHECK2: omp.inner.for.cond:
1611 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1612 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1613 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1614 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1615 // CHECK2: omp.inner.for.body:
1616 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1617 // CHECK2-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
1618 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1619 // CHECK2-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
1620 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
1621 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1622 // CHECK2: omp.inner.for.inc:
1623 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1624 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1625 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1626 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1627 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1628 // CHECK2: omp.inner.for.end:
1629 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1630 // CHECK2: omp.loop.exit:
1631 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1632 // CHECK2-NEXT: ret void
1633 //
1634 //
1635 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..7
1636 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1637 // CHECK2-NEXT: entry:
1638 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1639 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1640 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1641 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1642 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1643 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1644 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1645 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1646 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1647 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1648 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1649 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1650 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1651 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1652 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1653 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1654 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1655 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1656 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1657 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1658 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1659 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
1660 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1661 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
1662 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1663 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1664 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1665 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1666 // CHECK2-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1667 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1668 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
1669 // CHECK2-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
1670 // CHECK2: omp.dispatch.cond:
1671 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1672 // CHECK2-NEXT: [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1673 // CHECK2-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
1674 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
1675 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1676 // CHECK2: cond.true:
1677 // CHECK2-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1678 // CHECK2-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
1679 // CHECK2-NEXT: br label [[COND_END:%.*]]
1680 // CHECK2: cond.false:
1681 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1682 // CHECK2-NEXT: br label [[COND_END]]
1683 // CHECK2: cond.end:
1684 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
1685 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1686 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1687 // CHECK2-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
1688 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1689 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1690 // CHECK2-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
1691 // CHECK2-NEXT: br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1692 // CHECK2: omp.dispatch.body:
1693 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1694 // CHECK2: omp.inner.for.cond:
1695 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1696 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1697 // CHECK2-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
1698 // CHECK2-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1699 // CHECK2: omp.inner.for.body:
1700 // CHECK2-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1701 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
1702 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1703 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1704 // CHECK2-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1705 // CHECK2-NEXT: [[TMP15:%.*]] = load i32, i32* [[I]], align 4
1706 // CHECK2-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
1707 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
1708 // CHECK2-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1709 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1710 // CHECK2: omp.body.continue:
1711 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1712 // CHECK2: omp.inner.for.inc:
1713 // CHECK2-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1714 // CHECK2-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
1715 // CHECK2-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1716 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1717 // CHECK2: omp.inner.for.end:
1718 // CHECK2-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
1719 // CHECK2: omp.dispatch.inc:
1720 // CHECK2-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1721 // CHECK2-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1722 // CHECK2-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
1723 // CHECK2-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
1724 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1725 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1726 // CHECK2-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
1727 // CHECK2-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
1728 // CHECK2-NEXT: br label [[OMP_DISPATCH_COND]]
1729 // CHECK2: omp.dispatch.end:
1730 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1731 // CHECK2-NEXT: ret void
1732 //
1733 //
1734 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
1735 // CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1736 // CHECK2-NEXT: entry:
1737 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1738 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1739 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1740 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
1741 // CHECK2-NEXT: ret void
1742 //
1743 //
1744 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..10
1745 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1746 // CHECK2-NEXT: entry:
1747 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1748 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1749 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1750 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1751 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1752 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1753 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1754 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1755 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1756 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1757 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1758 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1759 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1760 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1761 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1762 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1763 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1764 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1765 // CHECK2-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1766 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1767 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1768 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1769 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1770 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1771 // CHECK2: cond.true:
1772 // CHECK2-NEXT: br label [[COND_END:%.*]]
1773 // CHECK2: cond.false:
1774 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1775 // CHECK2-NEXT: br label [[COND_END]]
1776 // CHECK2: cond.end:
1777 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1778 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1779 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1780 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1781 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1782 // CHECK2: omp.inner.for.cond:
1783 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1784 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1785 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1786 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1787 // CHECK2: omp.inner.for.body:
1788 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1789 // CHECK2-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
1790 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1791 // CHECK2-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
1792 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
1793 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1794 // CHECK2: omp.inner.for.inc:
1795 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1796 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1797 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1798 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1799 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1800 // CHECK2: omp.inner.for.end:
1801 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1802 // CHECK2: omp.loop.exit:
1803 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1804 // CHECK2-NEXT: ret void
1805 //
1806 //
1807 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..11
1808 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1809 // CHECK2-NEXT: entry:
1810 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1811 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1812 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1813 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1814 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1815 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1816 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1817 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1818 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1819 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1820 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1821 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1822 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1823 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1824 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1825 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1826 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1827 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1828 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1829 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1830 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1831 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
1832 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1833 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
1834 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1835 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1836 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1837 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1838 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1839 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1840 // CHECK2-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1841 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
1842 // CHECK2-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
1843 // CHECK2-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
1844 // CHECK2: omp.dispatch.cond:
1845 // CHECK2-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
1846 // CHECK2-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
1847 // CHECK2-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1848 // CHECK2: omp.dispatch.body:
1849 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1850 // CHECK2-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
1851 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1852 // CHECK2: omp.inner.for.cond:
1853 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
1854 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !10
1855 // CHECK2-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
1856 // CHECK2-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1857 // CHECK2: omp.inner.for.body:
1858 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
1859 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
1860 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1861 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !10
1862 // CHECK2-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1863 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !10
1864 // CHECK2-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
1865 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
1866 // CHECK2-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !10
1867 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1868 // CHECK2: omp.body.continue:
1869 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1870 // CHECK2: omp.inner.for.inc:
1871 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
1872 // CHECK2-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
1873 // CHECK2-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
1874 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
1875 // CHECK2: omp.inner.for.end:
1876 // CHECK2-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
1877 // CHECK2: omp.dispatch.inc:
1878 // CHECK2-NEXT: br label [[OMP_DISPATCH_COND]]
1879 // CHECK2: omp.dispatch.end:
1880 // CHECK2-NEXT: ret void
1881 //
1882 //
1883 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57
1884 // CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1885 // CHECK2-NEXT: entry:
1886 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1887 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1888 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1889 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
1890 // CHECK2-NEXT: ret void
1891 //
1892 //
1893 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..14
1894 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1895 // CHECK2-NEXT: entry:
1896 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1897 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1898 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1899 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1900 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1901 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1902 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1903 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1904 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1905 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1906 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1907 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1908 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1909 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1910 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1911 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1912 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1913 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1914 // CHECK2-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1915 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1916 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1917 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1918 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1919 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1920 // CHECK2: cond.true:
1921 // CHECK2-NEXT: br label [[COND_END:%.*]]
1922 // CHECK2: cond.false:
1923 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1924 // CHECK2-NEXT: br label [[COND_END]]
1925 // CHECK2: cond.end:
1926 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1927 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1928 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1929 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1930 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1931 // CHECK2: omp.inner.for.cond:
1932 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1933 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1934 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1935 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1936 // CHECK2: omp.inner.for.body:
1937 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1938 // CHECK2-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
1939 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1940 // CHECK2-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
1941 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
1942 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1943 // CHECK2: omp.inner.for.inc:
1944 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1945 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1946 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1947 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1948 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
1949 // CHECK2: omp.inner.for.end:
1950 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1951 // CHECK2: omp.loop.exit:
1952 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1953 // CHECK2-NEXT: ret void
1954 //
1955 //
1956 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..15
1957 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1958 // CHECK2-NEXT: entry:
1959 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1960 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1961 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1962 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1963 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1964 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1965 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1966 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1967 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1968 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1969 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1970 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1971 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1972 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1973 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1974 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1975 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1976 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1977 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1978 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1979 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1980 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
1981 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1982 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
1983 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1984 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1985 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1986 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1987 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1988 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1989 // CHECK2-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1990 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
1991 // CHECK2-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
1992 // CHECK2-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
1993 // CHECK2: omp.dispatch.cond:
1994 // CHECK2-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
1995 // CHECK2-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
1996 // CHECK2-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1997 // CHECK2: omp.dispatch.body:
1998 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1999 // CHECK2-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
2000 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2001 // CHECK2: omp.inner.for.cond:
2002 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
2003 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
2004 // CHECK2-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
2005 // CHECK2-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2006 // CHECK2: omp.inner.for.body:
2007 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
2008 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
2009 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2010 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
2011 // CHECK2-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2012 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
2013 // CHECK2-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
2014 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
2015 // CHECK2-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
2016 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2017 // CHECK2: omp.body.continue:
2018 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2019 // CHECK2: omp.inner.for.inc:
2020 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
2021 // CHECK2-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
2022 // CHECK2-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
2023 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
2024 // CHECK2: omp.inner.for.end:
2025 // CHECK2-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
2026 // CHECK2: omp.dispatch.inc:
2027 // CHECK2-NEXT: br label [[OMP_DISPATCH_COND]]
2028 // CHECK2: omp.dispatch.end:
2029 // CHECK2-NEXT: ret void
2030 //
2031 //
2032 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2033 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] {
2034 // CHECK2-NEXT: entry:
2035 // CHECK2-NEXT: call void @__tgt_register_requires(i64 1)
2036 // CHECK2-NEXT: ret void
2037 //
2038 //
2039 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2040 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
2041 // CHECK3-NEXT: entry:
2042 // CHECK3-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2043 // CHECK3-NEXT: [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
2044 // CHECK3-NEXT: ret i32 [[CALL]]
2045 //
2046 //
2047 // CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2048 // CHECK3-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2049 // CHECK3-NEXT: entry:
2050 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2051 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
2052 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
2053 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
2054 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
2055 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
2056 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
2057 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
2058 // CHECK3-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
2059 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 4
2060 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 4
2061 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 4
2062 // CHECK3-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
2063 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 4
2064 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 4
2065 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 4
2066 // CHECK3-NEXT: [[_TMP20:%.*]] = alloca i32, align 4
2067 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 4
2068 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 4
2069 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 4
2070 // CHECK3-NEXT: [[_TMP27:%.*]] = alloca i32, align 4
2071 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2072 // CHECK3-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2073 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
2074 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2075 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
2076 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
2077 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2078 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
2079 // CHECK3-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
2080 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2081 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
2082 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2083 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2084 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
2085 // CHECK3-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2086 // CHECK3-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
2087 // CHECK3-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2088 // CHECK3: omp_offload.failed:
2089 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
2090 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
2091 // CHECK3: omp_offload.cont:
2092 // CHECK3-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2093 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
2094 // CHECK3-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
2095 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 4
2096 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
2097 // CHECK3-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
2098 // CHECK3-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 4
2099 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
2100 // CHECK3-NEXT: store i8* null, i8** [[TMP13]], align 4
2101 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
2102 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
2103 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
2104 // CHECK3-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2105 // CHECK3-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
2106 // CHECK3-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
2107 // CHECK3: omp_offload.failed7:
2108 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40(%struct.SS* [[THIS1]]) #[[ATTR2]]
2109 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT8]]
2110 // CHECK3: omp_offload.cont8:
2111 // CHECK3-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2112 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
2113 // CHECK3-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
2114 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
2115 // CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
2116 // CHECK3-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
2117 // CHECK3-NEXT: store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 4
2118 // CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i32 0, i32 0
2119 // CHECK3-NEXT: store i8* null, i8** [[TMP22]], align 4
2120 // CHECK3-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
2121 // CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
2122 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
2123 // CHECK3-NEXT: [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2124 // CHECK3-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
2125 // CHECK3-NEXT: br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
2126 // CHECK3: omp_offload.failed14:
2127 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45(%struct.SS* [[THIS1]]) #[[ATTR2]]
2128 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT15]]
2129 // CHECK3: omp_offload.cont15:
2130 // CHECK3-NEXT: [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2131 // CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
2132 // CHECK3-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
2133 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 4
2134 // CHECK3-NEXT: [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
2135 // CHECK3-NEXT: [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
2136 // CHECK3-NEXT: store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 4
2137 // CHECK3-NEXT: [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i32 0, i32 0
2138 // CHECK3-NEXT: store i8* null, i8** [[TMP31]], align 4
2139 // CHECK3-NEXT: [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
2140 // CHECK3-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
2141 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
2142 // CHECK3-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2143 // CHECK3-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
2144 // CHECK3-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
2145 // CHECK3: omp_offload.failed21:
2146 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51(%struct.SS* [[THIS1]]) #[[ATTR2]]
2147 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT22]]
2148 // CHECK3: omp_offload.cont22:
2149 // CHECK3-NEXT: [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2150 // CHECK3-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
2151 // CHECK3-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
2152 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
2153 // CHECK3-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
2154 // CHECK3-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
2155 // CHECK3-NEXT: store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 4
2156 // CHECK3-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 0
2157 // CHECK3-NEXT: store i8* null, i8** [[TMP40]], align 4
2158 // CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
2159 // CHECK3-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
2160 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
2161 // CHECK3-NEXT: [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2162 // CHECK3-NEXT: [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
2163 // CHECK3-NEXT: br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
2164 // CHECK3: omp_offload.failed28:
2165 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57(%struct.SS* [[THIS1]]) #[[ATTR2]]
2166 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT29]]
2167 // CHECK3: omp_offload.cont29:
2168 // CHECK3-NEXT: [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2169 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i32 0, i32 0
2170 // CHECK3-NEXT: [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2171 // CHECK3-NEXT: ret i32 [[TMP45]]
2172 //
2173 //
2174 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35
2175 // CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
2176 // CHECK3-NEXT: entry:
2177 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2178 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2179 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2180 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
2181 // CHECK3-NEXT: ret void
2182 //
2183 //
2184 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
2185 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2186 // CHECK3-NEXT: entry:
2187 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2188 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2189 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2190 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2191 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
2192 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2193 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2194 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2195 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2196 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
2197 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2198 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2199 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2200 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2201 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2202 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
2203 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2204 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2205 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2206 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2207 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2208 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2209 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2210 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2211 // CHECK3: cond.true:
2212 // CHECK3-NEXT: br label [[COND_END:%.*]]
2213 // CHECK3: cond.false:
2214 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2215 // CHECK3-NEXT: br label [[COND_END]]
2216 // CHECK3: cond.end:
2217 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2218 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2219 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2220 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2221 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2222 // CHECK3: omp.inner.for.cond:
2223 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2224 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2225 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2226 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2227 // CHECK3: omp.inner.for.body:
2228 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2229 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2230 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
2231 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2232 // CHECK3: omp.inner.for.inc:
2233 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2234 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2235 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
2236 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2237 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
2238 // CHECK3: omp.inner.for.end:
2239 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2240 // CHECK3: omp.loop.exit:
2241 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2242 // CHECK3-NEXT: ret void
2243 //
2244 //
2245 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
2246 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2247 // CHECK3-NEXT: entry:
2248 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2249 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2250 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
2251 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
2252 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2253 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2254 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
2255 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2256 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2257 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2258 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2259 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
2260 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2261 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2262 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2263 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2264 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2265 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2266 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2267 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2268 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2269 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2270 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
2271 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
2272 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2273 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2274 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2275 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2276 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2277 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2278 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
2279 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2280 // CHECK3: cond.true:
2281 // CHECK3-NEXT: br label [[COND_END:%.*]]
2282 // CHECK3: cond.false:
2283 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2284 // CHECK3-NEXT: br label [[COND_END]]
2285 // CHECK3: cond.end:
2286 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2287 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2288 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2289 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
2290 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2291 // CHECK3: omp.inner.for.cond:
2292 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2293 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2294 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2295 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2296 // CHECK3: omp.inner.for.body:
2297 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2298 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2299 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2300 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2301 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2302 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
2303 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
2304 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2305 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2306 // CHECK3: omp.body.continue:
2307 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2308 // CHECK3: omp.inner.for.inc:
2309 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2310 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
2311 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2312 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
2313 // CHECK3: omp.inner.for.end:
2314 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2315 // CHECK3: omp.loop.exit:
2316 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
2317 // CHECK3-NEXT: ret void
2318 //
2319 //
2320 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40
2321 // CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2322 // CHECK3-NEXT: entry:
2323 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2324 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2325 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2326 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
2327 // CHECK3-NEXT: ret void
2328 //
2329 //
2330 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
2331 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2332 // CHECK3-NEXT: entry:
2333 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2334 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2335 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2336 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2337 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
2338 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2339 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2340 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2341 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2342 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
2343 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2344 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2345 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2346 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2347 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2348 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
2349 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2350 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2351 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2352 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2353 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2354 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2355 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2356 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2357 // CHECK3: cond.true:
2358 // CHECK3-NEXT: br label [[COND_END:%.*]]
2359 // CHECK3: cond.false:
2360 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2361 // CHECK3-NEXT: br label [[COND_END]]
2362 // CHECK3: cond.end:
2363 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2364 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2365 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2366 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2367 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2368 // CHECK3: omp.inner.for.cond:
2369 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2370 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2371 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2372 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2373 // CHECK3: omp.inner.for.body:
2374 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2375 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2376 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
2377 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2378 // CHECK3: omp.inner.for.inc:
2379 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2380 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2381 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
2382 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2383 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
2384 // CHECK3: omp.inner.for.end:
2385 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2386 // CHECK3: omp.loop.exit:
2387 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2388 // CHECK3-NEXT: ret void
2389 //
2390 //
2391 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
2392 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2393 // CHECK3-NEXT: entry:
2394 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2395 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2396 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
2397 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
2398 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2399 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2400 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
2401 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2402 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2403 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2404 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2405 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
2406 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2407 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2408 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2409 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2410 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2411 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2412 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2413 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2414 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2415 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2416 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
2417 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
2418 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2419 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2420 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2421 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2422 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2423 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2424 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
2425 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2426 // CHECK3: cond.true:
2427 // CHECK3-NEXT: br label [[COND_END:%.*]]
2428 // CHECK3: cond.false:
2429 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2430 // CHECK3-NEXT: br label [[COND_END]]
2431 // CHECK3: cond.end:
2432 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2433 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2434 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2435 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
2436 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2437 // CHECK3: omp.inner.for.cond:
2438 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2439 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2440 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2441 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2442 // CHECK3: omp.inner.for.body:
2443 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2444 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2445 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2446 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2447 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2448 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
2449 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
2450 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2451 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2452 // CHECK3: omp.body.continue:
2453 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2454 // CHECK3: omp.inner.for.inc:
2455 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2456 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
2457 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2458 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
2459 // CHECK3: omp.inner.for.end:
2460 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2461 // CHECK3: omp.loop.exit:
2462 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
2463 // CHECK3-NEXT: ret void
2464 //
2465 //
2466 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45
2467 // CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2468 // CHECK3-NEXT: entry:
2469 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2470 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2471 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2472 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
2473 // CHECK3-NEXT: ret void
2474 //
2475 //
2476 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..6
2477 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2478 // CHECK3-NEXT: entry:
2479 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2480 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2481 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2482 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2483 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
2484 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2485 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2486 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2487 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2488 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
2489 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2490 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2491 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2492 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2493 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2494 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
2495 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2496 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2497 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2498 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2499 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2500 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2501 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2502 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2503 // CHECK3: cond.true:
2504 // CHECK3-NEXT: br label [[COND_END:%.*]]
2505 // CHECK3: cond.false:
2506 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2507 // CHECK3-NEXT: br label [[COND_END]]
2508 // CHECK3: cond.end:
2509 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2510 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2511 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2512 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2513 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2514 // CHECK3: omp.inner.for.cond:
2515 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2516 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2517 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2518 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2519 // CHECK3: omp.inner.for.body:
2520 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2521 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2522 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
2523 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2524 // CHECK3: omp.inner.for.inc:
2525 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2526 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2527 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
2528 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2529 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
2530 // CHECK3: omp.inner.for.end:
2531 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2532 // CHECK3: omp.loop.exit:
2533 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2534 // CHECK3-NEXT: ret void
2535 //
2536 //
2537 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
2538 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2539 // CHECK3-NEXT: entry:
2540 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2541 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2542 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
2543 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
2544 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2545 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2546 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
2547 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2548 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2549 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2550 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2551 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
2552 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2553 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2554 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2555 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2556 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2557 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2558 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2559 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2560 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2561 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2562 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
2563 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
2564 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2565 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2566 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2567 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2568 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
2569 // CHECK3-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
2570 // CHECK3: omp.dispatch.cond:
2571 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2572 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2573 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
2574 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2575 // CHECK3: cond.true:
2576 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2577 // CHECK3-NEXT: br label [[COND_END:%.*]]
2578 // CHECK3: cond.false:
2579 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2580 // CHECK3-NEXT: br label [[COND_END]]
2581 // CHECK3: cond.end:
2582 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP7]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
2583 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2584 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2585 // CHECK3-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
2586 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2587 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2588 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
2589 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2590 // CHECK3: omp.dispatch.body:
2591 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2592 // CHECK3: omp.inner.for.cond:
2593 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2594 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2595 // CHECK3-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
2596 // CHECK3-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2597 // CHECK3: omp.inner.for.body:
2598 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2599 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
2600 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2601 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2602 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2603 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[I]], align 4
2604 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP15]]
2605 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2606 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2607 // CHECK3: omp.body.continue:
2608 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2609 // CHECK3: omp.inner.for.inc:
2610 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2611 // CHECK3-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1
2612 // CHECK3-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
2613 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
2614 // CHECK3: omp.inner.for.end:
2615 // CHECK3-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
2616 // CHECK3: omp.dispatch.inc:
2617 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2618 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2619 // CHECK3-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
2620 // CHECK3-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
2621 // CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2622 // CHECK3-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2623 // CHECK3-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
2624 // CHECK3-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
2625 // CHECK3-NEXT: br label [[OMP_DISPATCH_COND]]
2626 // CHECK3: omp.dispatch.end:
2627 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
2628 // CHECK3-NEXT: ret void
2629 //
2630 //
2631 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
2632 // CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2633 // CHECK3-NEXT: entry:
2634 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2635 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2636 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2637 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
2638 // CHECK3-NEXT: ret void
2639 //
2640 //
2641 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..10
2642 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2643 // CHECK3-NEXT: entry:
2644 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2645 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2646 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2647 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2648 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
2649 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2650 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2651 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2652 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2653 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
2654 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2655 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2656 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2657 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2658 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2659 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
2660 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2661 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2662 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2663 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2664 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2665 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2666 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2667 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2668 // CHECK3: cond.true:
2669 // CHECK3-NEXT: br label [[COND_END:%.*]]
2670 // CHECK3: cond.false:
2671 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2672 // CHECK3-NEXT: br label [[COND_END]]
2673 // CHECK3: cond.end:
2674 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2675 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2676 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2677 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2678 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2679 // CHECK3: omp.inner.for.cond:
2680 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2681 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2682 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2683 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2684 // CHECK3: omp.inner.for.body:
2685 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2686 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2687 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
2688 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2689 // CHECK3: omp.inner.for.inc:
2690 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2691 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2692 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
2693 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2694 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
2695 // CHECK3: omp.inner.for.end:
2696 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2697 // CHECK3: omp.loop.exit:
2698 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2699 // CHECK3-NEXT: ret void
2700 //
2701 //
2702 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
2703 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2704 // CHECK3-NEXT: entry:
2705 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2706 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2707 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
2708 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
2709 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2710 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2711 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
2712 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2713 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2714 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2715 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2716 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
2717 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2718 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2719 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2720 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2721 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2722 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2723 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2724 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2725 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2726 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2727 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
2728 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
2729 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2730 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2731 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2732 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2733 // CHECK3-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2734 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
2735 // CHECK3-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
2736 // CHECK3-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
2737 // CHECK3: omp.dispatch.cond:
2738 // CHECK3-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
2739 // CHECK3-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
2740 // CHECK3-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2741 // CHECK3: omp.dispatch.body:
2742 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2743 // CHECK3-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
2744 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2745 // CHECK3: omp.inner.for.cond:
2746 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
2747 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
2748 // CHECK3-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
2749 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2750 // CHECK3: omp.inner.for.body:
2751 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
2752 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
2753 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2754 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !11
2755 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2756 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !11
2757 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
2758 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !11
2759 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2760 // CHECK3: omp.body.continue:
2761 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2762 // CHECK3: omp.inner.for.inc:
2763 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
2764 // CHECK3-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
2765 // CHECK3-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
2766 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
2767 // CHECK3: omp.inner.for.end:
2768 // CHECK3-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
2769 // CHECK3: omp.dispatch.inc:
2770 // CHECK3-NEXT: br label [[OMP_DISPATCH_COND]]
2771 // CHECK3: omp.dispatch.end:
2772 // CHECK3-NEXT: ret void
2773 //
2774 //
2775 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57
2776 // CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2777 // CHECK3-NEXT: entry:
2778 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2779 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2780 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2781 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
2782 // CHECK3-NEXT: ret void
2783 //
2784 //
2785 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..14
2786 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2787 // CHECK3-NEXT: entry:
2788 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2789 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2790 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2791 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2792 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
2793 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2794 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2795 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2796 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2797 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
2798 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2799 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2800 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2801 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2802 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2803 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
2804 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2805 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2806 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2807 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2808 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2809 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2810 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2811 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2812 // CHECK3: cond.true:
2813 // CHECK3-NEXT: br label [[COND_END:%.*]]
2814 // CHECK3: cond.false:
2815 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2816 // CHECK3-NEXT: br label [[COND_END]]
2817 // CHECK3: cond.end:
2818 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2819 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2820 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2821 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2822 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2823 // CHECK3: omp.inner.for.cond:
2824 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2825 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2826 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2827 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2828 // CHECK3: omp.inner.for.body:
2829 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2830 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2831 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
2832 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2833 // CHECK3: omp.inner.for.inc:
2834 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2835 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2836 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
2837 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2838 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
2839 // CHECK3: omp.inner.for.end:
2840 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2841 // CHECK3: omp.loop.exit:
2842 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2843 // CHECK3-NEXT: ret void
2844 //
2845 //
2846 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..15
2847 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2848 // CHECK3-NEXT: entry:
2849 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2850 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2851 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
2852 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
2853 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2854 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2855 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
2856 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2857 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2858 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2859 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2860 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
2861 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2862 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2863 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2864 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2865 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2866 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2867 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2868 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2869 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2870 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2871 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
2872 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
2873 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2874 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2875 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2876 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2877 // CHECK3-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2878 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
2879 // CHECK3-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
2880 // CHECK3-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
2881 // CHECK3: omp.dispatch.cond:
2882 // CHECK3-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
2883 // CHECK3-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
2884 // CHECK3-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2885 // CHECK3: omp.dispatch.body:
2886 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2887 // CHECK3-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
2888 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2889 // CHECK3: omp.inner.for.cond:
2890 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
2891 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
2892 // CHECK3-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
2893 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2894 // CHECK3: omp.inner.for.body:
2895 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
2896 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
2897 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2898 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !14
2899 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2900 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !14
2901 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
2902 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !14
2903 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2904 // CHECK3: omp.body.continue:
2905 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2906 // CHECK3: omp.inner.for.inc:
2907 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
2908 // CHECK3-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
2909 // CHECK3-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
2910 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
2911 // CHECK3: omp.inner.for.end:
2912 // CHECK3-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
2913 // CHECK3: omp.dispatch.inc:
2914 // CHECK3-NEXT: br label [[OMP_DISPATCH_COND]]
2915 // CHECK3: omp.dispatch.end:
2916 // CHECK3-NEXT: ret void
2917 //
2918 //
2919 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2920 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
2921 // CHECK3-NEXT: entry:
2922 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
2923 // CHECK3-NEXT: ret void
2924 //
2925 //
2926 // CHECK4-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2927 // CHECK4-SAME: () #[[ATTR0:[0-9]+]] {
2928 // CHECK4-NEXT: entry:
2929 // CHECK4-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2930 // CHECK4-NEXT: [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
2931 // CHECK4-NEXT: ret i32 [[CALL]]
2932 //
2933 //
2934 // CHECK4-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2935 // CHECK4-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2936 // CHECK4-NEXT: entry:
2937 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2938 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
2939 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
2940 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
2941 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
2942 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
2943 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
2944 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
2945 // CHECK4-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
2946 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 4
2947 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 4
2948 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 4
2949 // CHECK4-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
2950 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 4
2951 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 4
2952 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 4
2953 // CHECK4-NEXT: [[_TMP20:%.*]] = alloca i32, align 4
2954 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 4
2955 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 4
2956 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 4
2957 // CHECK4-NEXT: [[_TMP27:%.*]] = alloca i32, align 4
2958 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2959 // CHECK4-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2960 // CHECK4-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
2961 // CHECK4-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2962 // CHECK4-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
2963 // CHECK4-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
2964 // CHECK4-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2965 // CHECK4-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
2966 // CHECK4-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
2967 // CHECK4-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2968 // CHECK4-NEXT: store i8* null, i8** [[TMP4]], align 4
2969 // CHECK4-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2970 // CHECK4-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2971 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
2972 // CHECK4-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2973 // CHECK4-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
2974 // CHECK4-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2975 // CHECK4: omp_offload.failed:
2976 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
2977 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT]]
2978 // CHECK4: omp_offload.cont:
2979 // CHECK4-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2980 // CHECK4-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
2981 // CHECK4-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
2982 // CHECK4-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 4
2983 // CHECK4-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
2984 // CHECK4-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
2985 // CHECK4-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 4
2986 // CHECK4-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
2987 // CHECK4-NEXT: store i8* null, i8** [[TMP13]], align 4
2988 // CHECK4-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
2989 // CHECK4-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
2990 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
2991 // CHECK4-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
2992 // CHECK4-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
2993 // CHECK4-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
2994 // CHECK4: omp_offload.failed7:
2995 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40(%struct.SS* [[THIS1]]) #[[ATTR2]]
2996 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT8]]
2997 // CHECK4: omp_offload.cont8:
2998 // CHECK4-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2999 // CHECK4-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
3000 // CHECK4-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
3001 // CHECK4-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
3002 // CHECK4-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
3003 // CHECK4-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
3004 // CHECK4-NEXT: store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 4
3005 // CHECK4-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i32 0, i32 0
3006 // CHECK4-NEXT: store i8* null, i8** [[TMP22]], align 4
3007 // CHECK4-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
3008 // CHECK4-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
3009 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
3010 // CHECK4-NEXT: [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3011 // CHECK4-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
3012 // CHECK4-NEXT: br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
3013 // CHECK4: omp_offload.failed14:
3014 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45(%struct.SS* [[THIS1]]) #[[ATTR2]]
3015 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT15]]
3016 // CHECK4: omp_offload.cont15:
3017 // CHECK4-NEXT: [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3018 // CHECK4-NEXT: [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
3019 // CHECK4-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
3020 // CHECK4-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 4
3021 // CHECK4-NEXT: [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
3022 // CHECK4-NEXT: [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
3023 // CHECK4-NEXT: store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 4
3024 // CHECK4-NEXT: [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i32 0, i32 0
3025 // CHECK4-NEXT: store i8* null, i8** [[TMP31]], align 4
3026 // CHECK4-NEXT: [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
3027 // CHECK4-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
3028 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
3029 // CHECK4-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3030 // CHECK4-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
3031 // CHECK4-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
3032 // CHECK4: omp_offload.failed21:
3033 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51(%struct.SS* [[THIS1]]) #[[ATTR2]]
3034 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT22]]
3035 // CHECK4: omp_offload.cont22:
3036 // CHECK4-NEXT: [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3037 // CHECK4-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
3038 // CHECK4-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
3039 // CHECK4-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
3040 // CHECK4-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
3041 // CHECK4-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
3042 // CHECK4-NEXT: store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 4
3043 // CHECK4-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 0
3044 // CHECK4-NEXT: store i8* null, i8** [[TMP40]], align 4
3045 // CHECK4-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
3046 // CHECK4-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
3047 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
3048 // CHECK4-NEXT: [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3049 // CHECK4-NEXT: [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
3050 // CHECK4-NEXT: br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
3051 // CHECK4: omp_offload.failed28:
3052 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57(%struct.SS* [[THIS1]]) #[[ATTR2]]
3053 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT29]]
3054 // CHECK4: omp_offload.cont29:
3055 // CHECK4-NEXT: [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3056 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i32 0, i32 0
3057 // CHECK4-NEXT: [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
3058 // CHECK4-NEXT: ret i32 [[TMP45]]
3059 //
3060 //
3061 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35
3062 // CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
3063 // CHECK4-NEXT: entry:
3064 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3065 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3066 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3067 // CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
3068 // CHECK4-NEXT: ret void
3069 //
3070 //
3071 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
3072 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3073 // CHECK4-NEXT: entry:
3074 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3075 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3076 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3077 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3078 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
3079 // CHECK4-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3080 // CHECK4-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3081 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3082 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3083 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
3084 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3085 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3086 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3087 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3088 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3089 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
3090 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3091 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3092 // CHECK4-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3093 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3094 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3095 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3096 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
3097 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3098 // CHECK4: cond.true:
3099 // CHECK4-NEXT: br label [[COND_END:%.*]]
3100 // CHECK4: cond.false:
3101 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3102 // CHECK4-NEXT: br label [[COND_END]]
3103 // CHECK4: cond.end:
3104 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3105 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3106 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3107 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3108 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3109 // CHECK4: omp.inner.for.cond:
3110 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3111 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3112 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3113 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3114 // CHECK4: omp.inner.for.body:
3115 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3116 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3117 // CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
3118 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3119 // CHECK4: omp.inner.for.inc:
3120 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3121 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3122 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
3123 // CHECK4-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3124 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]]
3125 // CHECK4: omp.inner.for.end:
3126 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3127 // CHECK4: omp.loop.exit:
3128 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3129 // CHECK4-NEXT: ret void
3130 //
3131 //
3132 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
3133 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3134 // CHECK4-NEXT: entry:
3135 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3136 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3137 // CHECK4-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
3138 // CHECK4-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
3139 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3140 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3141 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
3142 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3143 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3144 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3145 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3146 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
3147 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3148 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3149 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3150 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3151 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3152 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3153 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3154 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
3155 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3156 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3157 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
3158 // CHECK4-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
3159 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3160 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3161 // CHECK4-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3162 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
3163 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3164 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3165 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
3166 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3167 // CHECK4: cond.true:
3168 // CHECK4-NEXT: br label [[COND_END:%.*]]
3169 // CHECK4: cond.false:
3170 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3171 // CHECK4-NEXT: br label [[COND_END]]
3172 // CHECK4: cond.end:
3173 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
3174 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3175 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3176 // CHECK4-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
3177 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3178 // CHECK4: omp.inner.for.cond:
3179 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3180 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3181 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
3182 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3183 // CHECK4: omp.inner.for.body:
3184 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3185 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
3186 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3187 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3188 // CHECK4-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
3189 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
3190 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
3191 // CHECK4-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3192 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3193 // CHECK4: omp.body.continue:
3194 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3195 // CHECK4: omp.inner.for.inc:
3196 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3197 // CHECK4-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
3198 // CHECK4-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
3199 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]]
3200 // CHECK4: omp.inner.for.end:
3201 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3202 // CHECK4: omp.loop.exit:
3203 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
3204 // CHECK4-NEXT: ret void
3205 //
3206 //
3207 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40
3208 // CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3209 // CHECK4-NEXT: entry:
3210 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3211 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3212 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3213 // CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
3214 // CHECK4-NEXT: ret void
3215 //
3216 //
3217 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
3218 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3219 // CHECK4-NEXT: entry:
3220 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3221 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3222 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3223 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3224 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
3225 // CHECK4-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3226 // CHECK4-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3227 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3228 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3229 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
3230 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3231 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3232 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3233 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3234 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3235 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
3236 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3237 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3238 // CHECK4-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3239 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3240 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3241 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3242 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
3243 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3244 // CHECK4: cond.true:
3245 // CHECK4-NEXT: br label [[COND_END:%.*]]
3246 // CHECK4: cond.false:
3247 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3248 // CHECK4-NEXT: br label [[COND_END]]
3249 // CHECK4: cond.end:
3250 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3251 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3252 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3253 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3254 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3255 // CHECK4: omp.inner.for.cond:
3256 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3257 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3258 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3259 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3260 // CHECK4: omp.inner.for.body:
3261 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3262 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3263 // CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
3264 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3265 // CHECK4: omp.inner.for.inc:
3266 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3267 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3268 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
3269 // CHECK4-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3270 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]]
3271 // CHECK4: omp.inner.for.end:
3272 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3273 // CHECK4: omp.loop.exit:
3274 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3275 // CHECK4-NEXT: ret void
3276 //
3277 //
3278 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3
3279 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3280 // CHECK4-NEXT: entry:
3281 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3282 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3283 // CHECK4-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
3284 // CHECK4-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
3285 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3286 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3287 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
3288 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3289 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3290 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3291 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3292 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
3293 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3294 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3295 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3296 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3297 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3298 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3299 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3300 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
3301 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3302 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3303 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
3304 // CHECK4-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
3305 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3306 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3307 // CHECK4-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3308 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
3309 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3310 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3311 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
3312 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3313 // CHECK4: cond.true:
3314 // CHECK4-NEXT: br label [[COND_END:%.*]]
3315 // CHECK4: cond.false:
3316 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3317 // CHECK4-NEXT: br label [[COND_END]]
3318 // CHECK4: cond.end:
3319 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
3320 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3321 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3322 // CHECK4-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
3323 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3324 // CHECK4: omp.inner.for.cond:
3325 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3326 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3327 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
3328 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3329 // CHECK4: omp.inner.for.body:
3330 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3331 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
3332 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3333 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3334 // CHECK4-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
3335 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
3336 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
3337 // CHECK4-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3338 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3339 // CHECK4: omp.body.continue:
3340 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3341 // CHECK4: omp.inner.for.inc:
3342 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3343 // CHECK4-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
3344 // CHECK4-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
3345 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]]
3346 // CHECK4: omp.inner.for.end:
3347 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3348 // CHECK4: omp.loop.exit:
3349 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
3350 // CHECK4-NEXT: ret void
3351 //
3352 //
3353 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45
3354 // CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3355 // CHECK4-NEXT: entry:
3356 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3357 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3358 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3359 // CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
3360 // CHECK4-NEXT: ret void
3361 //
3362 //
3363 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..6
3364 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3365 // CHECK4-NEXT: entry:
3366 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3367 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3368 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3369 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3370 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
3371 // CHECK4-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3372 // CHECK4-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3373 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3374 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3375 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
3376 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3377 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3378 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3379 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3380 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3381 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
3382 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3383 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3384 // CHECK4-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3385 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3386 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3387 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3388 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
3389 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3390 // CHECK4: cond.true:
3391 // CHECK4-NEXT: br label [[COND_END:%.*]]
3392 // CHECK4: cond.false:
3393 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3394 // CHECK4-NEXT: br label [[COND_END]]
3395 // CHECK4: cond.end:
3396 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3397 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3398 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3399 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3400 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3401 // CHECK4: omp.inner.for.cond:
3402 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3403 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3404 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3405 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3406 // CHECK4: omp.inner.for.body:
3407 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3408 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3409 // CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
3410 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3411 // CHECK4: omp.inner.for.inc:
3412 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3413 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3414 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
3415 // CHECK4-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3416 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]]
3417 // CHECK4: omp.inner.for.end:
3418 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3419 // CHECK4: omp.loop.exit:
3420 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3421 // CHECK4-NEXT: ret void
3422 //
3423 //
3424 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..7
3425 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3426 // CHECK4-NEXT: entry:
3427 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3428 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3429 // CHECK4-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
3430 // CHECK4-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
3431 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3432 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3433 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
3434 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3435 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3436 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3437 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3438 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
3439 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3440 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3441 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3442 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3443 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3444 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3445 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3446 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
3447 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3448 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3449 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
3450 // CHECK4-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
3451 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3452 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3453 // CHECK4-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3454 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
3455 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
3456 // CHECK4-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
3457 // CHECK4: omp.dispatch.cond:
3458 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3459 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3460 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
3461 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3462 // CHECK4: cond.true:
3463 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3464 // CHECK4-NEXT: br label [[COND_END:%.*]]
3465 // CHECK4: cond.false:
3466 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3467 // CHECK4-NEXT: br label [[COND_END]]
3468 // CHECK4: cond.end:
3469 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ [[TMP7]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
3470 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3471 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3472 // CHECK4-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
3473 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3474 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3475 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
3476 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
3477 // CHECK4: omp.dispatch.body:
3478 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3479 // CHECK4: omp.inner.for.cond:
3480 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3481 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3482 // CHECK4-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
3483 // CHECK4-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3484 // CHECK4: omp.inner.for.body:
3485 // CHECK4-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3486 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
3487 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3488 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3489 // CHECK4-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
3490 // CHECK4-NEXT: [[TMP15:%.*]] = load i32, i32* [[I]], align 4
3491 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP15]]
3492 // CHECK4-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3493 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3494 // CHECK4: omp.body.continue:
3495 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3496 // CHECK4: omp.inner.for.inc:
3497 // CHECK4-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3498 // CHECK4-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1
3499 // CHECK4-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
3500 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]]
3501 // CHECK4: omp.inner.for.end:
3502 // CHECK4-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
3503 // CHECK4: omp.dispatch.inc:
3504 // CHECK4-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3505 // CHECK4-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3506 // CHECK4-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
3507 // CHECK4-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
3508 // CHECK4-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3509 // CHECK4-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3510 // CHECK4-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
3511 // CHECK4-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
3512 // CHECK4-NEXT: br label [[OMP_DISPATCH_COND]]
3513 // CHECK4: omp.dispatch.end:
3514 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
3515 // CHECK4-NEXT: ret void
3516 //
3517 //
3518 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
3519 // CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3520 // CHECK4-NEXT: entry:
3521 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3522 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3523 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3524 // CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
3525 // CHECK4-NEXT: ret void
3526 //
3527 //
3528 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..10
3529 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3530 // CHECK4-NEXT: entry:
3531 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3532 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3533 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3534 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3535 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
3536 // CHECK4-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3537 // CHECK4-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3538 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3539 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3540 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
3541 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3542 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3543 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3544 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3545 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3546 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
3547 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3548 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3549 // CHECK4-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3550 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3551 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3552 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3553 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
3554 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3555 // CHECK4: cond.true:
3556 // CHECK4-NEXT: br label [[COND_END:%.*]]
3557 // CHECK4: cond.false:
3558 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3559 // CHECK4-NEXT: br label [[COND_END]]
3560 // CHECK4: cond.end:
3561 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3562 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3563 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3564 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3565 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3566 // CHECK4: omp.inner.for.cond:
3567 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3568 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3569 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3570 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3571 // CHECK4: omp.inner.for.body:
3572 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3573 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3574 // CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
3575 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3576 // CHECK4: omp.inner.for.inc:
3577 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3578 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3579 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
3580 // CHECK4-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3581 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]]
3582 // CHECK4: omp.inner.for.end:
3583 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3584 // CHECK4: omp.loop.exit:
3585 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3586 // CHECK4-NEXT: ret void
3587 //
3588 //
3589 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..11
3590 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3591 // CHECK4-NEXT: entry:
3592 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3593 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3594 // CHECK4-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
3595 // CHECK4-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
3596 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3597 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3598 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
3599 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3600 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3601 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3602 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3603 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
3604 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3605 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3606 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3607 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3608 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3609 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3610 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3611 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
3612 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3613 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3614 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
3615 // CHECK4-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
3616 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3617 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3618 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3619 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3620 // CHECK4-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3621 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
3622 // CHECK4-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
3623 // CHECK4-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
3624 // CHECK4: omp.dispatch.cond:
3625 // CHECK4-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
3626 // CHECK4-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
3627 // CHECK4-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
3628 // CHECK4: omp.dispatch.body:
3629 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3630 // CHECK4-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
3631 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3632 // CHECK4: omp.inner.for.cond:
3633 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
3634 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
3635 // CHECK4-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
3636 // CHECK4-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3637 // CHECK4: omp.inner.for.body:
3638 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
3639 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
3640 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3641 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !11
3642 // CHECK4-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
3643 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !11
3644 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
3645 // CHECK4-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !11
3646 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3647 // CHECK4: omp.body.continue:
3648 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3649 // CHECK4: omp.inner.for.inc:
3650 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
3651 // CHECK4-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
3652 // CHECK4-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
3653 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
3654 // CHECK4: omp.inner.for.end:
3655 // CHECK4-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
3656 // CHECK4: omp.dispatch.inc:
3657 // CHECK4-NEXT: br label [[OMP_DISPATCH_COND]]
3658 // CHECK4: omp.dispatch.end:
3659 // CHECK4-NEXT: ret void
3660 //
3661 //
3662 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57
3663 // CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3664 // CHECK4-NEXT: entry:
3665 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3666 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3667 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3668 // CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
3669 // CHECK4-NEXT: ret void
3670 //
3671 //
3672 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..14
3673 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3674 // CHECK4-NEXT: entry:
3675 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3676 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3677 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3678 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3679 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
3680 // CHECK4-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3681 // CHECK4-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3682 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3683 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3684 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
3685 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3686 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3687 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3688 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3689 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3690 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
3691 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3692 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3693 // CHECK4-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3694 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3695 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3696 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3697 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
3698 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3699 // CHECK4: cond.true:
3700 // CHECK4-NEXT: br label [[COND_END:%.*]]
3701 // CHECK4: cond.false:
3702 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3703 // CHECK4-NEXT: br label [[COND_END]]
3704 // CHECK4: cond.end:
3705 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3706 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3707 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3708 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3709 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3710 // CHECK4: omp.inner.for.cond:
3711 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3712 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3713 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3714 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3715 // CHECK4: omp.inner.for.body:
3716 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3717 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3718 // CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
3719 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3720 // CHECK4: omp.inner.for.inc:
3721 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3722 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3723 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
3724 // CHECK4-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3725 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]]
3726 // CHECK4: omp.inner.for.end:
3727 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3728 // CHECK4: omp.loop.exit:
3729 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3730 // CHECK4-NEXT: ret void
3731 //
3732 //
3733 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..15
3734 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3735 // CHECK4-NEXT: entry:
3736 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3737 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3738 // CHECK4-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
3739 // CHECK4-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
3740 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
3741 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3742 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
3743 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3744 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3745 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3746 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3747 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
3748 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3749 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3750 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3751 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3752 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
3753 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
3754 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3755 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
3756 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3757 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3758 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
3759 // CHECK4-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
3760 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3761 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3762 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3763 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3764 // CHECK4-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3765 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
3766 // CHECK4-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
3767 // CHECK4-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
3768 // CHECK4: omp.dispatch.cond:
3769 // CHECK4-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
3770 // CHECK4-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
3771 // CHECK4-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
3772 // CHECK4: omp.dispatch.body:
3773 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3774 // CHECK4-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
3775 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3776 // CHECK4: omp.inner.for.cond:
3777 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3778 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
3779 // CHECK4-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
3780 // CHECK4-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3781 // CHECK4: omp.inner.for.body:
3782 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3783 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
3784 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3785 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !14
3786 // CHECK4-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
3787 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !14
3788 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
3789 // CHECK4-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !14
3790 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3791 // CHECK4: omp.body.continue:
3792 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3793 // CHECK4: omp.inner.for.inc:
3794 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3795 // CHECK4-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
3796 // CHECK4-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3797 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
3798 // CHECK4: omp.inner.for.end:
3799 // CHECK4-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
3800 // CHECK4: omp.dispatch.inc:
3801 // CHECK4-NEXT: br label [[OMP_DISPATCH_COND]]
3802 // CHECK4: omp.dispatch.end:
3803 // CHECK4-NEXT: ret void
3804 //
3805 //
3806 // CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3807 // CHECK4-SAME: () #[[ATTR3:[0-9]+]] {
3808 // CHECK4-NEXT: entry:
3809 // CHECK4-NEXT: call void @__tgt_register_requires(i64 1)
3810 // CHECK4-NEXT: ret void
3811 //
3812 //
3813 // CHECK5-LABEL: define {{[^@]+}}@_Z21teams_template_structv
3814 // CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
3815 // CHECK5-NEXT: entry:
3816 // CHECK5-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
3817 // CHECK5-NEXT: [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
3818 // CHECK5-NEXT: ret i32 [[CALL]]
3819 //
3820 //
3821 // CHECK5-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
3822 // CHECK5-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
3823 // CHECK5-NEXT: entry:
3824 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
3825 // CHECK5-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
3826 // CHECK5-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
3827 // CHECK5-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
3828 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3829 // CHECK5-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
3830 // CHECK5-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
3831 // CHECK5-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
3832 // CHECK5-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
3833 // CHECK5-NEXT: [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 8
3834 // CHECK5-NEXT: [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 8
3835 // CHECK5-NEXT: [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 8
3836 // CHECK5-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
3837 // CHECK5-NEXT: [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 8
3838 // CHECK5-NEXT: [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 8
3839 // CHECK5-NEXT: [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 8
3840 // CHECK5-NEXT: [[_TMP20:%.*]] = alloca i32, align 4
3841 // CHECK5-NEXT: [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 8
3842 // CHECK5-NEXT: [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 8
3843 // CHECK5-NEXT: [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 8
3844 // CHECK5-NEXT: [[_TMP27:%.*]] = alloca i32, align 4
3845 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
3846 // CHECK5-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
3847 // CHECK5-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
3848 // CHECK5-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3849 // CHECK5-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
3850 // CHECK5-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
3851 // CHECK5-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3852 // CHECK5-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
3853 // CHECK5-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
3854 // CHECK5-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3855 // CHECK5-NEXT: store i8* null, i8** [[TMP4]], align 8
3856 // CHECK5-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3857 // CHECK5-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3858 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
3859 // CHECK5-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3860 // CHECK5-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
3861 // CHECK5-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3862 // CHECK5: omp_offload.failed:
3863 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
3864 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT]]
3865 // CHECK5: omp_offload.cont:
3866 // CHECK5-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3867 // CHECK5-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
3868 // CHECK5-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
3869 // CHECK5-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 8
3870 // CHECK5-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
3871 // CHECK5-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
3872 // CHECK5-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 8
3873 // CHECK5-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
3874 // CHECK5-NEXT: store i8* null, i8** [[TMP13]], align 8
3875 // CHECK5-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
3876 // CHECK5-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
3877 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
3878 // CHECK5-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3879 // CHECK5-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
3880 // CHECK5-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
3881 // CHECK5: omp_offload.failed7:
3882 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40(%struct.SS* [[THIS1]]) #[[ATTR2]]
3883 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT8]]
3884 // CHECK5: omp_offload.cont8:
3885 // CHECK5-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3886 // CHECK5-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
3887 // CHECK5-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
3888 // CHECK5-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
3889 // CHECK5-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
3890 // CHECK5-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
3891 // CHECK5-NEXT: store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 8
3892 // CHECK5-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 0
3893 // CHECK5-NEXT: store i8* null, i8** [[TMP22]], align 8
3894 // CHECK5-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
3895 // CHECK5-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
3896 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
3897 // CHECK5-NEXT: [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3898 // CHECK5-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
3899 // CHECK5-NEXT: br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
3900 // CHECK5: omp_offload.failed14:
3901 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45(%struct.SS* [[THIS1]]) #[[ATTR2]]
3902 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT15]]
3903 // CHECK5: omp_offload.cont15:
3904 // CHECK5-NEXT: [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3905 // CHECK5-NEXT: [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
3906 // CHECK5-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
3907 // CHECK5-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 8
3908 // CHECK5-NEXT: [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
3909 // CHECK5-NEXT: [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
3910 // CHECK5-NEXT: store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 8
3911 // CHECK5-NEXT: [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i64 0, i64 0
3912 // CHECK5-NEXT: store i8* null, i8** [[TMP31]], align 8
3913 // CHECK5-NEXT: [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
3914 // CHECK5-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
3915 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
3916 // CHECK5-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3917 // CHECK5-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
3918 // CHECK5-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
3919 // CHECK5: omp_offload.failed21:
3920 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51(%struct.SS* [[THIS1]]) #[[ATTR2]]
3921 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT22]]
3922 // CHECK5: omp_offload.cont22:
3923 // CHECK5-NEXT: [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3924 // CHECK5-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
3925 // CHECK5-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
3926 // CHECK5-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
3927 // CHECK5-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
3928 // CHECK5-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
3929 // CHECK5-NEXT: store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 8
3930 // CHECK5-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 0
3931 // CHECK5-NEXT: store i8* null, i8** [[TMP40]], align 8
3932 // CHECK5-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
3933 // CHECK5-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
3934 // CHECK5-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
3935 // CHECK5-NEXT: [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
3936 // CHECK5-NEXT: [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
3937 // CHECK5-NEXT: br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
3938 // CHECK5: omp_offload.failed28:
3939 // CHECK5-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57(%struct.SS* [[THIS1]]) #[[ATTR2]]
3940 // CHECK5-NEXT: br label [[OMP_OFFLOAD_CONT29]]
3941 // CHECK5: omp_offload.cont29:
3942 // CHECK5-NEXT: [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3943 // CHECK5-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 0
3944 // CHECK5-NEXT: [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
3945 // CHECK5-NEXT: ret i32 [[TMP45]]
3946 //
3947 //
3948 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35
3949 // CHECK5-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
3950 // CHECK5-NEXT: entry:
3951 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
3952 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
3953 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
3954 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
3955 // CHECK5-NEXT: ret void
3956 //
3957 //
3958 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined.
3959 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
3960 // CHECK5-NEXT: entry:
3961 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3962 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3963 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
3964 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3965 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
3966 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3967 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3968 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3969 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3970 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
3971 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3972 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3973 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
3974 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
3975 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3976 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
3977 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3978 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3979 // CHECK5-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3980 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3981 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3982 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3983 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
3984 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3985 // CHECK5: cond.true:
3986 // CHECK5-NEXT: br label [[COND_END:%.*]]
3987 // CHECK5: cond.false:
3988 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3989 // CHECK5-NEXT: br label [[COND_END]]
3990 // CHECK5: cond.end:
3991 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3992 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3993 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3994 // CHECK5-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3995 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3996 // CHECK5: omp.inner.for.cond:
3997 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3998 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3999 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4000 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4001 // CHECK5: omp.inner.for.body:
4002 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4003 // CHECK5-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
4004 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4005 // CHECK5-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
4006 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
4007 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4008 // CHECK5: omp.inner.for.inc:
4009 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4010 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4011 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4012 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4013 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
4014 // CHECK5: omp.inner.for.end:
4015 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4016 // CHECK5: omp.loop.exit:
4017 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4018 // CHECK5-NEXT: ret void
4019 //
4020 //
4021 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..1
4022 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4023 // CHECK5-NEXT: entry:
4024 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4025 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4026 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4027 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4028 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4029 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4030 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
4031 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4032 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4033 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4034 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4035 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
4036 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4037 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4038 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4039 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4040 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4041 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4042 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4043 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
4044 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4045 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
4046 // CHECK5-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4047 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
4048 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4049 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4050 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4051 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4052 // CHECK5-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4053 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4054 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4055 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4056 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
4057 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4058 // CHECK5: cond.true:
4059 // CHECK5-NEXT: br label [[COND_END:%.*]]
4060 // CHECK5: cond.false:
4061 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4062 // CHECK5-NEXT: br label [[COND_END]]
4063 // CHECK5: cond.end:
4064 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
4065 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4066 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4067 // CHECK5-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
4068 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4069 // CHECK5: omp.inner.for.cond:
4070 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4071 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4072 // CHECK5-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
4073 // CHECK5-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4074 // CHECK5: omp.inner.for.body:
4075 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4076 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
4077 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4078 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4079 // CHECK5-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
4080 // CHECK5-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
4081 // CHECK5-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
4082 // CHECK5-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
4083 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4084 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4085 // CHECK5: omp.body.continue:
4086 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4087 // CHECK5: omp.inner.for.inc:
4088 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4089 // CHECK5-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
4090 // CHECK5-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
4091 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
4092 // CHECK5: omp.inner.for.end:
4093 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4094 // CHECK5: omp.loop.exit:
4095 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
4096 // CHECK5-NEXT: ret void
4097 //
4098 //
4099 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40
4100 // CHECK5-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4101 // CHECK5-NEXT: entry:
4102 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4103 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4104 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4105 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
4106 // CHECK5-NEXT: ret void
4107 //
4108 //
4109 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..2
4110 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4111 // CHECK5-NEXT: entry:
4112 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4113 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4114 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4115 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4116 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
4117 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4118 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4119 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4120 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4121 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
4122 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4123 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4124 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4125 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4126 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4127 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
4128 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4129 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4130 // CHECK5-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4131 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4132 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4133 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4134 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
4135 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4136 // CHECK5: cond.true:
4137 // CHECK5-NEXT: br label [[COND_END:%.*]]
4138 // CHECK5: cond.false:
4139 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4140 // CHECK5-NEXT: br label [[COND_END]]
4141 // CHECK5: cond.end:
4142 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4143 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4144 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4145 // CHECK5-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4146 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4147 // CHECK5: omp.inner.for.cond:
4148 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4149 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4150 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4151 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4152 // CHECK5: omp.inner.for.body:
4153 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4154 // CHECK5-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
4155 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4156 // CHECK5-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
4157 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
4158 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4159 // CHECK5: omp.inner.for.inc:
4160 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4161 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4162 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4163 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4164 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
4165 // CHECK5: omp.inner.for.end:
4166 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4167 // CHECK5: omp.loop.exit:
4168 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4169 // CHECK5-NEXT: ret void
4170 //
4171 //
4172 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..3
4173 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4174 // CHECK5-NEXT: entry:
4175 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4176 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4177 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4178 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4179 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4180 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4181 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
4182 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4183 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4184 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4185 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4186 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
4187 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4188 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4189 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4190 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4191 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4192 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4193 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4194 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
4195 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4196 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
4197 // CHECK5-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4198 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
4199 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4200 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4201 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4202 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4203 // CHECK5-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4204 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4205 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4206 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4207 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
4208 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4209 // CHECK5: cond.true:
4210 // CHECK5-NEXT: br label [[COND_END:%.*]]
4211 // CHECK5: cond.false:
4212 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4213 // CHECK5-NEXT: br label [[COND_END]]
4214 // CHECK5: cond.end:
4215 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
4216 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4217 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4218 // CHECK5-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
4219 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4220 // CHECK5: omp.inner.for.cond:
4221 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4222 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4223 // CHECK5-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
4224 // CHECK5-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4225 // CHECK5: omp.inner.for.body:
4226 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4227 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
4228 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4229 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4230 // CHECK5-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
4231 // CHECK5-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
4232 // CHECK5-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
4233 // CHECK5-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
4234 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4235 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4236 // CHECK5: omp.body.continue:
4237 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4238 // CHECK5: omp.inner.for.inc:
4239 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4240 // CHECK5-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
4241 // CHECK5-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
4242 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
4243 // CHECK5: omp.inner.for.end:
4244 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4245 // CHECK5: omp.loop.exit:
4246 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
4247 // CHECK5-NEXT: ret void
4248 //
4249 //
4250 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45
4251 // CHECK5-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4252 // CHECK5-NEXT: entry:
4253 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4254 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4255 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4256 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
4257 // CHECK5-NEXT: ret void
4258 //
4259 //
4260 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..6
4261 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4262 // CHECK5-NEXT: entry:
4263 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4264 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4265 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4266 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4267 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
4268 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4269 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4270 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4271 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4272 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
4273 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4274 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4275 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4276 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4277 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4278 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
4279 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4280 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4281 // CHECK5-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4282 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4283 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4284 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4285 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
4286 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4287 // CHECK5: cond.true:
4288 // CHECK5-NEXT: br label [[COND_END:%.*]]
4289 // CHECK5: cond.false:
4290 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4291 // CHECK5-NEXT: br label [[COND_END]]
4292 // CHECK5: cond.end:
4293 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4294 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4295 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4296 // CHECK5-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4297 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4298 // CHECK5: omp.inner.for.cond:
4299 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4300 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4301 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4302 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4303 // CHECK5: omp.inner.for.body:
4304 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4305 // CHECK5-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
4306 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4307 // CHECK5-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
4308 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
4309 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4310 // CHECK5: omp.inner.for.inc:
4311 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4312 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4313 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4314 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4315 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
4316 // CHECK5: omp.inner.for.end:
4317 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4318 // CHECK5: omp.loop.exit:
4319 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4320 // CHECK5-NEXT: ret void
4321 //
4322 //
4323 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..7
4324 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4325 // CHECK5-NEXT: entry:
4326 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4327 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4328 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4329 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4330 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4331 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4332 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
4333 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4334 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4335 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4336 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4337 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
4338 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4339 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4340 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4341 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4342 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4343 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4344 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4345 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
4346 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4347 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
4348 // CHECK5-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4349 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
4350 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4351 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4352 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4353 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4354 // CHECK5-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4355 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4356 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
4357 // CHECK5-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
4358 // CHECK5: omp.dispatch.cond:
4359 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4360 // CHECK5-NEXT: [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4361 // CHECK5-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
4362 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
4363 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4364 // CHECK5: cond.true:
4365 // CHECK5-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4366 // CHECK5-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
4367 // CHECK5-NEXT: br label [[COND_END:%.*]]
4368 // CHECK5: cond.false:
4369 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4370 // CHECK5-NEXT: br label [[COND_END]]
4371 // CHECK5: cond.end:
4372 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
4373 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4374 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4375 // CHECK5-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
4376 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4377 // CHECK5-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4378 // CHECK5-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
4379 // CHECK5-NEXT: br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
4380 // CHECK5: omp.dispatch.body:
4381 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4382 // CHECK5: omp.inner.for.cond:
4383 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4384 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4385 // CHECK5-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
4386 // CHECK5-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4387 // CHECK5: omp.inner.for.body:
4388 // CHECK5-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4389 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
4390 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4391 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4392 // CHECK5-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
4393 // CHECK5-NEXT: [[TMP15:%.*]] = load i32, i32* [[I]], align 4
4394 // CHECK5-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
4395 // CHECK5-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
4396 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4397 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4398 // CHECK5: omp.body.continue:
4399 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4400 // CHECK5: omp.inner.for.inc:
4401 // CHECK5-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4402 // CHECK5-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
4403 // CHECK5-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
4404 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
4405 // CHECK5: omp.inner.for.end:
4406 // CHECK5-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
4407 // CHECK5: omp.dispatch.inc:
4408 // CHECK5-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4409 // CHECK5-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4410 // CHECK5-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
4411 // CHECK5-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
4412 // CHECK5-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4413 // CHECK5-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4414 // CHECK5-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
4415 // CHECK5-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
4416 // CHECK5-NEXT: br label [[OMP_DISPATCH_COND]]
4417 // CHECK5: omp.dispatch.end:
4418 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
4419 // CHECK5-NEXT: ret void
4420 //
4421 //
4422 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
4423 // CHECK5-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4424 // CHECK5-NEXT: entry:
4425 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4426 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4427 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4428 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
4429 // CHECK5-NEXT: ret void
4430 //
4431 //
4432 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..10
4433 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4434 // CHECK5-NEXT: entry:
4435 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4436 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4437 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4438 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4439 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
4440 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4441 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4442 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4443 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4444 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
4445 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4446 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4447 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4448 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4449 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4450 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
4451 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4452 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4453 // CHECK5-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4454 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4455 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4456 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4457 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
4458 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4459 // CHECK5: cond.true:
4460 // CHECK5-NEXT: br label [[COND_END:%.*]]
4461 // CHECK5: cond.false:
4462 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4463 // CHECK5-NEXT: br label [[COND_END]]
4464 // CHECK5: cond.end:
4465 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4466 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4467 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4468 // CHECK5-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4469 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4470 // CHECK5: omp.inner.for.cond:
4471 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4472 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4473 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4474 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4475 // CHECK5: omp.inner.for.body:
4476 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4477 // CHECK5-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
4478 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4479 // CHECK5-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
4480 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
4481 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4482 // CHECK5: omp.inner.for.inc:
4483 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4484 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4485 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4486 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4487 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
4488 // CHECK5: omp.inner.for.end:
4489 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4490 // CHECK5: omp.loop.exit:
4491 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4492 // CHECK5-NEXT: ret void
4493 //
4494 //
4495 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..11
4496 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4497 // CHECK5-NEXT: entry:
4498 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4499 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4500 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4501 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4502 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4503 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4504 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
4505 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4506 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4507 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4508 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4509 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
4510 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4511 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4512 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4513 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4514 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4515 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4516 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4517 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
4518 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4519 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
4520 // CHECK5-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4521 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
4522 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4523 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4524 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4525 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4526 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4527 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4528 // CHECK5-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4529 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
4530 // CHECK5-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
4531 // CHECK5-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
4532 // CHECK5: omp.dispatch.cond:
4533 // CHECK5-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
4534 // CHECK5-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
4535 // CHECK5-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
4536 // CHECK5: omp.dispatch.body:
4537 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4538 // CHECK5-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
4539 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4540 // CHECK5: omp.inner.for.cond:
4541 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
4542 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !10
4543 // CHECK5-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
4544 // CHECK5-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4545 // CHECK5: omp.inner.for.body:
4546 // CHECK5-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
4547 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
4548 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4549 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !10
4550 // CHECK5-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
4551 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !10
4552 // CHECK5-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
4553 // CHECK5-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
4554 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !10
4555 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4556 // CHECK5: omp.body.continue:
4557 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4558 // CHECK5: omp.inner.for.inc:
4559 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
4560 // CHECK5-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
4561 // CHECK5-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
4562 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
4563 // CHECK5: omp.inner.for.end:
4564 // CHECK5-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
4565 // CHECK5: omp.dispatch.inc:
4566 // CHECK5-NEXT: br label [[OMP_DISPATCH_COND]]
4567 // CHECK5: omp.dispatch.end:
4568 // CHECK5-NEXT: ret void
4569 //
4570 //
4571 // CHECK5-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57
4572 // CHECK5-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4573 // CHECK5-NEXT: entry:
4574 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4575 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4576 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4577 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
4578 // CHECK5-NEXT: ret void
4579 //
4580 //
4581 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..14
4582 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4583 // CHECK5-NEXT: entry:
4584 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4585 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4586 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4587 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4588 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
4589 // CHECK5-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4590 // CHECK5-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4591 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4592 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4593 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
4594 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4595 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4596 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4597 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4598 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4599 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
4600 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4601 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4602 // CHECK5-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4603 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4604 // CHECK5-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4605 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4606 // CHECK5-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
4607 // CHECK5-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4608 // CHECK5: cond.true:
4609 // CHECK5-NEXT: br label [[COND_END:%.*]]
4610 // CHECK5: cond.false:
4611 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4612 // CHECK5-NEXT: br label [[COND_END]]
4613 // CHECK5: cond.end:
4614 // CHECK5-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4615 // CHECK5-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4616 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4617 // CHECK5-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4618 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4619 // CHECK5: omp.inner.for.cond:
4620 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4621 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4622 // CHECK5-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4623 // CHECK5-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4624 // CHECK5: omp.inner.for.body:
4625 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4626 // CHECK5-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
4627 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4628 // CHECK5-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
4629 // CHECK5-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
4630 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4631 // CHECK5: omp.inner.for.inc:
4632 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4633 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4634 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4635 // CHECK5-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4636 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]]
4637 // CHECK5: omp.inner.for.end:
4638 // CHECK5-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4639 // CHECK5: omp.loop.exit:
4640 // CHECK5-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4641 // CHECK5-NEXT: ret void
4642 //
4643 //
4644 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..15
4645 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4646 // CHECK5-NEXT: entry:
4647 // CHECK5-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4648 // CHECK5-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4649 // CHECK5-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4650 // CHECK5-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4651 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4652 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4653 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
4654 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4655 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4656 // CHECK5-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4657 // CHECK5-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4658 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
4659 // CHECK5-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4660 // CHECK5-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4661 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4662 // CHECK5-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4663 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4664 // CHECK5-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4665 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4666 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
4667 // CHECK5-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4668 // CHECK5-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
4669 // CHECK5-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4670 // CHECK5-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
4671 // CHECK5-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4672 // CHECK5-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4673 // CHECK5-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4674 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4675 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4676 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4677 // CHECK5-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4678 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
4679 // CHECK5-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
4680 // CHECK5-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
4681 // CHECK5: omp.dispatch.cond:
4682 // CHECK5-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
4683 // CHECK5-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
4684 // CHECK5-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
4685 // CHECK5: omp.dispatch.body:
4686 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4687 // CHECK5-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
4688 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4689 // CHECK5: omp.inner.for.cond:
4690 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
4691 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
4692 // CHECK5-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
4693 // CHECK5-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4694 // CHECK5: omp.inner.for.body:
4695 // CHECK5-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
4696 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
4697 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4698 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
4699 // CHECK5-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
4700 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
4701 // CHECK5-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
4702 // CHECK5-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
4703 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
4704 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4705 // CHECK5: omp.body.continue:
4706 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4707 // CHECK5: omp.inner.for.inc:
4708 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
4709 // CHECK5-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
4710 // CHECK5-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
4711 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
4712 // CHECK5: omp.inner.for.end:
4713 // CHECK5-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
4714 // CHECK5: omp.dispatch.inc:
4715 // CHECK5-NEXT: br label [[OMP_DISPATCH_COND]]
4716 // CHECK5: omp.dispatch.end:
4717 // CHECK5-NEXT: ret void
4718 //
4719 //
4720 // CHECK5-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4721 // CHECK5-SAME: () #[[ATTR3:[0-9]+]] {
4722 // CHECK5-NEXT: entry:
4723 // CHECK5-NEXT: call void @__tgt_register_requires(i64 1)
4724 // CHECK5-NEXT: ret void
4725 //
4726 //
4727 // CHECK6-LABEL: define {{[^@]+}}@_Z21teams_template_structv
4728 // CHECK6-SAME: () #[[ATTR0:[0-9]+]] {
4729 // CHECK6-NEXT: entry:
4730 // CHECK6-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
4731 // CHECK6-NEXT: [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
4732 // CHECK6-NEXT: ret i32 [[CALL]]
4733 //
4734 //
4735 // CHECK6-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
4736 // CHECK6-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
4737 // CHECK6-NEXT: entry:
4738 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4739 // CHECK6-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
4740 // CHECK6-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
4741 // CHECK6-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
4742 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4743 // CHECK6-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
4744 // CHECK6-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
4745 // CHECK6-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
4746 // CHECK6-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
4747 // CHECK6-NEXT: [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 8
4748 // CHECK6-NEXT: [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 8
4749 // CHECK6-NEXT: [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 8
4750 // CHECK6-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
4751 // CHECK6-NEXT: [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 8
4752 // CHECK6-NEXT: [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 8
4753 // CHECK6-NEXT: [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 8
4754 // CHECK6-NEXT: [[_TMP20:%.*]] = alloca i32, align 4
4755 // CHECK6-NEXT: [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 8
4756 // CHECK6-NEXT: [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 8
4757 // CHECK6-NEXT: [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 8
4758 // CHECK6-NEXT: [[_TMP27:%.*]] = alloca i32, align 4
4759 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4760 // CHECK6-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4761 // CHECK6-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
4762 // CHECK6-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4763 // CHECK6-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
4764 // CHECK6-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
4765 // CHECK6-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4766 // CHECK6-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
4767 // CHECK6-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
4768 // CHECK6-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4769 // CHECK6-NEXT: store i8* null, i8** [[TMP4]], align 8
4770 // CHECK6-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4771 // CHECK6-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4772 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
4773 // CHECK6-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4774 // CHECK6-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
4775 // CHECK6-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4776 // CHECK6: omp_offload.failed:
4777 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
4778 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT]]
4779 // CHECK6: omp_offload.cont:
4780 // CHECK6-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
4781 // CHECK6-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
4782 // CHECK6-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
4783 // CHECK6-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 8
4784 // CHECK6-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
4785 // CHECK6-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
4786 // CHECK6-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 8
4787 // CHECK6-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
4788 // CHECK6-NEXT: store i8* null, i8** [[TMP13]], align 8
4789 // CHECK6-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
4790 // CHECK6-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
4791 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
4792 // CHECK6-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4793 // CHECK6-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
4794 // CHECK6-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
4795 // CHECK6: omp_offload.failed7:
4796 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40(%struct.SS* [[THIS1]]) #[[ATTR2]]
4797 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT8]]
4798 // CHECK6: omp_offload.cont8:
4799 // CHECK6-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
4800 // CHECK6-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
4801 // CHECK6-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
4802 // CHECK6-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
4803 // CHECK6-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
4804 // CHECK6-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
4805 // CHECK6-NEXT: store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 8
4806 // CHECK6-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 0
4807 // CHECK6-NEXT: store i8* null, i8** [[TMP22]], align 8
4808 // CHECK6-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
4809 // CHECK6-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
4810 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
4811 // CHECK6-NEXT: [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4812 // CHECK6-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
4813 // CHECK6-NEXT: br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
4814 // CHECK6: omp_offload.failed14:
4815 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45(%struct.SS* [[THIS1]]) #[[ATTR2]]
4816 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT15]]
4817 // CHECK6: omp_offload.cont15:
4818 // CHECK6-NEXT: [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
4819 // CHECK6-NEXT: [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
4820 // CHECK6-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
4821 // CHECK6-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 8
4822 // CHECK6-NEXT: [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
4823 // CHECK6-NEXT: [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
4824 // CHECK6-NEXT: store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 8
4825 // CHECK6-NEXT: [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i64 0, i64 0
4826 // CHECK6-NEXT: store i8* null, i8** [[TMP31]], align 8
4827 // CHECK6-NEXT: [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
4828 // CHECK6-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
4829 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
4830 // CHECK6-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4831 // CHECK6-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
4832 // CHECK6-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
4833 // CHECK6: omp_offload.failed21:
4834 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51(%struct.SS* [[THIS1]]) #[[ATTR2]]
4835 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT22]]
4836 // CHECK6: omp_offload.cont22:
4837 // CHECK6-NEXT: [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
4838 // CHECK6-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
4839 // CHECK6-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
4840 // CHECK6-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
4841 // CHECK6-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
4842 // CHECK6-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
4843 // CHECK6-NEXT: store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 8
4844 // CHECK6-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i64 0, i64 0
4845 // CHECK6-NEXT: store i8* null, i8** [[TMP40]], align 8
4846 // CHECK6-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
4847 // CHECK6-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
4848 // CHECK6-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
4849 // CHECK6-NEXT: [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4850 // CHECK6-NEXT: [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
4851 // CHECK6-NEXT: br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
4852 // CHECK6: omp_offload.failed28:
4853 // CHECK6-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57(%struct.SS* [[THIS1]]) #[[ATTR2]]
4854 // CHECK6-NEXT: br label [[OMP_OFFLOAD_CONT29]]
4855 // CHECK6: omp_offload.cont29:
4856 // CHECK6-NEXT: [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
4857 // CHECK6-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 0
4858 // CHECK6-NEXT: [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
4859 // CHECK6-NEXT: ret i32 [[TMP45]]
4860 //
4861 //
4862 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35
4863 // CHECK6-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
4864 // CHECK6-NEXT: entry:
4865 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4866 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4867 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4868 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
4869 // CHECK6-NEXT: ret void
4870 //
4871 //
4872 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined.
4873 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4874 // CHECK6-NEXT: entry:
4875 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4876 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4877 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4878 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4879 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4880 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4881 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4882 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4883 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4884 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4885 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4886 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4887 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4888 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4889 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4890 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
4891 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4892 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4893 // CHECK6-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4894 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4895 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4896 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4897 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
4898 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4899 // CHECK6: cond.true:
4900 // CHECK6-NEXT: br label [[COND_END:%.*]]
4901 // CHECK6: cond.false:
4902 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4903 // CHECK6-NEXT: br label [[COND_END]]
4904 // CHECK6: cond.end:
4905 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4906 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4907 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4908 // CHECK6-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4909 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4910 // CHECK6: omp.inner.for.cond:
4911 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4912 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4913 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4914 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4915 // CHECK6: omp.inner.for.body:
4916 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4917 // CHECK6-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
4918 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4919 // CHECK6-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
4920 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
4921 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4922 // CHECK6: omp.inner.for.inc:
4923 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4924 // CHECK6-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4925 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4926 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4927 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
4928 // CHECK6: omp.inner.for.end:
4929 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4930 // CHECK6: omp.loop.exit:
4931 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4932 // CHECK6-NEXT: ret void
4933 //
4934 //
4935 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..1
4936 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
4937 // CHECK6-NEXT: entry:
4938 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4939 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4940 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4941 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4942 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
4943 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4944 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
4945 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4946 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4947 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4948 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4949 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
4950 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4951 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4952 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4953 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4954 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
4955 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
4956 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4957 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
4958 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4959 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
4960 // CHECK6-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4961 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
4962 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4963 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4964 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4965 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4966 // CHECK6-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4967 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4968 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4969 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4970 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
4971 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4972 // CHECK6: cond.true:
4973 // CHECK6-NEXT: br label [[COND_END:%.*]]
4974 // CHECK6: cond.false:
4975 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4976 // CHECK6-NEXT: br label [[COND_END]]
4977 // CHECK6: cond.end:
4978 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
4979 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4980 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4981 // CHECK6-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
4982 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4983 // CHECK6: omp.inner.for.cond:
4984 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4985 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4986 // CHECK6-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
4987 // CHECK6-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4988 // CHECK6: omp.inner.for.body:
4989 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4990 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
4991 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4992 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4993 // CHECK6-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
4994 // CHECK6-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
4995 // CHECK6-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
4996 // CHECK6-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
4997 // CHECK6-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4998 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4999 // CHECK6: omp.body.continue:
5000 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5001 // CHECK6: omp.inner.for.inc:
5002 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5003 // CHECK6-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
5004 // CHECK6-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
5005 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
5006 // CHECK6: omp.inner.for.end:
5007 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5008 // CHECK6: omp.loop.exit:
5009 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
5010 // CHECK6-NEXT: ret void
5011 //
5012 //
5013 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40
5014 // CHECK6-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5015 // CHECK6-NEXT: entry:
5016 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
5017 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
5018 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
5019 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
5020 // CHECK6-NEXT: ret void
5021 //
5022 //
5023 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..2
5024 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5025 // CHECK6-NEXT: entry:
5026 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5027 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5028 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
5029 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5030 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
5031 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5032 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5033 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5034 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5035 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
5036 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5037 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5038 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
5039 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
5040 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5041 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
5042 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5043 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5044 // CHECK6-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5045 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5046 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5047 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5048 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
5049 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5050 // CHECK6: cond.true:
5051 // CHECK6-NEXT: br label [[COND_END:%.*]]
5052 // CHECK6: cond.false:
5053 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5054 // CHECK6-NEXT: br label [[COND_END]]
5055 // CHECK6: cond.end:
5056 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5057 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5058 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5059 // CHECK6-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
5060 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5061 // CHECK6: omp.inner.for.cond:
5062 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5063 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5064 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5065 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5066 // CHECK6: omp.inner.for.body:
5067 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5068 // CHECK6-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
5069 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5070 // CHECK6-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
5071 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
5072 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5073 // CHECK6: omp.inner.for.inc:
5074 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5075 // CHECK6-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5076 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5077 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5078 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
5079 // CHECK6: omp.inner.for.end:
5080 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5081 // CHECK6: omp.loop.exit:
5082 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
5083 // CHECK6-NEXT: ret void
5084 //
5085 //
5086 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..3
5087 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5088 // CHECK6-NEXT: entry:
5089 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5090 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5091 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5092 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5093 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
5094 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5095 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
5096 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5097 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5098 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5099 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5100 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
5101 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5102 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5103 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5104 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5105 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
5106 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
5107 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5108 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
5109 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5110 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
5111 // CHECK6-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5112 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
5113 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5114 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5115 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5116 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5117 // CHECK6-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5118 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
5119 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5120 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5121 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
5122 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5123 // CHECK6: cond.true:
5124 // CHECK6-NEXT: br label [[COND_END:%.*]]
5125 // CHECK6: cond.false:
5126 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5127 // CHECK6-NEXT: br label [[COND_END]]
5128 // CHECK6: cond.end:
5129 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
5130 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5131 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5132 // CHECK6-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
5133 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5134 // CHECK6: omp.inner.for.cond:
5135 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5136 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5137 // CHECK6-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
5138 // CHECK6-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5139 // CHECK6: omp.inner.for.body:
5140 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5141 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
5142 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5143 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4
5144 // CHECK6-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
5145 // CHECK6-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
5146 // CHECK6-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
5147 // CHECK6-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
5148 // CHECK6-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
5149 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5150 // CHECK6: omp.body.continue:
5151 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5152 // CHECK6: omp.inner.for.inc:
5153 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5154 // CHECK6-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
5155 // CHECK6-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
5156 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
5157 // CHECK6: omp.inner.for.end:
5158 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5159 // CHECK6: omp.loop.exit:
5160 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
5161 // CHECK6-NEXT: ret void
5162 //
5163 //
5164 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45
5165 // CHECK6-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5166 // CHECK6-NEXT: entry:
5167 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
5168 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
5169 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
5170 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
5171 // CHECK6-NEXT: ret void
5172 //
5173 //
5174 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..6
5175 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5176 // CHECK6-NEXT: entry:
5177 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5178 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5179 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
5180 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5181 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
5182 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5183 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5184 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5185 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5186 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
5187 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5188 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5189 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
5190 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
5191 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5192 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
5193 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5194 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5195 // CHECK6-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5196 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5197 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5198 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5199 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
5200 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5201 // CHECK6: cond.true:
5202 // CHECK6-NEXT: br label [[COND_END:%.*]]
5203 // CHECK6: cond.false:
5204 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5205 // CHECK6-NEXT: br label [[COND_END]]
5206 // CHECK6: cond.end:
5207 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5208 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5209 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5210 // CHECK6-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
5211 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5212 // CHECK6: omp.inner.for.cond:
5213 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5214 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5215 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5216 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5217 // CHECK6: omp.inner.for.body:
5218 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5219 // CHECK6-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
5220 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5221 // CHECK6-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
5222 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
5223 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5224 // CHECK6: omp.inner.for.inc:
5225 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5226 // CHECK6-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5227 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5228 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5229 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
5230 // CHECK6: omp.inner.for.end:
5231 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5232 // CHECK6: omp.loop.exit:
5233 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
5234 // CHECK6-NEXT: ret void
5235 //
5236 //
5237 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..7
5238 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5239 // CHECK6-NEXT: entry:
5240 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5241 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5242 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5243 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5244 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
5245 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5246 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
5247 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5248 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5249 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5250 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5251 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
5252 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5253 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5254 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5255 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5256 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
5257 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
5258 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5259 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
5260 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5261 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
5262 // CHECK6-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5263 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
5264 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5265 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5266 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5267 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5268 // CHECK6-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5269 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
5270 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
5271 // CHECK6-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
5272 // CHECK6: omp.dispatch.cond:
5273 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5274 // CHECK6-NEXT: [[TMP6:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5275 // CHECK6-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP6]] to i32
5276 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[CONV2]]
5277 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5278 // CHECK6: cond.true:
5279 // CHECK6-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5280 // CHECK6-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
5281 // CHECK6-NEXT: br label [[COND_END:%.*]]
5282 // CHECK6: cond.false:
5283 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5284 // CHECK6-NEXT: br label [[COND_END]]
5285 // CHECK6: cond.end:
5286 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ [[CONV3]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
5287 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5288 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5289 // CHECK6-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
5290 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5291 // CHECK6-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5292 // CHECK6-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
5293 // CHECK6-NEXT: br i1 [[CMP4]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
5294 // CHECK6: omp.dispatch.body:
5295 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5296 // CHECK6: omp.inner.for.cond:
5297 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5298 // CHECK6-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5299 // CHECK6-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
5300 // CHECK6-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5301 // CHECK6: omp.inner.for.body:
5302 // CHECK6-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5303 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
5304 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5305 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4
5306 // CHECK6-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
5307 // CHECK6-NEXT: [[TMP15:%.*]] = load i32, i32* [[I]], align 4
5308 // CHECK6-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP15]] to i64
5309 // CHECK6-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
5310 // CHECK6-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
5311 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5312 // CHECK6: omp.body.continue:
5313 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5314 // CHECK6: omp.inner.for.inc:
5315 // CHECK6-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5316 // CHECK6-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP16]], 1
5317 // CHECK6-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
5318 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
5319 // CHECK6: omp.inner.for.end:
5320 // CHECK6-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
5321 // CHECK6: omp.dispatch.inc:
5322 // CHECK6-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5323 // CHECK6-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5324 // CHECK6-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
5325 // CHECK6-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_LB]], align 4
5326 // CHECK6-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5327 // CHECK6-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5328 // CHECK6-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
5329 // CHECK6-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_UB]], align 4
5330 // CHECK6-NEXT: br label [[OMP_DISPATCH_COND]]
5331 // CHECK6: omp.dispatch.end:
5332 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
5333 // CHECK6-NEXT: ret void
5334 //
5335 //
5336 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
5337 // CHECK6-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5338 // CHECK6-NEXT: entry:
5339 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
5340 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
5341 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
5342 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
5343 // CHECK6-NEXT: ret void
5344 //
5345 //
5346 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..10
5347 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5348 // CHECK6-NEXT: entry:
5349 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5350 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5351 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
5352 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5353 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
5354 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5355 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5356 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5357 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5358 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
5359 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5360 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5361 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
5362 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
5363 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5364 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
5365 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5366 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5367 // CHECK6-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5368 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5369 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5370 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5371 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
5372 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5373 // CHECK6: cond.true:
5374 // CHECK6-NEXT: br label [[COND_END:%.*]]
5375 // CHECK6: cond.false:
5376 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5377 // CHECK6-NEXT: br label [[COND_END]]
5378 // CHECK6: cond.end:
5379 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5380 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5381 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5382 // CHECK6-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
5383 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5384 // CHECK6: omp.inner.for.cond:
5385 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5386 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5387 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5388 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5389 // CHECK6: omp.inner.for.body:
5390 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5391 // CHECK6-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
5392 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5393 // CHECK6-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
5394 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
5395 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5396 // CHECK6: omp.inner.for.inc:
5397 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5398 // CHECK6-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5399 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5400 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5401 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
5402 // CHECK6: omp.inner.for.end:
5403 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5404 // CHECK6: omp.loop.exit:
5405 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
5406 // CHECK6-NEXT: ret void
5407 //
5408 //
5409 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..11
5410 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5411 // CHECK6-NEXT: entry:
5412 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5413 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5414 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5415 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5416 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
5417 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5418 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
5419 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5420 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5421 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5422 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5423 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
5424 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5425 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5426 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5427 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5428 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
5429 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
5430 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5431 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
5432 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5433 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
5434 // CHECK6-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5435 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
5436 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5437 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5438 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5439 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5440 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5441 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5442 // CHECK6-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5443 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
5444 // CHECK6-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
5445 // CHECK6-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
5446 // CHECK6: omp.dispatch.cond:
5447 // CHECK6-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
5448 // CHECK6-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
5449 // CHECK6-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
5450 // CHECK6: omp.dispatch.body:
5451 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5452 // CHECK6-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
5453 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5454 // CHECK6: omp.inner.for.cond:
5455 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
5456 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !10
5457 // CHECK6-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
5458 // CHECK6-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5459 // CHECK6: omp.inner.for.body:
5460 // CHECK6-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
5461 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
5462 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5463 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !10
5464 // CHECK6-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
5465 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !10
5466 // CHECK6-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
5467 // CHECK6-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
5468 // CHECK6-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !10
5469 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5470 // CHECK6: omp.body.continue:
5471 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5472 // CHECK6: omp.inner.for.inc:
5473 // CHECK6-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
5474 // CHECK6-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
5475 // CHECK6-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !10
5476 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
5477 // CHECK6: omp.inner.for.end:
5478 // CHECK6-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
5479 // CHECK6: omp.dispatch.inc:
5480 // CHECK6-NEXT: br label [[OMP_DISPATCH_COND]]
5481 // CHECK6: omp.dispatch.end:
5482 // CHECK6-NEXT: ret void
5483 //
5484 //
5485 // CHECK6-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57
5486 // CHECK6-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5487 // CHECK6-NEXT: entry:
5488 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
5489 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
5490 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
5491 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
5492 // CHECK6-NEXT: ret void
5493 //
5494 //
5495 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..14
5496 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5497 // CHECK6-NEXT: entry:
5498 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5499 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5500 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
5501 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5502 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
5503 // CHECK6-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5504 // CHECK6-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5505 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5506 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5507 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
5508 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5509 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5510 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
5511 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
5512 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5513 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
5514 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5515 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5516 // CHECK6-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5517 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5518 // CHECK6-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5519 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5520 // CHECK6-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
5521 // CHECK6-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5522 // CHECK6: cond.true:
5523 // CHECK6-NEXT: br label [[COND_END:%.*]]
5524 // CHECK6: cond.false:
5525 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5526 // CHECK6-NEXT: br label [[COND_END]]
5527 // CHECK6: cond.end:
5528 // CHECK6-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5529 // CHECK6-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5530 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5531 // CHECK6-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
5532 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5533 // CHECK6: omp.inner.for.cond:
5534 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5535 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5536 // CHECK6-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5537 // CHECK6-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5538 // CHECK6: omp.inner.for.body:
5539 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5540 // CHECK6-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
5541 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5542 // CHECK6-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
5543 // CHECK6-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
5544 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5545 // CHECK6: omp.inner.for.inc:
5546 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5547 // CHECK6-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5548 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5549 // CHECK6-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5550 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]]
5551 // CHECK6: omp.inner.for.end:
5552 // CHECK6-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5553 // CHECK6: omp.loop.exit:
5554 // CHECK6-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
5555 // CHECK6-NEXT: ret void
5556 //
5557 //
5558 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..15
5559 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5560 // CHECK6-NEXT: entry:
5561 // CHECK6-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5562 // CHECK6-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5563 // CHECK6-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5564 // CHECK6-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5565 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
5566 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5567 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
5568 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5569 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5570 // CHECK6-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5571 // CHECK6-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5572 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
5573 // CHECK6-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5574 // CHECK6-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5575 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5576 // CHECK6-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5577 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
5578 // CHECK6-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
5579 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5580 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
5581 // CHECK6-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5582 // CHECK6-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
5583 // CHECK6-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5584 // CHECK6-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
5585 // CHECK6-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5586 // CHECK6-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5587 // CHECK6-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5588 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5589 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5590 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5591 // CHECK6-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5592 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
5593 // CHECK6-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
5594 // CHECK6-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
5595 // CHECK6: omp.dispatch.cond:
5596 // CHECK6-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
5597 // CHECK6-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
5598 // CHECK6-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
5599 // CHECK6: omp.dispatch.body:
5600 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5601 // CHECK6-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
5602 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5603 // CHECK6: omp.inner.for.cond:
5604 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
5605 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
5606 // CHECK6-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
5607 // CHECK6-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5608 // CHECK6: omp.inner.for.body:
5609 // CHECK6-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
5610 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
5611 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5612 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
5613 // CHECK6-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
5614 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
5615 // CHECK6-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
5616 // CHECK6-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
5617 // CHECK6-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
5618 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5619 // CHECK6: omp.body.continue:
5620 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5621 // CHECK6: omp.inner.for.inc:
5622 // CHECK6-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
5623 // CHECK6-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
5624 // CHECK6-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
5625 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
5626 // CHECK6: omp.inner.for.end:
5627 // CHECK6-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
5628 // CHECK6: omp.dispatch.inc:
5629 // CHECK6-NEXT: br label [[OMP_DISPATCH_COND]]
5630 // CHECK6: omp.dispatch.end:
5631 // CHECK6-NEXT: ret void
5632 //
5633 //
5634 // CHECK6-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
5635 // CHECK6-SAME: () #[[ATTR3:[0-9]+]] {
5636 // CHECK6-NEXT: entry:
5637 // CHECK6-NEXT: call void @__tgt_register_requires(i64 1)
5638 // CHECK6-NEXT: ret void
5639 //
5640 //
5641 // CHECK7-LABEL: define {{[^@]+}}@_Z21teams_template_structv
5642 // CHECK7-SAME: () #[[ATTR0:[0-9]+]] {
5643 // CHECK7-NEXT: entry:
5644 // CHECK7-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
5645 // CHECK7-NEXT: [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
5646 // CHECK7-NEXT: ret i32 [[CALL]]
5647 //
5648 //
5649 // CHECK7-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
5650 // CHECK7-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
5651 // CHECK7-NEXT: entry:
5652 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
5653 // CHECK7-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
5654 // CHECK7-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
5655 // CHECK7-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
5656 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
5657 // CHECK7-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
5658 // CHECK7-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
5659 // CHECK7-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
5660 // CHECK7-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
5661 // CHECK7-NEXT: [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 4
5662 // CHECK7-NEXT: [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 4
5663 // CHECK7-NEXT: [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 4
5664 // CHECK7-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
5665 // CHECK7-NEXT: [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 4
5666 // CHECK7-NEXT: [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 4
5667 // CHECK7-NEXT: [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 4
5668 // CHECK7-NEXT: [[_TMP20:%.*]] = alloca i32, align 4
5669 // CHECK7-NEXT: [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 4
5670 // CHECK7-NEXT: [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 4
5671 // CHECK7-NEXT: [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 4
5672 // CHECK7-NEXT: [[_TMP27:%.*]] = alloca i32, align 4
5673 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
5674 // CHECK7-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
5675 // CHECK7-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
5676 // CHECK7-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5677 // CHECK7-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
5678 // CHECK7-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
5679 // CHECK7-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5680 // CHECK7-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
5681 // CHECK7-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
5682 // CHECK7-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
5683 // CHECK7-NEXT: store i8* null, i8** [[TMP4]], align 4
5684 // CHECK7-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5685 // CHECK7-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5686 // CHECK7-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
5687 // CHECK7-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5688 // CHECK7-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
5689 // CHECK7-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5690 // CHECK7: omp_offload.failed:
5691 // CHECK7-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
5692 // CHECK7-NEXT: br label [[OMP_OFFLOAD_CONT]]
5693 // CHECK7: omp_offload.cont:
5694 // CHECK7-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
5695 // CHECK7-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
5696 // CHECK7-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
5697 // CHECK7-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 4
5698 // CHECK7-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
5699 // CHECK7-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
5700 // CHECK7-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 4
5701 // CHECK7-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
5702 // CHECK7-NEXT: store i8* null, i8** [[TMP13]], align 4
5703 // CHECK7-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
5704 // CHECK7-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
5705 // CHECK7-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
5706 // CHECK7-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5707 // CHECK7-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
5708 // CHECK7-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
5709 // CHECK7: omp_offload.failed7:
5710 // CHECK7-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40(%struct.SS* [[THIS1]]) #[[ATTR2]]
5711 // CHECK7-NEXT: br label [[OMP_OFFLOAD_CONT8]]
5712 // CHECK7: omp_offload.cont8:
5713 // CHECK7-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
5714 // CHECK7-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
5715 // CHECK7-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
5716 // CHECK7-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
5717 // CHECK7-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
5718 // CHECK7-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
5719 // CHECK7-NEXT: store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 4
5720 // CHECK7-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i32 0, i32 0
5721 // CHECK7-NEXT: store i8* null, i8** [[TMP22]], align 4
5722 // CHECK7-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
5723 // CHECK7-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
5724 // CHECK7-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
5725 // CHECK7-NEXT: [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5726 // CHECK7-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
5727 // CHECK7-NEXT: br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
5728 // CHECK7: omp_offload.failed14:
5729 // CHECK7-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45(%struct.SS* [[THIS1]]) #[[ATTR2]]
5730 // CHECK7-NEXT: br label [[OMP_OFFLOAD_CONT15]]
5731 // CHECK7: omp_offload.cont15:
5732 // CHECK7-NEXT: [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
5733 // CHECK7-NEXT: [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
5734 // CHECK7-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
5735 // CHECK7-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 4
5736 // CHECK7-NEXT: [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
5737 // CHECK7-NEXT: [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
5738 // CHECK7-NEXT: store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 4
5739 // CHECK7-NEXT: [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i32 0, i32 0
5740 // CHECK7-NEXT: store i8* null, i8** [[TMP31]], align 4
5741 // CHECK7-NEXT: [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
5742 // CHECK7-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
5743 // CHECK7-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
5744 // CHECK7-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5745 // CHECK7-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
5746 // CHECK7-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
5747 // CHECK7: omp_offload.failed21:
5748 // CHECK7-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51(%struct.SS* [[THIS1]]) #[[ATTR2]]
5749 // CHECK7-NEXT: br label [[OMP_OFFLOAD_CONT22]]
5750 // CHECK7: omp_offload.cont22:
5751 // CHECK7-NEXT: [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
5752 // CHECK7-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
5753 // CHECK7-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
5754 // CHECK7-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
5755 // CHECK7-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
5756 // CHECK7-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
5757 // CHECK7-NEXT: store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 4
5758 // CHECK7-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 0
5759 // CHECK7-NEXT: store i8* null, i8** [[TMP40]], align 4
5760 // CHECK7-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
5761 // CHECK7-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
5762 // CHECK7-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
5763 // CHECK7-NEXT: [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5764 // CHECK7-NEXT: [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
5765 // CHECK7-NEXT: br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
5766 // CHECK7: omp_offload.failed28:
5767 // CHECK7-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57(%struct.SS* [[THIS1]]) #[[ATTR2]]
5768 // CHECK7-NEXT: br label [[OMP_OFFLOAD_CONT29]]
5769 // CHECK7: omp_offload.cont29:
5770 // CHECK7-NEXT: [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
5771 // CHECK7-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i32 0, i32 0
5772 // CHECK7-NEXT: [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
5773 // CHECK7-NEXT: ret i32 [[TMP45]]
5774 //
5775 //
5776 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35
5777 // CHECK7-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
5778 // CHECK7-NEXT: entry:
5779 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
5780 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
5781 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
5782 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
5783 // CHECK7-NEXT: ret void
5784 //
5785 //
5786 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined.
5787 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5788 // CHECK7-NEXT: entry:
5789 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5790 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5791 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
5792 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5793 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
5794 // CHECK7-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5795 // CHECK7-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5796 // CHECK7-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5797 // CHECK7-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5798 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
5799 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5800 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5801 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
5802 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
5803 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5804 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
5805 // CHECK7-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5806 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5807 // CHECK7-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5808 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5809 // CHECK7-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5810 // CHECK7-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5811 // CHECK7-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
5812 // CHECK7-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5813 // CHECK7: cond.true:
5814 // CHECK7-NEXT: br label [[COND_END:%.*]]
5815 // CHECK7: cond.false:
5816 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5817 // CHECK7-NEXT: br label [[COND_END]]
5818 // CHECK7: cond.end:
5819 // CHECK7-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5820 // CHECK7-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5821 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5822 // CHECK7-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
5823 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5824 // CHECK7: omp.inner.for.cond:
5825 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5826 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5827 // CHECK7-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5828 // CHECK7-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5829 // CHECK7: omp.inner.for.body:
5830 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5831 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5832 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
5833 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5834 // CHECK7: omp.inner.for.inc:
5835 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5836 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5837 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
5838 // CHECK7-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5839 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]]
5840 // CHECK7: omp.inner.for.end:
5841 // CHECK7-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5842 // CHECK7: omp.loop.exit:
5843 // CHECK7-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
5844 // CHECK7-NEXT: ret void
5845 //
5846 //
5847 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..1
5848 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5849 // CHECK7-NEXT: entry:
5850 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5851 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5852 // CHECK7-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
5853 // CHECK7-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
5854 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
5855 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5856 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
5857 // CHECK7-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5858 // CHECK7-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5859 // CHECK7-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5860 // CHECK7-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5861 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
5862 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5863 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5864 // CHECK7-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
5865 // CHECK7-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
5866 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
5867 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
5868 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5869 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
5870 // CHECK7-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
5871 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
5872 // CHECK7-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
5873 // CHECK7-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
5874 // CHECK7-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5875 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5876 // CHECK7-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5877 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
5878 // CHECK7-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5879 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5880 // CHECK7-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
5881 // CHECK7-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5882 // CHECK7: cond.true:
5883 // CHECK7-NEXT: br label [[COND_END:%.*]]
5884 // CHECK7: cond.false:
5885 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5886 // CHECK7-NEXT: br label [[COND_END]]
5887 // CHECK7: cond.end:
5888 // CHECK7-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
5889 // CHECK7-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5890 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5891 // CHECK7-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
5892 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5893 // CHECK7: omp.inner.for.cond:
5894 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5895 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5896 // CHECK7-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
5897 // CHECK7-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5898 // CHECK7: omp.inner.for.body:
5899 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5900 // CHECK7-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
5901 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5902 // CHECK7-NEXT: store i32 [[ADD]], i32* [[I]], align 4
5903 // CHECK7-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
5904 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
5905 // CHECK7-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
5906 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
5907 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5908 // CHECK7: omp.body.continue:
5909 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5910 // CHECK7: omp.inner.for.inc:
5911 // CHECK7-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5912 // CHECK7-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
5913 // CHECK7-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
5914 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]]
5915 // CHECK7: omp.inner.for.end:
5916 // CHECK7-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5917 // CHECK7: omp.loop.exit:
5918 // CHECK7-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
5919 // CHECK7-NEXT: ret void
5920 //
5921 //
5922 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40
5923 // CHECK7-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5924 // CHECK7-NEXT: entry:
5925 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
5926 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
5927 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
5928 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
5929 // CHECK7-NEXT: ret void
5930 //
5931 //
5932 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..2
5933 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5934 // CHECK7-NEXT: entry:
5935 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5936 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5937 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
5938 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5939 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
5940 // CHECK7-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5941 // CHECK7-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5942 // CHECK7-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5943 // CHECK7-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5944 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
5945 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
5946 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
5947 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
5948 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
5949 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5950 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
5951 // CHECK7-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5952 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5953 // CHECK7-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
5954 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5955 // CHECK7-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5956 // CHECK7-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5957 // CHECK7-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
5958 // CHECK7-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5959 // CHECK7: cond.true:
5960 // CHECK7-NEXT: br label [[COND_END:%.*]]
5961 // CHECK7: cond.false:
5962 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5963 // CHECK7-NEXT: br label [[COND_END]]
5964 // CHECK7: cond.end:
5965 // CHECK7-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5966 // CHECK7-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5967 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5968 // CHECK7-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
5969 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5970 // CHECK7: omp.inner.for.cond:
5971 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5972 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5973 // CHECK7-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5974 // CHECK7-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5975 // CHECK7: omp.inner.for.body:
5976 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5977 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5978 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
5979 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5980 // CHECK7: omp.inner.for.inc:
5981 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5982 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5983 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
5984 // CHECK7-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5985 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]]
5986 // CHECK7: omp.inner.for.end:
5987 // CHECK7-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5988 // CHECK7: omp.loop.exit:
5989 // CHECK7-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
5990 // CHECK7-NEXT: ret void
5991 //
5992 //
5993 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..3
5994 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
5995 // CHECK7-NEXT: entry:
5996 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
5997 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
5998 // CHECK7-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
5999 // CHECK7-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
6000 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6001 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6002 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
6003 // CHECK7-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6004 // CHECK7-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6005 // CHECK7-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6006 // CHECK7-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6007 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
6008 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6009 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6010 // CHECK7-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6011 // CHECK7-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6012 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6013 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6014 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6015 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
6016 // CHECK7-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6017 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6018 // CHECK7-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
6019 // CHECK7-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
6020 // CHECK7-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6021 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6022 // CHECK7-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6023 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
6024 // CHECK7-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6025 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6026 // CHECK7-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
6027 // CHECK7-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6028 // CHECK7: cond.true:
6029 // CHECK7-NEXT: br label [[COND_END:%.*]]
6030 // CHECK7: cond.false:
6031 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6032 // CHECK7-NEXT: br label [[COND_END]]
6033 // CHECK7: cond.end:
6034 // CHECK7-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
6035 // CHECK7-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6036 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6037 // CHECK7-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
6038 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6039 // CHECK7: omp.inner.for.cond:
6040 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6041 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6042 // CHECK7-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
6043 // CHECK7-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6044 // CHECK7: omp.inner.for.body:
6045 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6046 // CHECK7-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
6047 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6048 // CHECK7-NEXT: store i32 [[ADD]], i32* [[I]], align 4
6049 // CHECK7-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
6050 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
6051 // CHECK7-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
6052 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
6053 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6054 // CHECK7: omp.body.continue:
6055 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6056 // CHECK7: omp.inner.for.inc:
6057 // CHECK7-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6058 // CHECK7-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
6059 // CHECK7-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
6060 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]]
6061 // CHECK7: omp.inner.for.end:
6062 // CHECK7-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6063 // CHECK7: omp.loop.exit:
6064 // CHECK7-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
6065 // CHECK7-NEXT: ret void
6066 //
6067 //
6068 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45
6069 // CHECK7-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6070 // CHECK7-NEXT: entry:
6071 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6072 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6073 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6074 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
6075 // CHECK7-NEXT: ret void
6076 //
6077 //
6078 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..6
6079 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6080 // CHECK7-NEXT: entry:
6081 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6082 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6083 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6084 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6085 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
6086 // CHECK7-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6087 // CHECK7-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6088 // CHECK7-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6089 // CHECK7-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6090 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
6091 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6092 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6093 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6094 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6095 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6096 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
6097 // CHECK7-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6098 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6099 // CHECK7-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6100 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6101 // CHECK7-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6102 // CHECK7-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6103 // CHECK7-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
6104 // CHECK7-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6105 // CHECK7: cond.true:
6106 // CHECK7-NEXT: br label [[COND_END:%.*]]
6107 // CHECK7: cond.false:
6108 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6109 // CHECK7-NEXT: br label [[COND_END]]
6110 // CHECK7: cond.end:
6111 // CHECK7-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
6112 // CHECK7-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6113 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6114 // CHECK7-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
6115 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6116 // CHECK7: omp.inner.for.cond:
6117 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6118 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6119 // CHECK7-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6120 // CHECK7-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6121 // CHECK7: omp.inner.for.body:
6122 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6123 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6124 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
6125 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6126 // CHECK7: omp.inner.for.inc:
6127 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6128 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6129 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
6130 // CHECK7-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6131 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]]
6132 // CHECK7: omp.inner.for.end:
6133 // CHECK7-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6134 // CHECK7: omp.loop.exit:
6135 // CHECK7-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
6136 // CHECK7-NEXT: ret void
6137 //
6138 //
6139 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..7
6140 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6141 // CHECK7-NEXT: entry:
6142 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6143 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6144 // CHECK7-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
6145 // CHECK7-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
6146 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6147 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6148 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
6149 // CHECK7-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6150 // CHECK7-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6151 // CHECK7-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6152 // CHECK7-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6153 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
6154 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6155 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6156 // CHECK7-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6157 // CHECK7-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6158 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6159 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6160 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6161 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
6162 // CHECK7-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6163 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6164 // CHECK7-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
6165 // CHECK7-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
6166 // CHECK7-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6167 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6168 // CHECK7-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6169 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
6170 // CHECK7-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
6171 // CHECK7-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
6172 // CHECK7: omp.dispatch.cond:
6173 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6174 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6175 // CHECK7-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
6176 // CHECK7-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6177 // CHECK7: cond.true:
6178 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6179 // CHECK7-NEXT: br label [[COND_END:%.*]]
6180 // CHECK7: cond.false:
6181 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6182 // CHECK7-NEXT: br label [[COND_END]]
6183 // CHECK7: cond.end:
6184 // CHECK7-NEXT: [[COND:%.*]] = phi i32 [ [[TMP7]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
6185 // CHECK7-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6186 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6187 // CHECK7-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
6188 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6189 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6190 // CHECK7-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
6191 // CHECK7-NEXT: br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
6192 // CHECK7: omp.dispatch.body:
6193 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6194 // CHECK7: omp.inner.for.cond:
6195 // CHECK7-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6196 // CHECK7-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6197 // CHECK7-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
6198 // CHECK7-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6199 // CHECK7: omp.inner.for.body:
6200 // CHECK7-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6201 // CHECK7-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
6202 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6203 // CHECK7-NEXT: store i32 [[ADD]], i32* [[I]], align 4
6204 // CHECK7-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
6205 // CHECK7-NEXT: [[TMP15:%.*]] = load i32, i32* [[I]], align 4
6206 // CHECK7-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP15]]
6207 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
6208 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6209 // CHECK7: omp.body.continue:
6210 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6211 // CHECK7: omp.inner.for.inc:
6212 // CHECK7-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6213 // CHECK7-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1
6214 // CHECK7-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
6215 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]]
6216 // CHECK7: omp.inner.for.end:
6217 // CHECK7-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
6218 // CHECK7: omp.dispatch.inc:
6219 // CHECK7-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6220 // CHECK7-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6221 // CHECK7-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
6222 // CHECK7-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
6223 // CHECK7-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6224 // CHECK7-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6225 // CHECK7-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
6226 // CHECK7-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
6227 // CHECK7-NEXT: br label [[OMP_DISPATCH_COND]]
6228 // CHECK7: omp.dispatch.end:
6229 // CHECK7-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
6230 // CHECK7-NEXT: ret void
6231 //
6232 //
6233 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
6234 // CHECK7-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6235 // CHECK7-NEXT: entry:
6236 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6237 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6238 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6239 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
6240 // CHECK7-NEXT: ret void
6241 //
6242 //
6243 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..10
6244 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6245 // CHECK7-NEXT: entry:
6246 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6247 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6248 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6249 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6250 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
6251 // CHECK7-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6252 // CHECK7-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6253 // CHECK7-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6254 // CHECK7-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6255 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
6256 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6257 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6258 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6259 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6260 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6261 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
6262 // CHECK7-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6263 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6264 // CHECK7-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6265 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6266 // CHECK7-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6267 // CHECK7-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6268 // CHECK7-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
6269 // CHECK7-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6270 // CHECK7: cond.true:
6271 // CHECK7-NEXT: br label [[COND_END:%.*]]
6272 // CHECK7: cond.false:
6273 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6274 // CHECK7-NEXT: br label [[COND_END]]
6275 // CHECK7: cond.end:
6276 // CHECK7-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
6277 // CHECK7-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6278 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6279 // CHECK7-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
6280 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6281 // CHECK7: omp.inner.for.cond:
6282 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6283 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6284 // CHECK7-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6285 // CHECK7-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6286 // CHECK7: omp.inner.for.body:
6287 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6288 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6289 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
6290 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6291 // CHECK7: omp.inner.for.inc:
6292 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6293 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6294 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
6295 // CHECK7-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6296 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]]
6297 // CHECK7: omp.inner.for.end:
6298 // CHECK7-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6299 // CHECK7: omp.loop.exit:
6300 // CHECK7-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
6301 // CHECK7-NEXT: ret void
6302 //
6303 //
6304 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..11
6305 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6306 // CHECK7-NEXT: entry:
6307 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6308 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6309 // CHECK7-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
6310 // CHECK7-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
6311 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6312 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6313 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
6314 // CHECK7-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6315 // CHECK7-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6316 // CHECK7-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6317 // CHECK7-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6318 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
6319 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6320 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6321 // CHECK7-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6322 // CHECK7-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6323 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6324 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6325 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6326 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
6327 // CHECK7-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6328 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6329 // CHECK7-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
6330 // CHECK7-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
6331 // CHECK7-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6332 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6333 // CHECK7-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6334 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6335 // CHECK7-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6336 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
6337 // CHECK7-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
6338 // CHECK7-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
6339 // CHECK7: omp.dispatch.cond:
6340 // CHECK7-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
6341 // CHECK7-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
6342 // CHECK7-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
6343 // CHECK7: omp.dispatch.body:
6344 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6345 // CHECK7-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
6346 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6347 // CHECK7: omp.inner.for.cond:
6348 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
6349 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
6350 // CHECK7-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
6351 // CHECK7-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6352 // CHECK7: omp.inner.for.body:
6353 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
6354 // CHECK7-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
6355 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6356 // CHECK7-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !11
6357 // CHECK7-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
6358 // CHECK7-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !11
6359 // CHECK7-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
6360 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !11
6361 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6362 // CHECK7: omp.body.continue:
6363 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6364 // CHECK7: omp.inner.for.inc:
6365 // CHECK7-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
6366 // CHECK7-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
6367 // CHECK7-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
6368 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
6369 // CHECK7: omp.inner.for.end:
6370 // CHECK7-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
6371 // CHECK7: omp.dispatch.inc:
6372 // CHECK7-NEXT: br label [[OMP_DISPATCH_COND]]
6373 // CHECK7: omp.dispatch.end:
6374 // CHECK7-NEXT: ret void
6375 //
6376 //
6377 // CHECK7-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57
6378 // CHECK7-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6379 // CHECK7-NEXT: entry:
6380 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6381 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6382 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6383 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
6384 // CHECK7-NEXT: ret void
6385 //
6386 //
6387 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..14
6388 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6389 // CHECK7-NEXT: entry:
6390 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6391 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6392 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6393 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6394 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
6395 // CHECK7-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6396 // CHECK7-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6397 // CHECK7-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6398 // CHECK7-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6399 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
6400 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6401 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6402 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6403 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6404 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6405 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
6406 // CHECK7-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6407 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6408 // CHECK7-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6409 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6410 // CHECK7-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6411 // CHECK7-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6412 // CHECK7-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
6413 // CHECK7-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6414 // CHECK7: cond.true:
6415 // CHECK7-NEXT: br label [[COND_END:%.*]]
6416 // CHECK7: cond.false:
6417 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6418 // CHECK7-NEXT: br label [[COND_END]]
6419 // CHECK7: cond.end:
6420 // CHECK7-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
6421 // CHECK7-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6422 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6423 // CHECK7-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
6424 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6425 // CHECK7: omp.inner.for.cond:
6426 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6427 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6428 // CHECK7-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6429 // CHECK7-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6430 // CHECK7: omp.inner.for.body:
6431 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6432 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6433 // CHECK7-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
6434 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6435 // CHECK7: omp.inner.for.inc:
6436 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6437 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6438 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
6439 // CHECK7-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6440 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]]
6441 // CHECK7: omp.inner.for.end:
6442 // CHECK7-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6443 // CHECK7: omp.loop.exit:
6444 // CHECK7-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
6445 // CHECK7-NEXT: ret void
6446 //
6447 //
6448 // CHECK7-LABEL: define {{[^@]+}}@.omp_outlined..15
6449 // CHECK7-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6450 // CHECK7-NEXT: entry:
6451 // CHECK7-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6452 // CHECK7-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6453 // CHECK7-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
6454 // CHECK7-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
6455 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6456 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6457 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
6458 // CHECK7-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6459 // CHECK7-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6460 // CHECK7-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6461 // CHECK7-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6462 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
6463 // CHECK7-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6464 // CHECK7-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6465 // CHECK7-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6466 // CHECK7-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6467 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6468 // CHECK7-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6469 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6470 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
6471 // CHECK7-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6472 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6473 // CHECK7-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
6474 // CHECK7-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
6475 // CHECK7-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6476 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6477 // CHECK7-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6478 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6479 // CHECK7-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6480 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
6481 // CHECK7-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
6482 // CHECK7-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
6483 // CHECK7: omp.dispatch.cond:
6484 // CHECK7-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
6485 // CHECK7-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
6486 // CHECK7-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
6487 // CHECK7: omp.dispatch.body:
6488 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6489 // CHECK7-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
6490 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6491 // CHECK7: omp.inner.for.cond:
6492 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
6493 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
6494 // CHECK7-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
6495 // CHECK7-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6496 // CHECK7: omp.inner.for.body:
6497 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
6498 // CHECK7-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
6499 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6500 // CHECK7-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !14
6501 // CHECK7-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
6502 // CHECK7-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !14
6503 // CHECK7-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
6504 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !14
6505 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6506 // CHECK7: omp.body.continue:
6507 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6508 // CHECK7: omp.inner.for.inc:
6509 // CHECK7-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
6510 // CHECK7-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
6511 // CHECK7-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
6512 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
6513 // CHECK7: omp.inner.for.end:
6514 // CHECK7-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
6515 // CHECK7: omp.dispatch.inc:
6516 // CHECK7-NEXT: br label [[OMP_DISPATCH_COND]]
6517 // CHECK7: omp.dispatch.end:
6518 // CHECK7-NEXT: ret void
6519 //
6520 //
6521 // CHECK7-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
6522 // CHECK7-SAME: () #[[ATTR3:[0-9]+]] {
6523 // CHECK7-NEXT: entry:
6524 // CHECK7-NEXT: call void @__tgt_register_requires(i64 1)
6525 // CHECK7-NEXT: ret void
6526 //
6527 //
6528 // CHECK8-LABEL: define {{[^@]+}}@_Z21teams_template_structv
6529 // CHECK8-SAME: () #[[ATTR0:[0-9]+]] {
6530 // CHECK8-NEXT: entry:
6531 // CHECK8-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
6532 // CHECK8-NEXT: [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
6533 // CHECK8-NEXT: ret i32 [[CALL]]
6534 //
6535 //
6536 // CHECK8-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
6537 // CHECK8-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
6538 // CHECK8-NEXT: entry:
6539 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6540 // CHECK8-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
6541 // CHECK8-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
6542 // CHECK8-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
6543 // CHECK8-NEXT: [[TMP:%.*]] = alloca i32, align 4
6544 // CHECK8-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
6545 // CHECK8-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
6546 // CHECK8-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
6547 // CHECK8-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
6548 // CHECK8-NEXT: [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 4
6549 // CHECK8-NEXT: [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 4
6550 // CHECK8-NEXT: [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 4
6551 // CHECK8-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
6552 // CHECK8-NEXT: [[DOTOFFLOAD_BASEPTRS17:%.*]] = alloca [1 x i8*], align 4
6553 // CHECK8-NEXT: [[DOTOFFLOAD_PTRS18:%.*]] = alloca [1 x i8*], align 4
6554 // CHECK8-NEXT: [[DOTOFFLOAD_MAPPERS19:%.*]] = alloca [1 x i8*], align 4
6555 // CHECK8-NEXT: [[_TMP20:%.*]] = alloca i32, align 4
6556 // CHECK8-NEXT: [[DOTOFFLOAD_BASEPTRS24:%.*]] = alloca [1 x i8*], align 4
6557 // CHECK8-NEXT: [[DOTOFFLOAD_PTRS25:%.*]] = alloca [1 x i8*], align 4
6558 // CHECK8-NEXT: [[DOTOFFLOAD_MAPPERS26:%.*]] = alloca [1 x i8*], align 4
6559 // CHECK8-NEXT: [[_TMP27:%.*]] = alloca i32, align 4
6560 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6561 // CHECK8-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6562 // CHECK8-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
6563 // CHECK8-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6564 // CHECK8-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
6565 // CHECK8-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
6566 // CHECK8-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6567 // CHECK8-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
6568 // CHECK8-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
6569 // CHECK8-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
6570 // CHECK8-NEXT: store i8* null, i8** [[TMP4]], align 4
6571 // CHECK8-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6572 // CHECK8-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6573 // CHECK8-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
6574 // CHECK8-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
6575 // CHECK8-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
6576 // CHECK8-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6577 // CHECK8: omp_offload.failed:
6578 // CHECK8-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
6579 // CHECK8-NEXT: br label [[OMP_OFFLOAD_CONT]]
6580 // CHECK8: omp_offload.cont:
6581 // CHECK8-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
6582 // CHECK8-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
6583 // CHECK8-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
6584 // CHECK8-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 4
6585 // CHECK8-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
6586 // CHECK8-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
6587 // CHECK8-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 4
6588 // CHECK8-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
6589 // CHECK8-NEXT: store i8* null, i8** [[TMP13]], align 4
6590 // CHECK8-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
6591 // CHECK8-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
6592 // CHECK8-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
6593 // CHECK8-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
6594 // CHECK8-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
6595 // CHECK8-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
6596 // CHECK8: omp_offload.failed7:
6597 // CHECK8-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40(%struct.SS* [[THIS1]]) #[[ATTR2]]
6598 // CHECK8-NEXT: br label [[OMP_OFFLOAD_CONT8]]
6599 // CHECK8: omp_offload.cont8:
6600 // CHECK8-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
6601 // CHECK8-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
6602 // CHECK8-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
6603 // CHECK8-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
6604 // CHECK8-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
6605 // CHECK8-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
6606 // CHECK8-NEXT: store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 4
6607 // CHECK8-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i32 0, i32 0
6608 // CHECK8-NEXT: store i8* null, i8** [[TMP22]], align 4
6609 // CHECK8-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
6610 // CHECK8-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
6611 // CHECK8-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
6612 // CHECK8-NEXT: [[TMP25:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45.region_id, i32 1, i8** [[TMP23]], i8** [[TMP24]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
6613 // CHECK8-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
6614 // CHECK8-NEXT: br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
6615 // CHECK8: omp_offload.failed14:
6616 // CHECK8-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45(%struct.SS* [[THIS1]]) #[[ATTR2]]
6617 // CHECK8-NEXT: br label [[OMP_OFFLOAD_CONT15]]
6618 // CHECK8: omp_offload.cont15:
6619 // CHECK8-NEXT: [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
6620 // CHECK8-NEXT: [[TMP27:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
6621 // CHECK8-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to %struct.SS**
6622 // CHECK8-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP28]], align 4
6623 // CHECK8-NEXT: [[TMP29:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
6624 // CHECK8-NEXT: [[TMP30:%.*]] = bitcast i8** [[TMP29]] to [123 x i32]**
6625 // CHECK8-NEXT: store [123 x i32]* [[A16]], [123 x i32]** [[TMP30]], align 4
6626 // CHECK8-NEXT: [[TMP31:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS19]], i32 0, i32 0
6627 // CHECK8-NEXT: store i8* null, i8** [[TMP31]], align 4
6628 // CHECK8-NEXT: [[TMP32:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS17]], i32 0, i32 0
6629 // CHECK8-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS18]], i32 0, i32 0
6630 // CHECK8-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
6631 // CHECK8-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51.region_id, i32 1, i8** [[TMP32]], i8** [[TMP33]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
6632 // CHECK8-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
6633 // CHECK8-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED21:%.*]], label [[OMP_OFFLOAD_CONT22:%.*]]
6634 // CHECK8: omp_offload.failed21:
6635 // CHECK8-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51(%struct.SS* [[THIS1]]) #[[ATTR2]]
6636 // CHECK8-NEXT: br label [[OMP_OFFLOAD_CONT22]]
6637 // CHECK8: omp_offload.cont22:
6638 // CHECK8-NEXT: [[A23:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
6639 // CHECK8-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
6640 // CHECK8-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
6641 // CHECK8-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
6642 // CHECK8-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
6643 // CHECK8-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
6644 // CHECK8-NEXT: store [123 x i32]* [[A23]], [123 x i32]** [[TMP39]], align 4
6645 // CHECK8-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS26]], i32 0, i32 0
6646 // CHECK8-NEXT: store i8* null, i8** [[TMP40]], align 4
6647 // CHECK8-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS24]], i32 0, i32 0
6648 // CHECK8-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS25]], i32 0, i32 0
6649 // CHECK8-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
6650 // CHECK8-NEXT: [[TMP43:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57.region_id, i32 1, i8** [[TMP41]], i8** [[TMP42]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
6651 // CHECK8-NEXT: [[TMP44:%.*]] = icmp ne i32 [[TMP43]], 0
6652 // CHECK8-NEXT: br i1 [[TMP44]], label [[OMP_OFFLOAD_FAILED28:%.*]], label [[OMP_OFFLOAD_CONT29:%.*]]
6653 // CHECK8: omp_offload.failed28:
6654 // CHECK8-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57(%struct.SS* [[THIS1]]) #[[ATTR2]]
6655 // CHECK8-NEXT: br label [[OMP_OFFLOAD_CONT29]]
6656 // CHECK8: omp_offload.cont29:
6657 // CHECK8-NEXT: [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
6658 // CHECK8-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i32 0, i32 0
6659 // CHECK8-NEXT: [[TMP45:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
6660 // CHECK8-NEXT: ret i32 [[TMP45]]
6661 //
6662 //
6663 // CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l35
6664 // CHECK8-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
6665 // CHECK8-NEXT: entry:
6666 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6667 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6668 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6669 // CHECK8-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
6670 // CHECK8-NEXT: ret void
6671 //
6672 //
6673 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined.
6674 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6675 // CHECK8-NEXT: entry:
6676 // CHECK8-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6677 // CHECK8-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6678 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6679 // CHECK8-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6680 // CHECK8-NEXT: [[TMP:%.*]] = alloca i32, align 4
6681 // CHECK8-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6682 // CHECK8-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6683 // CHECK8-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6684 // CHECK8-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6685 // CHECK8-NEXT: [[I:%.*]] = alloca i32, align 4
6686 // CHECK8-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6687 // CHECK8-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6688 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6689 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6690 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6691 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
6692 // CHECK8-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6693 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6694 // CHECK8-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6695 // CHECK8-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6696 // CHECK8-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6697 // CHECK8-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6698 // CHECK8-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
6699 // CHECK8-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6700 // CHECK8: cond.true:
6701 // CHECK8-NEXT: br label [[COND_END:%.*]]
6702 // CHECK8: cond.false:
6703 // CHECK8-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6704 // CHECK8-NEXT: br label [[COND_END]]
6705 // CHECK8: cond.end:
6706 // CHECK8-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
6707 // CHECK8-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6708 // CHECK8-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6709 // CHECK8-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
6710 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6711 // CHECK8: omp.inner.for.cond:
6712 // CHECK8-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6713 // CHECK8-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6714 // CHECK8-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6715 // CHECK8-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6716 // CHECK8: omp.inner.for.body:
6717 // CHECK8-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6718 // CHECK8-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6719 // CHECK8-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
6720 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6721 // CHECK8: omp.inner.for.inc:
6722 // CHECK8-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6723 // CHECK8-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6724 // CHECK8-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
6725 // CHECK8-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6726 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND]]
6727 // CHECK8: omp.inner.for.end:
6728 // CHECK8-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6729 // CHECK8: omp.loop.exit:
6730 // CHECK8-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
6731 // CHECK8-NEXT: ret void
6732 //
6733 //
6734 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..1
6735 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6736 // CHECK8-NEXT: entry:
6737 // CHECK8-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6738 // CHECK8-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6739 // CHECK8-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
6740 // CHECK8-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
6741 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6742 // CHECK8-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6743 // CHECK8-NEXT: [[TMP:%.*]] = alloca i32, align 4
6744 // CHECK8-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6745 // CHECK8-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6746 // CHECK8-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6747 // CHECK8-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6748 // CHECK8-NEXT: [[I:%.*]] = alloca i32, align 4
6749 // CHECK8-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6750 // CHECK8-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6751 // CHECK8-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6752 // CHECK8-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6753 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6754 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6755 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6756 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
6757 // CHECK8-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6758 // CHECK8-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6759 // CHECK8-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
6760 // CHECK8-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
6761 // CHECK8-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6762 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6763 // CHECK8-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6764 // CHECK8-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
6765 // CHECK8-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6766 // CHECK8-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6767 // CHECK8-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
6768 // CHECK8-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6769 // CHECK8: cond.true:
6770 // CHECK8-NEXT: br label [[COND_END:%.*]]
6771 // CHECK8: cond.false:
6772 // CHECK8-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6773 // CHECK8-NEXT: br label [[COND_END]]
6774 // CHECK8: cond.end:
6775 // CHECK8-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
6776 // CHECK8-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6777 // CHECK8-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6778 // CHECK8-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
6779 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6780 // CHECK8: omp.inner.for.cond:
6781 // CHECK8-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6782 // CHECK8-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6783 // CHECK8-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
6784 // CHECK8-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6785 // CHECK8: omp.inner.for.body:
6786 // CHECK8-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6787 // CHECK8-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
6788 // CHECK8-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6789 // CHECK8-NEXT: store i32 [[ADD]], i32* [[I]], align 4
6790 // CHECK8-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
6791 // CHECK8-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
6792 // CHECK8-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
6793 // CHECK8-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
6794 // CHECK8-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6795 // CHECK8: omp.body.continue:
6796 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6797 // CHECK8: omp.inner.for.inc:
6798 // CHECK8-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6799 // CHECK8-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
6800 // CHECK8-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
6801 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND]]
6802 // CHECK8: omp.inner.for.end:
6803 // CHECK8-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6804 // CHECK8: omp.loop.exit:
6805 // CHECK8-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
6806 // CHECK8-NEXT: ret void
6807 //
6808 //
6809 // CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l40
6810 // CHECK8-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6811 // CHECK8-NEXT: entry:
6812 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6813 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6814 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6815 // CHECK8-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
6816 // CHECK8-NEXT: ret void
6817 //
6818 //
6819 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..2
6820 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6821 // CHECK8-NEXT: entry:
6822 // CHECK8-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6823 // CHECK8-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6824 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6825 // CHECK8-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6826 // CHECK8-NEXT: [[TMP:%.*]] = alloca i32, align 4
6827 // CHECK8-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6828 // CHECK8-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6829 // CHECK8-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6830 // CHECK8-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6831 // CHECK8-NEXT: [[I:%.*]] = alloca i32, align 4
6832 // CHECK8-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6833 // CHECK8-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6834 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6835 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6836 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6837 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
6838 // CHECK8-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6839 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6840 // CHECK8-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6841 // CHECK8-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6842 // CHECK8-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6843 // CHECK8-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6844 // CHECK8-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
6845 // CHECK8-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6846 // CHECK8: cond.true:
6847 // CHECK8-NEXT: br label [[COND_END:%.*]]
6848 // CHECK8: cond.false:
6849 // CHECK8-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6850 // CHECK8-NEXT: br label [[COND_END]]
6851 // CHECK8: cond.end:
6852 // CHECK8-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
6853 // CHECK8-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6854 // CHECK8-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6855 // CHECK8-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
6856 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6857 // CHECK8: omp.inner.for.cond:
6858 // CHECK8-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6859 // CHECK8-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6860 // CHECK8-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6861 // CHECK8-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6862 // CHECK8: omp.inner.for.body:
6863 // CHECK8-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6864 // CHECK8-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6865 // CHECK8-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
6866 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6867 // CHECK8: omp.inner.for.inc:
6868 // CHECK8-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6869 // CHECK8-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6870 // CHECK8-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
6871 // CHECK8-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6872 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND]]
6873 // CHECK8: omp.inner.for.end:
6874 // CHECK8-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6875 // CHECK8: omp.loop.exit:
6876 // CHECK8-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
6877 // CHECK8-NEXT: ret void
6878 //
6879 //
6880 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..3
6881 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6882 // CHECK8-NEXT: entry:
6883 // CHECK8-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6884 // CHECK8-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6885 // CHECK8-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
6886 // CHECK8-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
6887 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6888 // CHECK8-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6889 // CHECK8-NEXT: [[TMP:%.*]] = alloca i32, align 4
6890 // CHECK8-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6891 // CHECK8-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6892 // CHECK8-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6893 // CHECK8-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6894 // CHECK8-NEXT: [[I:%.*]] = alloca i32, align 4
6895 // CHECK8-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6896 // CHECK8-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6897 // CHECK8-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6898 // CHECK8-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6899 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6900 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6901 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6902 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
6903 // CHECK8-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6904 // CHECK8-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6905 // CHECK8-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
6906 // CHECK8-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
6907 // CHECK8-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6908 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6909 // CHECK8-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6910 // CHECK8-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
6911 // CHECK8-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6912 // CHECK8-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6913 // CHECK8-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
6914 // CHECK8-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6915 // CHECK8: cond.true:
6916 // CHECK8-NEXT: br label [[COND_END:%.*]]
6917 // CHECK8: cond.false:
6918 // CHECK8-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6919 // CHECK8-NEXT: br label [[COND_END]]
6920 // CHECK8: cond.end:
6921 // CHECK8-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
6922 // CHECK8-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6923 // CHECK8-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6924 // CHECK8-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
6925 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6926 // CHECK8: omp.inner.for.cond:
6927 // CHECK8-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6928 // CHECK8-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6929 // CHECK8-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
6930 // CHECK8-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6931 // CHECK8: omp.inner.for.body:
6932 // CHECK8-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6933 // CHECK8-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
6934 // CHECK8-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6935 // CHECK8-NEXT: store i32 [[ADD]], i32* [[I]], align 4
6936 // CHECK8-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
6937 // CHECK8-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
6938 // CHECK8-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
6939 // CHECK8-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
6940 // CHECK8-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6941 // CHECK8: omp.body.continue:
6942 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6943 // CHECK8: omp.inner.for.inc:
6944 // CHECK8-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6945 // CHECK8-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
6946 // CHECK8-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
6947 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND]]
6948 // CHECK8: omp.inner.for.end:
6949 // CHECK8-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6950 // CHECK8: omp.loop.exit:
6951 // CHECK8-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
6952 // CHECK8-NEXT: ret void
6953 //
6954 //
6955 // CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l45
6956 // CHECK8-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6957 // CHECK8-NEXT: entry:
6958 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6959 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6960 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6961 // CHECK8-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
6962 // CHECK8-NEXT: ret void
6963 //
6964 //
6965 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..6
6966 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
6967 // CHECK8-NEXT: entry:
6968 // CHECK8-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6969 // CHECK8-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6970 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
6971 // CHECK8-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6972 // CHECK8-NEXT: [[TMP:%.*]] = alloca i32, align 4
6973 // CHECK8-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6974 // CHECK8-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6975 // CHECK8-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6976 // CHECK8-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6977 // CHECK8-NEXT: [[I:%.*]] = alloca i32, align 4
6978 // CHECK8-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6979 // CHECK8-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6980 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
6981 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
6982 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6983 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
6984 // CHECK8-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6985 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6986 // CHECK8-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6987 // CHECK8-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6988 // CHECK8-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6989 // CHECK8-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6990 // CHECK8-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
6991 // CHECK8-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6992 // CHECK8: cond.true:
6993 // CHECK8-NEXT: br label [[COND_END:%.*]]
6994 // CHECK8: cond.false:
6995 // CHECK8-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6996 // CHECK8-NEXT: br label [[COND_END]]
6997 // CHECK8: cond.end:
6998 // CHECK8-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
6999 // CHECK8-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7000 // CHECK8-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7001 // CHECK8-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
7002 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7003 // CHECK8: omp.inner.for.cond:
7004 // CHECK8-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7005 // CHECK8-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7006 // CHECK8-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7007 // CHECK8-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7008 // CHECK8: omp.inner.for.body:
7009 // CHECK8-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7010 // CHECK8-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7011 // CHECK8-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
7012 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7013 // CHECK8: omp.inner.for.inc:
7014 // CHECK8-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7015 // CHECK8-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7016 // CHECK8-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
7017 // CHECK8-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7018 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND]]
7019 // CHECK8: omp.inner.for.end:
7020 // CHECK8-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7021 // CHECK8: omp.loop.exit:
7022 // CHECK8-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
7023 // CHECK8-NEXT: ret void
7024 //
7025 //
7026 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..7
7027 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
7028 // CHECK8-NEXT: entry:
7029 // CHECK8-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7030 // CHECK8-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7031 // CHECK8-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
7032 // CHECK8-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
7033 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
7034 // CHECK8-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7035 // CHECK8-NEXT: [[TMP:%.*]] = alloca i32, align 4
7036 // CHECK8-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7037 // CHECK8-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7038 // CHECK8-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7039 // CHECK8-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7040 // CHECK8-NEXT: [[I:%.*]] = alloca i32, align 4
7041 // CHECK8-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7042 // CHECK8-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7043 // CHECK8-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7044 // CHECK8-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7045 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
7046 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
7047 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7048 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
7049 // CHECK8-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7050 // CHECK8-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7051 // CHECK8-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
7052 // CHECK8-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
7053 // CHECK8-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7054 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7055 // CHECK8-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7056 // CHECK8-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
7057 // CHECK8-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
7058 // CHECK8-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
7059 // CHECK8: omp.dispatch.cond:
7060 // CHECK8-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7061 // CHECK8-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7062 // CHECK8-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], [[TMP6]]
7063 // CHECK8-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7064 // CHECK8: cond.true:
7065 // CHECK8-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7066 // CHECK8-NEXT: br label [[COND_END:%.*]]
7067 // CHECK8: cond.false:
7068 // CHECK8-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7069 // CHECK8-NEXT: br label [[COND_END]]
7070 // CHECK8: cond.end:
7071 // CHECK8-NEXT: [[COND:%.*]] = phi i32 [ [[TMP7]], [[COND_TRUE]] ], [ [[TMP8]], [[COND_FALSE]] ]
7072 // CHECK8-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7073 // CHECK8-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7074 // CHECK8-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
7075 // CHECK8-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7076 // CHECK8-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7077 // CHECK8-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
7078 // CHECK8-NEXT: br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
7079 // CHECK8: omp.dispatch.body:
7080 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7081 // CHECK8: omp.inner.for.cond:
7082 // CHECK8-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7083 // CHECK8-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7084 // CHECK8-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
7085 // CHECK8-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7086 // CHECK8: omp.inner.for.body:
7087 // CHECK8-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7088 // CHECK8-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
7089 // CHECK8-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7090 // CHECK8-NEXT: store i32 [[ADD]], i32* [[I]], align 4
7091 // CHECK8-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
7092 // CHECK8-NEXT: [[TMP15:%.*]] = load i32, i32* [[I]], align 4
7093 // CHECK8-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP15]]
7094 // CHECK8-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
7095 // CHECK8-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7096 // CHECK8: omp.body.continue:
7097 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7098 // CHECK8: omp.inner.for.inc:
7099 // CHECK8-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7100 // CHECK8-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], 1
7101 // CHECK8-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
7102 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND]]
7103 // CHECK8: omp.inner.for.end:
7104 // CHECK8-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
7105 // CHECK8: omp.dispatch.inc:
7106 // CHECK8-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7107 // CHECK8-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7108 // CHECK8-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
7109 // CHECK8-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
7110 // CHECK8-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7111 // CHECK8-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7112 // CHECK8-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP19]], [[TMP20]]
7113 // CHECK8-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
7114 // CHECK8-NEXT: br label [[OMP_DISPATCH_COND]]
7115 // CHECK8: omp.dispatch.end:
7116 // CHECK8-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
7117 // CHECK8-NEXT: ret void
7118 //
7119 //
7120 // CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l51
7121 // CHECK8-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
7122 // CHECK8-NEXT: entry:
7123 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
7124 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
7125 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
7126 // CHECK8-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
7127 // CHECK8-NEXT: ret void
7128 //
7129 //
7130 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..10
7131 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
7132 // CHECK8-NEXT: entry:
7133 // CHECK8-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7134 // CHECK8-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7135 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
7136 // CHECK8-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7137 // CHECK8-NEXT: [[TMP:%.*]] = alloca i32, align 4
7138 // CHECK8-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7139 // CHECK8-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7140 // CHECK8-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7141 // CHECK8-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7142 // CHECK8-NEXT: [[I:%.*]] = alloca i32, align 4
7143 // CHECK8-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7144 // CHECK8-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7145 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
7146 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
7147 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7148 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
7149 // CHECK8-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7150 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7151 // CHECK8-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7152 // CHECK8-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
7153 // CHECK8-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7154 // CHECK8-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7155 // CHECK8-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
7156 // CHECK8-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7157 // CHECK8: cond.true:
7158 // CHECK8-NEXT: br label [[COND_END:%.*]]
7159 // CHECK8: cond.false:
7160 // CHECK8-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7161 // CHECK8-NEXT: br label [[COND_END]]
7162 // CHECK8: cond.end:
7163 // CHECK8-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
7164 // CHECK8-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7165 // CHECK8-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7166 // CHECK8-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
7167 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7168 // CHECK8: omp.inner.for.cond:
7169 // CHECK8-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7170 // CHECK8-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7171 // CHECK8-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7172 // CHECK8-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7173 // CHECK8: omp.inner.for.body:
7174 // CHECK8-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7175 // CHECK8-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7176 // CHECK8-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
7177 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7178 // CHECK8: omp.inner.for.inc:
7179 // CHECK8-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7180 // CHECK8-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7181 // CHECK8-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
7182 // CHECK8-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7183 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND]]
7184 // CHECK8: omp.inner.for.end:
7185 // CHECK8-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7186 // CHECK8: omp.loop.exit:
7187 // CHECK8-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
7188 // CHECK8-NEXT: ret void
7189 //
7190 //
7191 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..11
7192 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
7193 // CHECK8-NEXT: entry:
7194 // CHECK8-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7195 // CHECK8-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7196 // CHECK8-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
7197 // CHECK8-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
7198 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
7199 // CHECK8-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7200 // CHECK8-NEXT: [[TMP:%.*]] = alloca i32, align 4
7201 // CHECK8-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7202 // CHECK8-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7203 // CHECK8-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7204 // CHECK8-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7205 // CHECK8-NEXT: [[I:%.*]] = alloca i32, align 4
7206 // CHECK8-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7207 // CHECK8-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7208 // CHECK8-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7209 // CHECK8-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7210 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
7211 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
7212 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7213 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
7214 // CHECK8-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7215 // CHECK8-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7216 // CHECK8-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
7217 // CHECK8-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
7218 // CHECK8-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7219 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7220 // CHECK8-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7221 // CHECK8-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7222 // CHECK8-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7223 // CHECK8-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
7224 // CHECK8-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
7225 // CHECK8-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
7226 // CHECK8: omp.dispatch.cond:
7227 // CHECK8-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
7228 // CHECK8-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
7229 // CHECK8-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
7230 // CHECK8: omp.dispatch.body:
7231 // CHECK8-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7232 // CHECK8-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
7233 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7234 // CHECK8: omp.inner.for.cond:
7235 // CHECK8-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
7236 // CHECK8-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
7237 // CHECK8-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
7238 // CHECK8-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7239 // CHECK8: omp.inner.for.body:
7240 // CHECK8-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
7241 // CHECK8-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
7242 // CHECK8-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7243 // CHECK8-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !11
7244 // CHECK8-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
7245 // CHECK8-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !11
7246 // CHECK8-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
7247 // CHECK8-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !11
7248 // CHECK8-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7249 // CHECK8: omp.body.continue:
7250 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7251 // CHECK8: omp.inner.for.inc:
7252 // CHECK8-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
7253 // CHECK8-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
7254 // CHECK8-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
7255 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
7256 // CHECK8: omp.inner.for.end:
7257 // CHECK8-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
7258 // CHECK8: omp.dispatch.inc:
7259 // CHECK8-NEXT: br label [[OMP_DISPATCH_COND]]
7260 // CHECK8: omp.dispatch.end:
7261 // CHECK8-NEXT: ret void
7262 //
7263 //
7264 // CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l57
7265 // CHECK8-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
7266 // CHECK8-NEXT: entry:
7267 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
7268 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
7269 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
7270 // CHECK8-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
7271 // CHECK8-NEXT: ret void
7272 //
7273 //
7274 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..14
7275 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
7276 // CHECK8-NEXT: entry:
7277 // CHECK8-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7278 // CHECK8-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7279 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
7280 // CHECK8-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7281 // CHECK8-NEXT: [[TMP:%.*]] = alloca i32, align 4
7282 // CHECK8-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7283 // CHECK8-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7284 // CHECK8-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7285 // CHECK8-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7286 // CHECK8-NEXT: [[I:%.*]] = alloca i32, align 4
7287 // CHECK8-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7288 // CHECK8-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7289 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
7290 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
7291 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7292 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
7293 // CHECK8-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7294 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7295 // CHECK8-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7296 // CHECK8-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
7297 // CHECK8-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7298 // CHECK8-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7299 // CHECK8-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
7300 // CHECK8-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7301 // CHECK8: cond.true:
7302 // CHECK8-NEXT: br label [[COND_END:%.*]]
7303 // CHECK8: cond.false:
7304 // CHECK8-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7305 // CHECK8-NEXT: br label [[COND_END]]
7306 // CHECK8: cond.end:
7307 // CHECK8-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
7308 // CHECK8-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7309 // CHECK8-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7310 // CHECK8-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
7311 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7312 // CHECK8: omp.inner.for.cond:
7313 // CHECK8-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7314 // CHECK8-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7315 // CHECK8-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7316 // CHECK8-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7317 // CHECK8: omp.inner.for.body:
7318 // CHECK8-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7319 // CHECK8-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7320 // CHECK8-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
7321 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7322 // CHECK8: omp.inner.for.inc:
7323 // CHECK8-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7324 // CHECK8-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7325 // CHECK8-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
7326 // CHECK8-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7327 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND]]
7328 // CHECK8: omp.inner.for.end:
7329 // CHECK8-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7330 // CHECK8: omp.loop.exit:
7331 // CHECK8-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
7332 // CHECK8-NEXT: ret void
7333 //
7334 //
7335 // CHECK8-LABEL: define {{[^@]+}}@.omp_outlined..15
7336 // CHECK8-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
7337 // CHECK8-NEXT: entry:
7338 // CHECK8-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7339 // CHECK8-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7340 // CHECK8-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
7341 // CHECK8-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
7342 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
7343 // CHECK8-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7344 // CHECK8-NEXT: [[TMP:%.*]] = alloca i32, align 4
7345 // CHECK8-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7346 // CHECK8-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7347 // CHECK8-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7348 // CHECK8-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7349 // CHECK8-NEXT: [[I:%.*]] = alloca i32, align 4
7350 // CHECK8-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7351 // CHECK8-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7352 // CHECK8-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7353 // CHECK8-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7354 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
7355 // CHECK8-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
7356 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7357 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
7358 // CHECK8-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7359 // CHECK8-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7360 // CHECK8-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
7361 // CHECK8-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
7362 // CHECK8-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7363 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7364 // CHECK8-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7365 // CHECK8-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7366 // CHECK8-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7367 // CHECK8-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
7368 // CHECK8-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 61)
7369 // CHECK8-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
7370 // CHECK8: omp.dispatch.cond:
7371 // CHECK8-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
7372 // CHECK8-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
7373 // CHECK8-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
7374 // CHECK8: omp.dispatch.body:
7375 // CHECK8-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7376 // CHECK8-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
7377 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7378 // CHECK8: omp.inner.for.cond:
7379 // CHECK8-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
7380 // CHECK8-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
7381 // CHECK8-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
7382 // CHECK8-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7383 // CHECK8: omp.inner.for.body:
7384 // CHECK8-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
7385 // CHECK8-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
7386 // CHECK8-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7387 // CHECK8-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !14
7388 // CHECK8-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
7389 // CHECK8-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !14
7390 // CHECK8-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP12]]
7391 // CHECK8-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !14
7392 // CHECK8-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7393 // CHECK8: omp.body.continue:
7394 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7395 // CHECK8: omp.inner.for.inc:
7396 // CHECK8-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
7397 // CHECK8-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
7398 // CHECK8-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
7399 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
7400 // CHECK8: omp.inner.for.end:
7401 // CHECK8-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
7402 // CHECK8: omp.dispatch.inc:
7403 // CHECK8-NEXT: br label [[OMP_DISPATCH_COND]]
7404 // CHECK8: omp.dispatch.end:
7405 // CHECK8-NEXT: ret void
7406 //
7407 //
7408 // CHECK8-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
7409 // CHECK8-SAME: () #[[ATTR3:[0-9]+]] {
7410 // CHECK8-NEXT: entry:
7411 // CHECK8-NEXT: call void @__tgt_register_requires(i64 1)
7412 // CHECK8-NEXT: ret void
7413 //
7414 //
7415 // CHECK13-LABEL: define {{[^@]+}}@main
7416 // CHECK13-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
7417 // CHECK13-NEXT: entry:
7418 // CHECK13-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
7419 // CHECK13-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
7420 // CHECK13-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
7421 // CHECK13-NEXT: [[N:%.*]] = alloca i32, align 4
7422 // CHECK13-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
7423 // CHECK13-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
7424 // CHECK13-NEXT: [[M:%.*]] = alloca i32, align 4
7425 // CHECK13-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
7426 // CHECK13-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
7427 // CHECK13-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
7428 // CHECK13-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
7429 // CHECK13-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
7430 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
7431 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
7432 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
7433 // CHECK13-NEXT: [[N_CASTED3:%.*]] = alloca i64, align 8
7434 // CHECK13-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
7435 // CHECK13-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
7436 // CHECK13-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
7437 // CHECK13-NEXT: [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
7438 // CHECK13-NEXT: [[_TMP9:%.*]] = alloca i32, align 4
7439 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
7440 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
7441 // CHECK13-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
7442 // CHECK13-NEXT: [[N_CASTED19:%.*]] = alloca i64, align 8
7443 // CHECK13-NEXT: [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [4 x i8*], align 8
7444 // CHECK13-NEXT: [[DOTOFFLOAD_PTRS22:%.*]] = alloca [4 x i8*], align 8
7445 // CHECK13-NEXT: [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [4 x i8*], align 8
7446 // CHECK13-NEXT: [[DOTOFFLOAD_SIZES24:%.*]] = alloca [4 x i64], align 8
7447 // CHECK13-NEXT: [[_TMP25:%.*]] = alloca i32, align 4
7448 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
7449 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
7450 // CHECK13-NEXT: [[N_CASTED34:%.*]] = alloca i64, align 8
7451 // CHECK13-NEXT: [[DOTOFFLOAD_BASEPTRS36:%.*]] = alloca [3 x i8*], align 8
7452 // CHECK13-NEXT: [[DOTOFFLOAD_PTRS37:%.*]] = alloca [3 x i8*], align 8
7453 // CHECK13-NEXT: [[DOTOFFLOAD_MAPPERS38:%.*]] = alloca [3 x i8*], align 8
7454 // CHECK13-NEXT: [[DOTOFFLOAD_SIZES39:%.*]] = alloca [3 x i64], align 8
7455 // CHECK13-NEXT: [[_TMP40:%.*]] = alloca i32, align 4
7456 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
7457 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
7458 // CHECK13-NEXT: [[M_CASTED49:%.*]] = alloca i64, align 8
7459 // CHECK13-NEXT: [[N_CASTED51:%.*]] = alloca i64, align 8
7460 // CHECK13-NEXT: [[DOTOFFLOAD_BASEPTRS53:%.*]] = alloca [4 x i8*], align 8
7461 // CHECK13-NEXT: [[DOTOFFLOAD_PTRS54:%.*]] = alloca [4 x i8*], align 8
7462 // CHECK13-NEXT: [[DOTOFFLOAD_MAPPERS55:%.*]] = alloca [4 x i8*], align 8
7463 // CHECK13-NEXT: [[DOTOFFLOAD_SIZES56:%.*]] = alloca [4 x i64], align 8
7464 // CHECK13-NEXT: [[_TMP57:%.*]] = alloca i32, align 4
7465 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_58:%.*]] = alloca i32, align 4
7466 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_59:%.*]] = alloca i32, align 4
7467 // CHECK13-NEXT: store i32 0, i32* [[RETVAL]], align 4
7468 // CHECK13-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
7469 // CHECK13-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
7470 // CHECK13-NEXT: store i32 100, i32* [[N]], align 4
7471 // CHECK13-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
7472 // CHECK13-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
7473 // CHECK13-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
7474 // CHECK13-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
7475 // CHECK13-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
7476 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
7477 // CHECK13-NEXT: store i32 10, i32* [[M]], align 4
7478 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
7479 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
7480 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
7481 // CHECK13-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
7482 // CHECK13-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
7483 // CHECK13-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7484 // CHECK13-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
7485 // CHECK13-NEXT: store i64 [[TMP4]], i64* [[TMP7]], align 8
7486 // CHECK13-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7487 // CHECK13-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
7488 // CHECK13-NEXT: store i64 [[TMP4]], i64* [[TMP9]], align 8
7489 // CHECK13-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
7490 // CHECK13-NEXT: store i64 4, i64* [[TMP10]], align 8
7491 // CHECK13-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
7492 // CHECK13-NEXT: store i8* null, i8** [[TMP11]], align 8
7493 // CHECK13-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
7494 // CHECK13-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
7495 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
7496 // CHECK13-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
7497 // CHECK13-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
7498 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
7499 // CHECK13-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
7500 // CHECK13-NEXT: store i64 8, i64* [[TMP16]], align 8
7501 // CHECK13-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
7502 // CHECK13-NEXT: store i8* null, i8** [[TMP17]], align 8
7503 // CHECK13-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
7504 // CHECK13-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
7505 // CHECK13-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 8
7506 // CHECK13-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
7507 // CHECK13-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
7508 // CHECK13-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 8
7509 // CHECK13-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
7510 // CHECK13-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 8
7511 // CHECK13-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
7512 // CHECK13-NEXT: store i8* null, i8** [[TMP23]], align 8
7513 // CHECK13-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7514 // CHECK13-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7515 // CHECK13-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
7516 // CHECK13-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
7517 // CHECK13-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
7518 // CHECK13-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7519 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
7520 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
7521 // CHECK13-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
7522 // CHECK13-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
7523 // CHECK13-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7524 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
7525 // CHECK13-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
7526 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
7527 // CHECK13-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
7528 // CHECK13-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
7529 // CHECK13-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
7530 // CHECK13: omp_offload.failed:
7531 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
7532 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT]]
7533 // CHECK13: omp_offload.cont:
7534 // CHECK13-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
7535 // CHECK13-NEXT: [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
7536 // CHECK13-NEXT: store i32 [[TMP33]], i32* [[CONV4]], align 4
7537 // CHECK13-NEXT: [[TMP34:%.*]] = load i64, i64* [[N_CASTED3]], align 8
7538 // CHECK13-NEXT: [[TMP35:%.*]] = mul nuw i64 [[TMP1]], 4
7539 // CHECK13-NEXT: [[TMP36:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
7540 // CHECK13-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
7541 // CHECK13-NEXT: store i64 [[TMP34]], i64* [[TMP37]], align 8
7542 // CHECK13-NEXT: [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
7543 // CHECK13-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
7544 // CHECK13-NEXT: store i64 [[TMP34]], i64* [[TMP39]], align 8
7545 // CHECK13-NEXT: [[TMP40:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
7546 // CHECK13-NEXT: store i64 4, i64* [[TMP40]], align 8
7547 // CHECK13-NEXT: [[TMP41:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
7548 // CHECK13-NEXT: store i8* null, i8** [[TMP41]], align 8
7549 // CHECK13-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
7550 // CHECK13-NEXT: [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
7551 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[TMP43]], align 8
7552 // CHECK13-NEXT: [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
7553 // CHECK13-NEXT: [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i64*
7554 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[TMP45]], align 8
7555 // CHECK13-NEXT: [[TMP46:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 1
7556 // CHECK13-NEXT: store i64 8, i64* [[TMP46]], align 8
7557 // CHECK13-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
7558 // CHECK13-NEXT: store i8* null, i8** [[TMP47]], align 8
7559 // CHECK13-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
7560 // CHECK13-NEXT: [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32**
7561 // CHECK13-NEXT: store i32* [[VLA]], i32** [[TMP49]], align 8
7562 // CHECK13-NEXT: [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
7563 // CHECK13-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i32**
7564 // CHECK13-NEXT: store i32* [[VLA]], i32** [[TMP51]], align 8
7565 // CHECK13-NEXT: [[TMP52:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
7566 // CHECK13-NEXT: store i64 [[TMP35]], i64* [[TMP52]], align 8
7567 // CHECK13-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
7568 // CHECK13-NEXT: store i8* null, i8** [[TMP53]], align 8
7569 // CHECK13-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
7570 // CHECK13-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
7571 // CHECK13-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
7572 // CHECK13-NEXT: [[TMP57:%.*]] = load i32, i32* [[N]], align 4
7573 // CHECK13-NEXT: store i32 [[TMP57]], i32* [[DOTCAPTURE_EXPR_10]], align 4
7574 // CHECK13-NEXT: [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
7575 // CHECK13-NEXT: [[SUB12:%.*]] = sub nsw i32 [[TMP58]], 0
7576 // CHECK13-NEXT: [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
7577 // CHECK13-NEXT: [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
7578 // CHECK13-NEXT: store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
7579 // CHECK13-NEXT: [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
7580 // CHECK13-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP59]], 1
7581 // CHECK13-NEXT: [[TMP60:%.*]] = zext i32 [[ADD15]] to i64
7582 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP60]])
7583 // CHECK13-NEXT: [[TMP61:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153.region_id, i32 3, i8** [[TMP54]], i8** [[TMP55]], i64* [[TMP56]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
7584 // CHECK13-NEXT: [[TMP62:%.*]] = icmp ne i32 [[TMP61]], 0
7585 // CHECK13-NEXT: br i1 [[TMP62]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
7586 // CHECK13: omp_offload.failed16:
7587 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153(i64 [[TMP34]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
7588 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT17]]
7589 // CHECK13: omp_offload.cont17:
7590 // CHECK13-NEXT: [[TMP63:%.*]] = load i32, i32* [[M]], align 4
7591 // CHECK13-NEXT: [[CONV18:%.*]] = bitcast i64* [[M_CASTED]] to i32*
7592 // CHECK13-NEXT: store i32 [[TMP63]], i32* [[CONV18]], align 4
7593 // CHECK13-NEXT: [[TMP64:%.*]] = load i64, i64* [[M_CASTED]], align 8
7594 // CHECK13-NEXT: [[TMP65:%.*]] = load i32, i32* [[N]], align 4
7595 // CHECK13-NEXT: [[CONV20:%.*]] = bitcast i64* [[N_CASTED19]] to i32*
7596 // CHECK13-NEXT: store i32 [[TMP65]], i32* [[CONV20]], align 4
7597 // CHECK13-NEXT: [[TMP66:%.*]] = load i64, i64* [[N_CASTED19]], align 8
7598 // CHECK13-NEXT: [[TMP67:%.*]] = mul nuw i64 [[TMP1]], 4
7599 // CHECK13-NEXT: [[TMP68:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
7600 // CHECK13-NEXT: [[TMP69:%.*]] = bitcast i8** [[TMP68]] to i64*
7601 // CHECK13-NEXT: store i64 [[TMP64]], i64* [[TMP69]], align 8
7602 // CHECK13-NEXT: [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
7603 // CHECK13-NEXT: [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i64*
7604 // CHECK13-NEXT: store i64 [[TMP64]], i64* [[TMP71]], align 8
7605 // CHECK13-NEXT: [[TMP72:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
7606 // CHECK13-NEXT: store i64 4, i64* [[TMP72]], align 8
7607 // CHECK13-NEXT: [[TMP73:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
7608 // CHECK13-NEXT: store i8* null, i8** [[TMP73]], align 8
7609 // CHECK13-NEXT: [[TMP74:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
7610 // CHECK13-NEXT: [[TMP75:%.*]] = bitcast i8** [[TMP74]] to i64*
7611 // CHECK13-NEXT: store i64 [[TMP66]], i64* [[TMP75]], align 8
7612 // CHECK13-NEXT: [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
7613 // CHECK13-NEXT: [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i64*
7614 // CHECK13-NEXT: store i64 [[TMP66]], i64* [[TMP77]], align 8
7615 // CHECK13-NEXT: [[TMP78:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 1
7616 // CHECK13-NEXT: store i64 4, i64* [[TMP78]], align 8
7617 // CHECK13-NEXT: [[TMP79:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
7618 // CHECK13-NEXT: store i8* null, i8** [[TMP79]], align 8
7619 // CHECK13-NEXT: [[TMP80:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 2
7620 // CHECK13-NEXT: [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
7621 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[TMP81]], align 8
7622 // CHECK13-NEXT: [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 2
7623 // CHECK13-NEXT: [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
7624 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[TMP83]], align 8
7625 // CHECK13-NEXT: [[TMP84:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 2
7626 // CHECK13-NEXT: store i64 8, i64* [[TMP84]], align 8
7627 // CHECK13-NEXT: [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 2
7628 // CHECK13-NEXT: store i8* null, i8** [[TMP85]], align 8
7629 // CHECK13-NEXT: [[TMP86:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 3
7630 // CHECK13-NEXT: [[TMP87:%.*]] = bitcast i8** [[TMP86]] to i32**
7631 // CHECK13-NEXT: store i32* [[VLA]], i32** [[TMP87]], align 8
7632 // CHECK13-NEXT: [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 3
7633 // CHECK13-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
7634 // CHECK13-NEXT: store i32* [[VLA]], i32** [[TMP89]], align 8
7635 // CHECK13-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 3
7636 // CHECK13-NEXT: store i64 [[TMP67]], i64* [[TMP90]], align 8
7637 // CHECK13-NEXT: [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 3
7638 // CHECK13-NEXT: store i8* null, i8** [[TMP91]], align 8
7639 // CHECK13-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
7640 // CHECK13-NEXT: [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
7641 // CHECK13-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
7642 // CHECK13-NEXT: [[TMP95:%.*]] = load i32, i32* [[N]], align 4
7643 // CHECK13-NEXT: store i32 [[TMP95]], i32* [[DOTCAPTURE_EXPR_26]], align 4
7644 // CHECK13-NEXT: [[TMP96:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_26]], align 4
7645 // CHECK13-NEXT: [[SUB28:%.*]] = sub nsw i32 [[TMP96]], 0
7646 // CHECK13-NEXT: [[DIV29:%.*]] = sdiv i32 [[SUB28]], 1
7647 // CHECK13-NEXT: [[SUB30:%.*]] = sub nsw i32 [[DIV29]], 1
7648 // CHECK13-NEXT: store i32 [[SUB30]], i32* [[DOTCAPTURE_EXPR_27]], align 4
7649 // CHECK13-NEXT: [[TMP97:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
7650 // CHECK13-NEXT: [[ADD31:%.*]] = add nsw i32 [[TMP97]], 1
7651 // CHECK13-NEXT: [[TMP98:%.*]] = zext i32 [[ADD31]] to i64
7652 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP98]])
7653 // CHECK13-NEXT: [[TMP99:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158.region_id, i32 4, i8** [[TMP92]], i8** [[TMP93]], i64* [[TMP94]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
7654 // CHECK13-NEXT: [[TMP100:%.*]] = icmp ne i32 [[TMP99]], 0
7655 // CHECK13-NEXT: br i1 [[TMP100]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
7656 // CHECK13: omp_offload.failed32:
7657 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158(i64 [[TMP64]], i64 [[TMP66]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
7658 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT33]]
7659 // CHECK13: omp_offload.cont33:
7660 // CHECK13-NEXT: [[TMP101:%.*]] = load i32, i32* [[N]], align 4
7661 // CHECK13-NEXT: [[CONV35:%.*]] = bitcast i64* [[N_CASTED34]] to i32*
7662 // CHECK13-NEXT: store i32 [[TMP101]], i32* [[CONV35]], align 4
7663 // CHECK13-NEXT: [[TMP102:%.*]] = load i64, i64* [[N_CASTED34]], align 8
7664 // CHECK13-NEXT: [[TMP103:%.*]] = mul nuw i64 [[TMP1]], 4
7665 // CHECK13-NEXT: [[TMP104:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
7666 // CHECK13-NEXT: [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i64*
7667 // CHECK13-NEXT: store i64 [[TMP102]], i64* [[TMP105]], align 8
7668 // CHECK13-NEXT: [[TMP106:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
7669 // CHECK13-NEXT: [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i64*
7670 // CHECK13-NEXT: store i64 [[TMP102]], i64* [[TMP107]], align 8
7671 // CHECK13-NEXT: [[TMP108:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
7672 // CHECK13-NEXT: store i64 4, i64* [[TMP108]], align 8
7673 // CHECK13-NEXT: [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 0
7674 // CHECK13-NEXT: store i8* null, i8** [[TMP109]], align 8
7675 // CHECK13-NEXT: [[TMP110:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 1
7676 // CHECK13-NEXT: [[TMP111:%.*]] = bitcast i8** [[TMP110]] to i64*
7677 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[TMP111]], align 8
7678 // CHECK13-NEXT: [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 1
7679 // CHECK13-NEXT: [[TMP113:%.*]] = bitcast i8** [[TMP112]] to i64*
7680 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[TMP113]], align 8
7681 // CHECK13-NEXT: [[TMP114:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 1
7682 // CHECK13-NEXT: store i64 8, i64* [[TMP114]], align 8
7683 // CHECK13-NEXT: [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 1
7684 // CHECK13-NEXT: store i8* null, i8** [[TMP115]], align 8
7685 // CHECK13-NEXT: [[TMP116:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 2
7686 // CHECK13-NEXT: [[TMP117:%.*]] = bitcast i8** [[TMP116]] to i32**
7687 // CHECK13-NEXT: store i32* [[VLA]], i32** [[TMP117]], align 8
7688 // CHECK13-NEXT: [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 2
7689 // CHECK13-NEXT: [[TMP119:%.*]] = bitcast i8** [[TMP118]] to i32**
7690 // CHECK13-NEXT: store i32* [[VLA]], i32** [[TMP119]], align 8
7691 // CHECK13-NEXT: [[TMP120:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 2
7692 // CHECK13-NEXT: store i64 [[TMP103]], i64* [[TMP120]], align 8
7693 // CHECK13-NEXT: [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 2
7694 // CHECK13-NEXT: store i8* null, i8** [[TMP121]], align 8
7695 // CHECK13-NEXT: [[TMP122:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
7696 // CHECK13-NEXT: [[TMP123:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
7697 // CHECK13-NEXT: [[TMP124:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
7698 // CHECK13-NEXT: [[TMP125:%.*]] = load i32, i32* [[N]], align 4
7699 // CHECK13-NEXT: store i32 [[TMP125]], i32* [[DOTCAPTURE_EXPR_41]], align 4
7700 // CHECK13-NEXT: [[TMP126:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
7701 // CHECK13-NEXT: [[SUB43:%.*]] = sub nsw i32 [[TMP126]], 0
7702 // CHECK13-NEXT: [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
7703 // CHECK13-NEXT: [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
7704 // CHECK13-NEXT: store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
7705 // CHECK13-NEXT: [[TMP127:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
7706 // CHECK13-NEXT: [[ADD46:%.*]] = add nsw i32 [[TMP127]], 1
7707 // CHECK13-NEXT: [[TMP128:%.*]] = zext i32 [[ADD46]] to i64
7708 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP128]])
7709 // CHECK13-NEXT: [[TMP129:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163.region_id, i32 3, i8** [[TMP122]], i8** [[TMP123]], i64* [[TMP124]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
7710 // CHECK13-NEXT: [[TMP130:%.*]] = icmp ne i32 [[TMP129]], 0
7711 // CHECK13-NEXT: br i1 [[TMP130]], label [[OMP_OFFLOAD_FAILED47:%.*]], label [[OMP_OFFLOAD_CONT48:%.*]]
7712 // CHECK13: omp_offload.failed47:
7713 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163(i64 [[TMP102]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
7714 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT48]]
7715 // CHECK13: omp_offload.cont48:
7716 // CHECK13-NEXT: [[TMP131:%.*]] = load i32, i32* [[M]], align 4
7717 // CHECK13-NEXT: [[CONV50:%.*]] = bitcast i64* [[M_CASTED49]] to i32*
7718 // CHECK13-NEXT: store i32 [[TMP131]], i32* [[CONV50]], align 4
7719 // CHECK13-NEXT: [[TMP132:%.*]] = load i64, i64* [[M_CASTED49]], align 8
7720 // CHECK13-NEXT: [[TMP133:%.*]] = load i32, i32* [[N]], align 4
7721 // CHECK13-NEXT: [[CONV52:%.*]] = bitcast i64* [[N_CASTED51]] to i32*
7722 // CHECK13-NEXT: store i32 [[TMP133]], i32* [[CONV52]], align 4
7723 // CHECK13-NEXT: [[TMP134:%.*]] = load i64, i64* [[N_CASTED51]], align 8
7724 // CHECK13-NEXT: [[TMP135:%.*]] = mul nuw i64 [[TMP1]], 4
7725 // CHECK13-NEXT: [[TMP136:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
7726 // CHECK13-NEXT: [[TMP137:%.*]] = bitcast i8** [[TMP136]] to i64*
7727 // CHECK13-NEXT: store i64 [[TMP132]], i64* [[TMP137]], align 8
7728 // CHECK13-NEXT: [[TMP138:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
7729 // CHECK13-NEXT: [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i64*
7730 // CHECK13-NEXT: store i64 [[TMP132]], i64* [[TMP139]], align 8
7731 // CHECK13-NEXT: [[TMP140:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
7732 // CHECK13-NEXT: store i64 4, i64* [[TMP140]], align 8
7733 // CHECK13-NEXT: [[TMP141:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 0
7734 // CHECK13-NEXT: store i8* null, i8** [[TMP141]], align 8
7735 // CHECK13-NEXT: [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 1
7736 // CHECK13-NEXT: [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i64*
7737 // CHECK13-NEXT: store i64 [[TMP134]], i64* [[TMP143]], align 8
7738 // CHECK13-NEXT: [[TMP144:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 1
7739 // CHECK13-NEXT: [[TMP145:%.*]] = bitcast i8** [[TMP144]] to i64*
7740 // CHECK13-NEXT: store i64 [[TMP134]], i64* [[TMP145]], align 8
7741 // CHECK13-NEXT: [[TMP146:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 1
7742 // CHECK13-NEXT: store i64 4, i64* [[TMP146]], align 8
7743 // CHECK13-NEXT: [[TMP147:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 1
7744 // CHECK13-NEXT: store i8* null, i8** [[TMP147]], align 8
7745 // CHECK13-NEXT: [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 2
7746 // CHECK13-NEXT: [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
7747 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[TMP149]], align 8
7748 // CHECK13-NEXT: [[TMP150:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 2
7749 // CHECK13-NEXT: [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
7750 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[TMP151]], align 8
7751 // CHECK13-NEXT: [[TMP152:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 2
7752 // CHECK13-NEXT: store i64 8, i64* [[TMP152]], align 8
7753 // CHECK13-NEXT: [[TMP153:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 2
7754 // CHECK13-NEXT: store i8* null, i8** [[TMP153]], align 8
7755 // CHECK13-NEXT: [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 3
7756 // CHECK13-NEXT: [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32**
7757 // CHECK13-NEXT: store i32* [[VLA]], i32** [[TMP155]], align 8
7758 // CHECK13-NEXT: [[TMP156:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 3
7759 // CHECK13-NEXT: [[TMP157:%.*]] = bitcast i8** [[TMP156]] to i32**
7760 // CHECK13-NEXT: store i32* [[VLA]], i32** [[TMP157]], align 8
7761 // CHECK13-NEXT: [[TMP158:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 3
7762 // CHECK13-NEXT: store i64 [[TMP135]], i64* [[TMP158]], align 8
7763 // CHECK13-NEXT: [[TMP159:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 3
7764 // CHECK13-NEXT: store i8* null, i8** [[TMP159]], align 8
7765 // CHECK13-NEXT: [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
7766 // CHECK13-NEXT: [[TMP161:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
7767 // CHECK13-NEXT: [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
7768 // CHECK13-NEXT: [[TMP163:%.*]] = load i32, i32* [[N]], align 4
7769 // CHECK13-NEXT: store i32 [[TMP163]], i32* [[DOTCAPTURE_EXPR_58]], align 4
7770 // CHECK13-NEXT: [[TMP164:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_58]], align 4
7771 // CHECK13-NEXT: [[SUB60:%.*]] = sub nsw i32 [[TMP164]], 0
7772 // CHECK13-NEXT: [[DIV61:%.*]] = sdiv i32 [[SUB60]], 1
7773 // CHECK13-NEXT: [[SUB62:%.*]] = sub nsw i32 [[DIV61]], 1
7774 // CHECK13-NEXT: store i32 [[SUB62]], i32* [[DOTCAPTURE_EXPR_59]], align 4
7775 // CHECK13-NEXT: [[TMP165:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_59]], align 4
7776 // CHECK13-NEXT: [[ADD63:%.*]] = add nsw i32 [[TMP165]], 1
7777 // CHECK13-NEXT: [[TMP166:%.*]] = zext i32 [[ADD63]] to i64
7778 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP166]])
7779 // CHECK13-NEXT: [[TMP167:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168.region_id, i32 4, i8** [[TMP160]], i8** [[TMP161]], i64* [[TMP162]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
7780 // CHECK13-NEXT: [[TMP168:%.*]] = icmp ne i32 [[TMP167]], 0
7781 // CHECK13-NEXT: br i1 [[TMP168]], label [[OMP_OFFLOAD_FAILED64:%.*]], label [[OMP_OFFLOAD_CONT65:%.*]]
7782 // CHECK13: omp_offload.failed64:
7783 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168(i64 [[TMP132]], i64 [[TMP134]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
7784 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT65]]
7785 // CHECK13: omp_offload.cont65:
7786 // CHECK13-NEXT: [[TMP169:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
7787 // CHECK13-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP169]])
7788 // CHECK13-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
7789 // CHECK13-NEXT: [[TMP170:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
7790 // CHECK13-NEXT: call void @llvm.stackrestore(i8* [[TMP170]])
7791 // CHECK13-NEXT: [[TMP171:%.*]] = load i32, i32* [[RETVAL]], align 4
7792 // CHECK13-NEXT: ret i32 [[TMP171]]
7793 //
7794 //
7795 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148
7796 // CHECK13-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
7797 // CHECK13-NEXT: entry:
7798 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
7799 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
7800 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
7801 // CHECK13-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
7802 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
7803 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
7804 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
7805 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
7806 // CHECK13-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
7807 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
7808 // CHECK13-NEXT: ret void
7809 //
7810 //
7811 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined.
7812 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
7813 // CHECK13-NEXT: entry:
7814 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7815 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7816 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
7817 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
7818 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
7819 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7820 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
7821 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
7822 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
7823 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
7824 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7825 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7826 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7827 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7828 // CHECK13-NEXT: [[I3:%.*]] = alloca i32, align 4
7829 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7830 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7831 // CHECK13-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
7832 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
7833 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
7834 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
7835 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
7836 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
7837 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
7838 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
7839 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7840 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
7841 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
7842 // CHECK13-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
7843 // CHECK13-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
7844 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
7845 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7846 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
7847 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
7848 // CHECK13: omp.precond.then:
7849 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7850 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7851 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
7852 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7853 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7854 // CHECK13-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7855 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
7856 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7857 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7858 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7859 // CHECK13-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
7860 // CHECK13-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7861 // CHECK13: cond.true:
7862 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7863 // CHECK13-NEXT: br label [[COND_END:%.*]]
7864 // CHECK13: cond.false:
7865 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7866 // CHECK13-NEXT: br label [[COND_END]]
7867 // CHECK13: cond.end:
7868 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
7869 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7870 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7871 // CHECK13-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
7872 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7873 // CHECK13: omp.inner.for.cond:
7874 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7875 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7876 // CHECK13-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
7877 // CHECK13-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7878 // CHECK13: omp.inner.for.body:
7879 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7880 // CHECK13-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
7881 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7882 // CHECK13-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
7883 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
7884 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7885 // CHECK13: omp.inner.for.inc:
7886 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7887 // CHECK13-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7888 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
7889 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7890 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
7891 // CHECK13: omp.inner.for.end:
7892 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7893 // CHECK13: omp.loop.exit:
7894 // CHECK13-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7895 // CHECK13-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
7896 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
7897 // CHECK13-NEXT: br label [[OMP_PRECOND_END]]
7898 // CHECK13: omp.precond.end:
7899 // CHECK13-NEXT: ret void
7900 //
7901 //
7902 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..1
7903 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
7904 // CHECK13-NEXT: entry:
7905 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7906 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7907 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
7908 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
7909 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
7910 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
7911 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
7912 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7913 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
7914 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
7915 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
7916 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
7917 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7918 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7919 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7920 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7921 // CHECK13-NEXT: [[I4:%.*]] = alloca i32, align 4
7922 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7923 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7924 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7925 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7926 // CHECK13-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
7927 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
7928 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
7929 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
7930 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
7931 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
7932 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
7933 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
7934 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7935 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
7936 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
7937 // CHECK13-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
7938 // CHECK13-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
7939 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
7940 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7941 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
7942 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
7943 // CHECK13: omp.precond.then:
7944 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7945 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7946 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
7947 // CHECK13-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7948 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
7949 // CHECK13-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7950 // CHECK13-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
7951 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
7952 // CHECK13-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
7953 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7954 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7955 // CHECK13-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7956 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
7957 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7958 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7959 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7960 // CHECK13-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
7961 // CHECK13-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7962 // CHECK13: cond.true:
7963 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7964 // CHECK13-NEXT: br label [[COND_END:%.*]]
7965 // CHECK13: cond.false:
7966 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7967 // CHECK13-NEXT: br label [[COND_END]]
7968 // CHECK13: cond.end:
7969 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
7970 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7971 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7972 // CHECK13-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
7973 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7974 // CHECK13: omp.inner.for.cond:
7975 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7976 // CHECK13-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7977 // CHECK13-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
7978 // CHECK13-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7979 // CHECK13: omp.inner.for.body:
7980 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7981 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
7982 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7983 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
7984 // CHECK13-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
7985 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
7986 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
7987 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
7988 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7989 // CHECK13: omp.body.continue:
7990 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7991 // CHECK13: omp.inner.for.inc:
7992 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7993 // CHECK13-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
7994 // CHECK13-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
7995 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
7996 // CHECK13: omp.inner.for.end:
7997 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7998 // CHECK13: omp.loop.exit:
7999 // CHECK13-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8000 // CHECK13-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
8001 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
8002 // CHECK13-NEXT: br label [[OMP_PRECOND_END]]
8003 // CHECK13: omp.precond.end:
8004 // CHECK13-NEXT: ret void
8005 //
8006 //
8007 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153
8008 // CHECK13-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
8009 // CHECK13-NEXT: entry:
8010 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
8011 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
8012 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
8013 // CHECK13-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
8014 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8015 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
8016 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
8017 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8018 // CHECK13-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
8019 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
8020 // CHECK13-NEXT: ret void
8021 //
8022 //
8023 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..2
8024 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
8025 // CHECK13-NEXT: entry:
8026 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8027 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8028 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
8029 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
8030 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
8031 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8032 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8033 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8034 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8035 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8036 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8037 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8038 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8039 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8040 // CHECK13-NEXT: [[I3:%.*]] = alloca i32, align 4
8041 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8042 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8043 // CHECK13-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
8044 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8045 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
8046 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
8047 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8048 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
8049 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8050 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
8051 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8052 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8053 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8054 // CHECK13-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
8055 // CHECK13-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8056 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
8057 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8058 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8059 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8060 // CHECK13: omp.precond.then:
8061 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8062 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8063 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
8064 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8065 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8066 // CHECK13-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8067 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
8068 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8069 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8070 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8071 // CHECK13-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
8072 // CHECK13-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8073 // CHECK13: cond.true:
8074 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8075 // CHECK13-NEXT: br label [[COND_END:%.*]]
8076 // CHECK13: cond.false:
8077 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8078 // CHECK13-NEXT: br label [[COND_END]]
8079 // CHECK13: cond.end:
8080 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
8081 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8082 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8083 // CHECK13-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
8084 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8085 // CHECK13: omp.inner.for.cond:
8086 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8087 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8088 // CHECK13-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
8089 // CHECK13-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8090 // CHECK13: omp.inner.for.body:
8091 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8092 // CHECK13-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
8093 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8094 // CHECK13-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
8095 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
8096 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8097 // CHECK13: omp.inner.for.inc:
8098 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8099 // CHECK13-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8100 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
8101 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
8102 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8103 // CHECK13: omp.inner.for.end:
8104 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8105 // CHECK13: omp.loop.exit:
8106 // CHECK13-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8107 // CHECK13-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
8108 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
8109 // CHECK13-NEXT: br label [[OMP_PRECOND_END]]
8110 // CHECK13: omp.precond.end:
8111 // CHECK13-NEXT: ret void
8112 //
8113 //
8114 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..3
8115 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
8116 // CHECK13-NEXT: entry:
8117 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8118 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8119 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8120 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8121 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
8122 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
8123 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
8124 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8125 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8126 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8127 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8128 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8129 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8130 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8131 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8132 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8133 // CHECK13-NEXT: [[I4:%.*]] = alloca i32, align 4
8134 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8135 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8136 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8137 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8138 // CHECK13-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
8139 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8140 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
8141 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
8142 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8143 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
8144 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8145 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
8146 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8147 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8148 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8149 // CHECK13-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
8150 // CHECK13-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8151 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
8152 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8153 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8154 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8155 // CHECK13: omp.precond.then:
8156 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8157 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8158 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
8159 // CHECK13-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8160 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
8161 // CHECK13-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8162 // CHECK13-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
8163 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8164 // CHECK13-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
8165 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8166 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8167 // CHECK13-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8168 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
8169 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8170 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8171 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8172 // CHECK13-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
8173 // CHECK13-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8174 // CHECK13: cond.true:
8175 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8176 // CHECK13-NEXT: br label [[COND_END:%.*]]
8177 // CHECK13: cond.false:
8178 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8179 // CHECK13-NEXT: br label [[COND_END]]
8180 // CHECK13: cond.end:
8181 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
8182 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8183 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8184 // CHECK13-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
8185 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8186 // CHECK13: omp.inner.for.cond:
8187 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8188 // CHECK13-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8189 // CHECK13-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
8190 // CHECK13-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8191 // CHECK13: omp.inner.for.body:
8192 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8193 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
8194 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8195 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
8196 // CHECK13-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
8197 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
8198 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
8199 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
8200 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8201 // CHECK13: omp.body.continue:
8202 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8203 // CHECK13: omp.inner.for.inc:
8204 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8205 // CHECK13-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
8206 // CHECK13-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
8207 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8208 // CHECK13: omp.inner.for.end:
8209 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8210 // CHECK13: omp.loop.exit:
8211 // CHECK13-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8212 // CHECK13-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
8213 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
8214 // CHECK13-NEXT: br label [[OMP_PRECOND_END]]
8215 // CHECK13: omp.precond.end:
8216 // CHECK13-NEXT: ret void
8217 //
8218 //
8219 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158
8220 // CHECK13-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
8221 // CHECK13-NEXT: entry:
8222 // CHECK13-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
8223 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
8224 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
8225 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
8226 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8227 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
8228 // CHECK13-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
8229 // CHECK13-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
8230 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8231 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
8232 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
8233 // CHECK13-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
8234 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8235 // CHECK13-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
8236 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
8237 // CHECK13-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
8238 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8239 // CHECK13-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
8240 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
8241 // CHECK13-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
8242 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
8243 // CHECK13-NEXT: ret void
8244 //
8245 //
8246 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..5
8247 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
8248 // CHECK13-NEXT: entry:
8249 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8250 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8251 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
8252 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
8253 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
8254 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
8255 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8256 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8257 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8258 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
8259 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8260 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8261 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8262 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8263 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8264 // CHECK13-NEXT: [[I4:%.*]] = alloca i32, align 4
8265 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
8266 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8267 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8268 // CHECK13-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
8269 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8270 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
8271 // CHECK13-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
8272 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
8273 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8274 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
8275 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
8276 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8277 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8278 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8279 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8280 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8281 // CHECK13-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
8282 // CHECK13-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
8283 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
8284 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8285 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8286 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8287 // CHECK13: omp.precond.then:
8288 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8289 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8290 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
8291 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8292 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8293 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[CONV]], align 8
8294 // CHECK13-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8295 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
8296 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
8297 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8298 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8299 // CHECK13-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
8300 // CHECK13-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8301 // CHECK13: cond.true:
8302 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8303 // CHECK13-NEXT: br label [[COND_END:%.*]]
8304 // CHECK13: cond.false:
8305 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8306 // CHECK13-NEXT: br label [[COND_END]]
8307 // CHECK13: cond.end:
8308 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
8309 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8310 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8311 // CHECK13-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
8312 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8313 // CHECK13: omp.inner.for.cond:
8314 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8315 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8316 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
8317 // CHECK13-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
8318 // CHECK13-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8319 // CHECK13: omp.inner.for.body:
8320 // CHECK13-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8321 // CHECK13-NEXT: [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
8322 // CHECK13-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8323 // CHECK13-NEXT: [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
8324 // CHECK13-NEXT: [[TMP21:%.*]] = load i32, i32* [[CONV]], align 8
8325 // CHECK13-NEXT: [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
8326 // CHECK13-NEXT: store i32 [[TMP21]], i32* [[CONV7]], align 4
8327 // CHECK13-NEXT: [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
8328 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]])
8329 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8330 // CHECK13: omp.inner.for.inc:
8331 // CHECK13-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8332 // CHECK13-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8333 // CHECK13-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
8334 // CHECK13-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
8335 // CHECK13-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8336 // CHECK13-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8337 // CHECK13-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
8338 // CHECK13-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4
8339 // CHECK13-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8340 // CHECK13-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8341 // CHECK13-NEXT: [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
8342 // CHECK13-NEXT: store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4
8343 // CHECK13-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8344 // CHECK13-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8345 // CHECK13-NEXT: [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
8346 // CHECK13-NEXT: br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
8347 // CHECK13: cond.true12:
8348 // CHECK13-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8349 // CHECK13-NEXT: br label [[COND_END14:%.*]]
8350 // CHECK13: cond.false13:
8351 // CHECK13-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8352 // CHECK13-NEXT: br label [[COND_END14]]
8353 // CHECK13: cond.end14:
8354 // CHECK13-NEXT: [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
8355 // CHECK13-NEXT: store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4
8356 // CHECK13-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8357 // CHECK13-NEXT: store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4
8358 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8359 // CHECK13: omp.inner.for.end:
8360 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8361 // CHECK13: omp.loop.exit:
8362 // CHECK13-NEXT: [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8363 // CHECK13-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
8364 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
8365 // CHECK13-NEXT: br label [[OMP_PRECOND_END]]
8366 // CHECK13: omp.precond.end:
8367 // CHECK13-NEXT: ret void
8368 //
8369 //
8370 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..6
8371 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
8372 // CHECK13-NEXT: entry:
8373 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8374 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8375 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8376 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8377 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
8378 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
8379 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
8380 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
8381 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8382 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8383 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8384 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
8385 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8386 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8387 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8388 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8389 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8390 // CHECK13-NEXT: [[I6:%.*]] = alloca i32, align 4
8391 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8392 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8393 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8394 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8395 // CHECK13-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
8396 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8397 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
8398 // CHECK13-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
8399 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
8400 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8401 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
8402 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
8403 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8404 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8405 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8406 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8407 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8408 // CHECK13-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
8409 // CHECK13-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
8410 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
8411 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8412 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8413 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8414 // CHECK13: omp.precond.then:
8415 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8416 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8417 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
8418 // CHECK13-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8419 // CHECK13-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
8420 // CHECK13-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8421 // CHECK13-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
8422 // CHECK13-NEXT: store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
8423 // CHECK13-NEXT: store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
8424 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8425 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8426 // CHECK13-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8427 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
8428 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8429 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8430 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8431 // CHECK13-NEXT: [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
8432 // CHECK13-NEXT: br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8433 // CHECK13: cond.true:
8434 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8435 // CHECK13-NEXT: br label [[COND_END:%.*]]
8436 // CHECK13: cond.false:
8437 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8438 // CHECK13-NEXT: br label [[COND_END]]
8439 // CHECK13: cond.end:
8440 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
8441 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8442 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8443 // CHECK13-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
8444 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8445 // CHECK13: omp.inner.for.cond:
8446 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8447 // CHECK13-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8448 // CHECK13-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
8449 // CHECK13-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8450 // CHECK13: omp.inner.for.body:
8451 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8452 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
8453 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8454 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I6]], align 4
8455 // CHECK13-NEXT: [[TMP19:%.*]] = load i32, i32* [[I6]], align 4
8456 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
8457 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
8458 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
8459 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8460 // CHECK13: omp.body.continue:
8461 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8462 // CHECK13: omp.inner.for.inc:
8463 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8464 // CHECK13-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
8465 // CHECK13-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
8466 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8467 // CHECK13: omp.inner.for.end:
8468 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8469 // CHECK13: omp.loop.exit:
8470 // CHECK13-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8471 // CHECK13-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
8472 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
8473 // CHECK13-NEXT: br label [[OMP_PRECOND_END]]
8474 // CHECK13: omp.precond.end:
8475 // CHECK13-NEXT: ret void
8476 //
8477 //
8478 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163
8479 // CHECK13-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
8480 // CHECK13-NEXT: entry:
8481 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
8482 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
8483 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
8484 // CHECK13-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
8485 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8486 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
8487 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
8488 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8489 // CHECK13-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
8490 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
8491 // CHECK13-NEXT: ret void
8492 //
8493 //
8494 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..8
8495 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
8496 // CHECK13-NEXT: entry:
8497 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8498 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8499 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
8500 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
8501 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
8502 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8503 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8504 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8505 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8506 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8507 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8508 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8509 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8510 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8511 // CHECK13-NEXT: [[I3:%.*]] = alloca i32, align 4
8512 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8513 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8514 // CHECK13-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
8515 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8516 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
8517 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
8518 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8519 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
8520 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8521 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
8522 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8523 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8524 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8525 // CHECK13-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
8526 // CHECK13-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8527 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
8528 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8529 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8530 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8531 // CHECK13: omp.precond.then:
8532 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8533 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8534 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
8535 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8536 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8537 // CHECK13-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8538 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
8539 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8540 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8541 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8542 // CHECK13-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
8543 // CHECK13-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8544 // CHECK13: cond.true:
8545 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8546 // CHECK13-NEXT: br label [[COND_END:%.*]]
8547 // CHECK13: cond.false:
8548 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8549 // CHECK13-NEXT: br label [[COND_END]]
8550 // CHECK13: cond.end:
8551 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
8552 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8553 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8554 // CHECK13-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
8555 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8556 // CHECK13: omp.inner.for.cond:
8557 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8558 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8559 // CHECK13-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
8560 // CHECK13-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8561 // CHECK13: omp.inner.for.body:
8562 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8563 // CHECK13-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
8564 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8565 // CHECK13-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
8566 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
8567 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8568 // CHECK13: omp.inner.for.inc:
8569 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8570 // CHECK13-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8571 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
8572 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
8573 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8574 // CHECK13: omp.inner.for.end:
8575 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8576 // CHECK13: omp.loop.exit:
8577 // CHECK13-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8578 // CHECK13-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
8579 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
8580 // CHECK13-NEXT: br label [[OMP_PRECOND_END]]
8581 // CHECK13: omp.precond.end:
8582 // CHECK13-NEXT: ret void
8583 //
8584 //
8585 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..9
8586 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
8587 // CHECK13-NEXT: entry:
8588 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8589 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8590 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8591 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8592 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
8593 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
8594 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
8595 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8596 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8597 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8598 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8599 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8600 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8601 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8602 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8603 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8604 // CHECK13-NEXT: [[I4:%.*]] = alloca i32, align 4
8605 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8606 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8607 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8608 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8609 // CHECK13-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
8610 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8611 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
8612 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
8613 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8614 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
8615 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8616 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
8617 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8618 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8619 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8620 // CHECK13-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
8621 // CHECK13-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8622 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
8623 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8624 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8625 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8626 // CHECK13: omp.precond.then:
8627 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8628 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8629 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
8630 // CHECK13-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8631 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
8632 // CHECK13-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8633 // CHECK13-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
8634 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8635 // CHECK13-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
8636 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8637 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8638 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8639 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8640 // CHECK13-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8641 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
8642 // CHECK13-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 1073741859, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
8643 // CHECK13-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
8644 // CHECK13: omp.dispatch.cond:
8645 // CHECK13-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8646 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
8647 // CHECK13-NEXT: [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
8648 // CHECK13-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
8649 // CHECK13-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
8650 // CHECK13: omp.dispatch.body:
8651 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8652 // CHECK13-NEXT: store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
8653 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8654 // CHECK13: omp.inner.for.cond:
8655 // CHECK13-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
8656 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !15
8657 // CHECK13-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
8658 // CHECK13-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8659 // CHECK13: omp.inner.for.body:
8660 // CHECK13-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
8661 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
8662 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8663 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !15
8664 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !15
8665 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
8666 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
8667 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !15
8668 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8669 // CHECK13: omp.body.continue:
8670 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8671 // CHECK13: omp.inner.for.inc:
8672 // CHECK13-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
8673 // CHECK13-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP21]], 1
8674 // CHECK13-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
8675 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
8676 // CHECK13: omp.inner.for.end:
8677 // CHECK13-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
8678 // CHECK13: omp.dispatch.inc:
8679 // CHECK13-NEXT: br label [[OMP_DISPATCH_COND]]
8680 // CHECK13: omp.dispatch.end:
8681 // CHECK13-NEXT: br label [[OMP_PRECOND_END]]
8682 // CHECK13: omp.precond.end:
8683 // CHECK13-NEXT: ret void
8684 //
8685 //
8686 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168
8687 // CHECK13-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
8688 // CHECK13-NEXT: entry:
8689 // CHECK13-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
8690 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
8691 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
8692 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
8693 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8694 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
8695 // CHECK13-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
8696 // CHECK13-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
8697 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8698 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
8699 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
8700 // CHECK13-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
8701 // CHECK13-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8702 // CHECK13-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
8703 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
8704 // CHECK13-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
8705 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8706 // CHECK13-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
8707 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
8708 // CHECK13-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
8709 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
8710 // CHECK13-NEXT: ret void
8711 //
8712 //
8713 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..11
8714 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
8715 // CHECK13-NEXT: entry:
8716 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8717 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8718 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
8719 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
8720 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
8721 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
8722 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8723 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8724 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8725 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
8726 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8727 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8728 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8729 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8730 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8731 // CHECK13-NEXT: [[I4:%.*]] = alloca i32, align 4
8732 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
8733 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8734 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8735 // CHECK13-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
8736 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8737 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
8738 // CHECK13-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
8739 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
8740 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8741 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
8742 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
8743 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8744 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8745 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8746 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8747 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8748 // CHECK13-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
8749 // CHECK13-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
8750 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
8751 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8752 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8753 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8754 // CHECK13: omp.precond.then:
8755 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8756 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8757 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
8758 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8759 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8760 // CHECK13-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8761 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
8762 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
8763 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8764 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8765 // CHECK13-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
8766 // CHECK13-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8767 // CHECK13: cond.true:
8768 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8769 // CHECK13-NEXT: br label [[COND_END:%.*]]
8770 // CHECK13: cond.false:
8771 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8772 // CHECK13-NEXT: br label [[COND_END]]
8773 // CHECK13: cond.end:
8774 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
8775 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8776 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8777 // CHECK13-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
8778 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8779 // CHECK13: omp.inner.for.cond:
8780 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8781 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8782 // CHECK13-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
8783 // CHECK13-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8784 // CHECK13: omp.inner.for.body:
8785 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8786 // CHECK13-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
8787 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8788 // CHECK13-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
8789 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[CONV]], align 8
8790 // CHECK13-NEXT: [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
8791 // CHECK13-NEXT: store i32 [[TMP20]], i32* [[CONV7]], align 4
8792 // CHECK13-NEXT: [[TMP21:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
8793 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP21]])
8794 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8795 // CHECK13: omp.inner.for.inc:
8796 // CHECK13-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8797 // CHECK13-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8798 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
8799 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
8800 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
8801 // CHECK13: omp.inner.for.end:
8802 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8803 // CHECK13: omp.loop.exit:
8804 // CHECK13-NEXT: [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8805 // CHECK13-NEXT: [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
8806 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
8807 // CHECK13-NEXT: br label [[OMP_PRECOND_END]]
8808 // CHECK13: omp.precond.end:
8809 // CHECK13-NEXT: ret void
8810 //
8811 //
8812 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..12
8813 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
8814 // CHECK13-NEXT: entry:
8815 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8816 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8817 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8818 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8819 // CHECK13-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
8820 // CHECK13-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
8821 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
8822 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
8823 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8824 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8825 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8826 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
8827 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
8828 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8829 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8830 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8831 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8832 // CHECK13-NEXT: [[I6:%.*]] = alloca i32, align 4
8833 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8834 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8835 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8836 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8837 // CHECK13-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
8838 // CHECK13-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
8839 // CHECK13-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
8840 // CHECK13-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
8841 // CHECK13-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
8842 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
8843 // CHECK13-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
8844 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
8845 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8846 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8847 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8848 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8849 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8850 // CHECK13-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
8851 // CHECK13-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
8852 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
8853 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8854 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8855 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8856 // CHECK13: omp.precond.then:
8857 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8858 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8859 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
8860 // CHECK13-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8861 // CHECK13-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
8862 // CHECK13-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8863 // CHECK13-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
8864 // CHECK13-NEXT: store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
8865 // CHECK13-NEXT: store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
8866 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8867 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8868 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
8869 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8870 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8871 // CHECK13-NEXT: [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8872 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
8873 // CHECK13-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 1073741859, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
8874 // CHECK13-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
8875 // CHECK13: omp.dispatch.cond:
8876 // CHECK13-NEXT: [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8877 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
8878 // CHECK13-NEXT: [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
8879 // CHECK13-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
8880 // CHECK13-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
8881 // CHECK13: omp.dispatch.body:
8882 // CHECK13-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8883 // CHECK13-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
8884 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8885 // CHECK13: omp.inner.for.cond:
8886 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
8887 // CHECK13-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
8888 // CHECK13-NEXT: [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
8889 // CHECK13-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8890 // CHECK13: omp.inner.for.body:
8891 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
8892 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
8893 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8894 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !18
8895 // CHECK13-NEXT: [[TMP21:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !18
8896 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
8897 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
8898 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
8899 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8900 // CHECK13: omp.body.continue:
8901 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8902 // CHECK13: omp.inner.for.inc:
8903 // CHECK13-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
8904 // CHECK13-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
8905 // CHECK13-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
8906 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
8907 // CHECK13: omp.inner.for.end:
8908 // CHECK13-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
8909 // CHECK13: omp.dispatch.inc:
8910 // CHECK13-NEXT: br label [[OMP_DISPATCH_COND]]
8911 // CHECK13: omp.dispatch.end:
8912 // CHECK13-NEXT: br label [[OMP_PRECOND_END]]
8913 // CHECK13: omp.precond.end:
8914 // CHECK13-NEXT: ret void
8915 //
8916 //
8917 // CHECK13-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
8918 // CHECK13-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
8919 // CHECK13-NEXT: entry:
8920 // CHECK13-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
8921 // CHECK13-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
8922 // CHECK13-NEXT: [[M:%.*]] = alloca i32, align 4
8923 // CHECK13-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
8924 // CHECK13-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
8925 // CHECK13-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
8926 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
8927 // CHECK13-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
8928 // CHECK13-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
8929 // CHECK13-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
8930 // CHECK13-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
8931 // CHECK13-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
8932 // CHECK13-NEXT: [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
8933 // CHECK13-NEXT: [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
8934 // CHECK13-NEXT: [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
8935 // CHECK13-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
8936 // CHECK13-NEXT: [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 8
8937 // CHECK13-NEXT: [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 8
8938 // CHECK13-NEXT: [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 8
8939 // CHECK13-NEXT: [[_TMP16:%.*]] = alloca i32, align 4
8940 // CHECK13-NEXT: [[M_CASTED19:%.*]] = alloca i64, align 8
8941 // CHECK13-NEXT: [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [2 x i8*], align 8
8942 // CHECK13-NEXT: [[DOTOFFLOAD_PTRS22:%.*]] = alloca [2 x i8*], align 8
8943 // CHECK13-NEXT: [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [2 x i8*], align 8
8944 // CHECK13-NEXT: [[_TMP24:%.*]] = alloca i32, align 4
8945 // CHECK13-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
8946 // CHECK13-NEXT: store i32 10, i32* [[M]], align 4
8947 // CHECK13-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8948 // CHECK13-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
8949 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
8950 // CHECK13-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8951 // CHECK13-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
8952 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
8953 // CHECK13-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
8954 // CHECK13-NEXT: store i8* null, i8** [[TMP4]], align 8
8955 // CHECK13-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8956 // CHECK13-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8957 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
8958 // CHECK13-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8959 // CHECK13-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
8960 // CHECK13-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8961 // CHECK13: omp_offload.failed:
8962 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116([10 x i32]* [[A]]) #[[ATTR3]]
8963 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT]]
8964 // CHECK13: omp_offload.cont:
8965 // CHECK13-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
8966 // CHECK13-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
8967 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 8
8968 // CHECK13-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
8969 // CHECK13-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
8970 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 8
8971 // CHECK13-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
8972 // CHECK13-NEXT: store i8* null, i8** [[TMP13]], align 8
8973 // CHECK13-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
8974 // CHECK13-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
8975 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
8976 // CHECK13-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8977 // CHECK13-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
8978 // CHECK13-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
8979 // CHECK13: omp_offload.failed5:
8980 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121([10 x i32]* [[A]]) #[[ATTR3]]
8981 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT6]]
8982 // CHECK13: omp_offload.cont6:
8983 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[M]], align 4
8984 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
8985 // CHECK13-NEXT: store i32 [[TMP18]], i32* [[CONV]], align 4
8986 // CHECK13-NEXT: [[TMP19:%.*]] = load i64, i64* [[M_CASTED]], align 8
8987 // CHECK13-NEXT: [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
8988 // CHECK13-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
8989 // CHECK13-NEXT: store i64 [[TMP19]], i64* [[TMP21]], align 8
8990 // CHECK13-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
8991 // CHECK13-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
8992 // CHECK13-NEXT: store i64 [[TMP19]], i64* [[TMP23]], align 8
8993 // CHECK13-NEXT: [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
8994 // CHECK13-NEXT: store i8* null, i8** [[TMP24]], align 8
8995 // CHECK13-NEXT: [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
8996 // CHECK13-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
8997 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 8
8998 // CHECK13-NEXT: [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
8999 // CHECK13-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
9000 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 8
9001 // CHECK13-NEXT: [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
9002 // CHECK13-NEXT: store i8* null, i8** [[TMP29]], align 8
9003 // CHECK13-NEXT: [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
9004 // CHECK13-NEXT: [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
9005 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
9006 // CHECK13-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
9007 // CHECK13-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
9008 // CHECK13-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
9009 // CHECK13: omp_offload.failed11:
9010 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126(i64 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
9011 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT12]]
9012 // CHECK13: omp_offload.cont12:
9013 // CHECK13-NEXT: [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
9014 // CHECK13-NEXT: [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
9015 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 8
9016 // CHECK13-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
9017 // CHECK13-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
9018 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 8
9019 // CHECK13-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i64 0, i64 0
9020 // CHECK13-NEXT: store i8* null, i8** [[TMP38]], align 8
9021 // CHECK13-NEXT: [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
9022 // CHECK13-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
9023 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
9024 // CHECK13-NEXT: [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
9025 // CHECK13-NEXT: [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
9026 // CHECK13-NEXT: br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
9027 // CHECK13: omp_offload.failed17:
9028 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131([10 x i32]* [[A]]) #[[ATTR3]]
9029 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT18]]
9030 // CHECK13: omp_offload.cont18:
9031 // CHECK13-NEXT: [[TMP43:%.*]] = load i32, i32* [[M]], align 4
9032 // CHECK13-NEXT: [[CONV20:%.*]] = bitcast i64* [[M_CASTED19]] to i32*
9033 // CHECK13-NEXT: store i32 [[TMP43]], i32* [[CONV20]], align 4
9034 // CHECK13-NEXT: [[TMP44:%.*]] = load i64, i64* [[M_CASTED19]], align 8
9035 // CHECK13-NEXT: [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
9036 // CHECK13-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
9037 // CHECK13-NEXT: store i64 [[TMP44]], i64* [[TMP46]], align 8
9038 // CHECK13-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
9039 // CHECK13-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
9040 // CHECK13-NEXT: store i64 [[TMP44]], i64* [[TMP48]], align 8
9041 // CHECK13-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
9042 // CHECK13-NEXT: store i8* null, i8** [[TMP49]], align 8
9043 // CHECK13-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
9044 // CHECK13-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
9045 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 8
9046 // CHECK13-NEXT: [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
9047 // CHECK13-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
9048 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 8
9049 // CHECK13-NEXT: [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
9050 // CHECK13-NEXT: store i8* null, i8** [[TMP54]], align 8
9051 // CHECK13-NEXT: [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
9052 // CHECK13-NEXT: [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
9053 // CHECK13-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
9054 // CHECK13-NEXT: [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
9055 // CHECK13-NEXT: [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
9056 // CHECK13-NEXT: br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED25:%.*]], label [[OMP_OFFLOAD_CONT26:%.*]]
9057 // CHECK13: omp_offload.failed25:
9058 // CHECK13-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136(i64 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
9059 // CHECK13-NEXT: br label [[OMP_OFFLOAD_CONT26]]
9060 // CHECK13: omp_offload.cont26:
9061 // CHECK13-NEXT: ret i32 0
9062 //
9063 //
9064 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116
9065 // CHECK13-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9066 // CHECK13-NEXT: entry:
9067 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9068 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9069 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9070 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
9071 // CHECK13-NEXT: ret void
9072 //
9073 //
9074 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..14
9075 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9076 // CHECK13-NEXT: entry:
9077 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9078 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9079 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9080 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9081 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9082 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9083 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9084 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9085 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9086 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9087 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9088 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9089 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9090 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9091 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9092 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
9093 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9094 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9095 // CHECK13-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9096 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9097 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9098 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9099 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
9100 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9101 // CHECK13: cond.true:
9102 // CHECK13-NEXT: br label [[COND_END:%.*]]
9103 // CHECK13: cond.false:
9104 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9105 // CHECK13-NEXT: br label [[COND_END]]
9106 // CHECK13: cond.end:
9107 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9108 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9109 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9110 // CHECK13-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
9111 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9112 // CHECK13: omp.inner.for.cond:
9113 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9114 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9115 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9116 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9117 // CHECK13: omp.inner.for.body:
9118 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9119 // CHECK13-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
9120 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9121 // CHECK13-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
9122 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
9123 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9124 // CHECK13: omp.inner.for.inc:
9125 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9126 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9127 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
9128 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9129 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
9130 // CHECK13: omp.inner.for.end:
9131 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9132 // CHECK13: omp.loop.exit:
9133 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
9134 // CHECK13-NEXT: ret void
9135 //
9136 //
9137 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..15
9138 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9139 // CHECK13-NEXT: entry:
9140 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9141 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9142 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9143 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9144 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9145 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9146 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9147 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9148 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9149 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9150 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9151 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9152 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9153 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9154 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9155 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9156 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9157 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9158 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9159 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
9160 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9161 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
9162 // CHECK13-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9163 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
9164 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9165 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9166 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9167 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9168 // CHECK13-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9169 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
9170 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9171 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9172 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
9173 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9174 // CHECK13: cond.true:
9175 // CHECK13-NEXT: br label [[COND_END:%.*]]
9176 // CHECK13: cond.false:
9177 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9178 // CHECK13-NEXT: br label [[COND_END]]
9179 // CHECK13: cond.end:
9180 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
9181 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9182 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9183 // CHECK13-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
9184 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9185 // CHECK13: omp.inner.for.cond:
9186 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9187 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9188 // CHECK13-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
9189 // CHECK13-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9190 // CHECK13: omp.inner.for.body:
9191 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9192 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
9193 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9194 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9195 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
9196 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
9197 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
9198 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
9199 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9200 // CHECK13: omp.body.continue:
9201 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9202 // CHECK13: omp.inner.for.inc:
9203 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9204 // CHECK13-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
9205 // CHECK13-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
9206 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
9207 // CHECK13: omp.inner.for.end:
9208 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9209 // CHECK13: omp.loop.exit:
9210 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
9211 // CHECK13-NEXT: ret void
9212 //
9213 //
9214 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121
9215 // CHECK13-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9216 // CHECK13-NEXT: entry:
9217 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9218 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9219 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9220 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
9221 // CHECK13-NEXT: ret void
9222 //
9223 //
9224 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..17
9225 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9226 // CHECK13-NEXT: entry:
9227 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9228 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9229 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9230 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9231 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9232 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9233 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9234 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9235 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9236 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9237 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9238 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9239 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9240 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9241 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9242 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
9243 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9244 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9245 // CHECK13-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9246 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9247 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9248 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9249 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
9250 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9251 // CHECK13: cond.true:
9252 // CHECK13-NEXT: br label [[COND_END:%.*]]
9253 // CHECK13: cond.false:
9254 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9255 // CHECK13-NEXT: br label [[COND_END]]
9256 // CHECK13: cond.end:
9257 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9258 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9259 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9260 // CHECK13-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
9261 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9262 // CHECK13: omp.inner.for.cond:
9263 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9264 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9265 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9266 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9267 // CHECK13: omp.inner.for.body:
9268 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9269 // CHECK13-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
9270 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9271 // CHECK13-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
9272 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
9273 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9274 // CHECK13: omp.inner.for.inc:
9275 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9276 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9277 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
9278 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9279 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
9280 // CHECK13: omp.inner.for.end:
9281 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9282 // CHECK13: omp.loop.exit:
9283 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
9284 // CHECK13-NEXT: ret void
9285 //
9286 //
9287 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..18
9288 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9289 // CHECK13-NEXT: entry:
9290 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9291 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9292 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9293 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9294 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9295 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9296 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9297 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9298 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9299 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9300 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9301 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9302 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9303 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9304 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9305 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9306 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9307 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9308 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9309 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
9310 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9311 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
9312 // CHECK13-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9313 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
9314 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9315 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9316 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9317 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9318 // CHECK13-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9319 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
9320 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9321 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9322 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
9323 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9324 // CHECK13: cond.true:
9325 // CHECK13-NEXT: br label [[COND_END:%.*]]
9326 // CHECK13: cond.false:
9327 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9328 // CHECK13-NEXT: br label [[COND_END]]
9329 // CHECK13: cond.end:
9330 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
9331 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9332 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9333 // CHECK13-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
9334 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9335 // CHECK13: omp.inner.for.cond:
9336 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9337 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9338 // CHECK13-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
9339 // CHECK13-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9340 // CHECK13: omp.inner.for.body:
9341 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9342 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
9343 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9344 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9345 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
9346 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
9347 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
9348 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
9349 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9350 // CHECK13: omp.body.continue:
9351 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9352 // CHECK13: omp.inner.for.inc:
9353 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9354 // CHECK13-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
9355 // CHECK13-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
9356 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
9357 // CHECK13: omp.inner.for.end:
9358 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9359 // CHECK13: omp.loop.exit:
9360 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
9361 // CHECK13-NEXT: ret void
9362 //
9363 //
9364 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126
9365 // CHECK13-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9366 // CHECK13-NEXT: entry:
9367 // CHECK13-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
9368 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9369 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
9370 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
9371 // CHECK13-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
9372 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9373 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
9374 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9375 // CHECK13-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
9376 // CHECK13-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
9377 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9378 // CHECK13-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
9379 // CHECK13-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
9380 // CHECK13-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
9381 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
9382 // CHECK13-NEXT: ret void
9383 //
9384 //
9385 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..21
9386 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
9387 // CHECK13-NEXT: entry:
9388 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9389 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9390 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9391 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
9392 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9393 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9394 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9395 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9396 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9397 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9398 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9399 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
9400 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9401 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9402 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9403 // CHECK13-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
9404 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9405 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
9406 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9407 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
9408 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9409 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9410 // CHECK13-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9411 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9412 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9413 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9414 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
9415 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9416 // CHECK13: cond.true:
9417 // CHECK13-NEXT: br label [[COND_END:%.*]]
9418 // CHECK13: cond.false:
9419 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9420 // CHECK13-NEXT: br label [[COND_END]]
9421 // CHECK13: cond.end:
9422 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9423 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9424 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9425 // CHECK13-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
9426 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9427 // CHECK13: omp.inner.for.cond:
9428 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9429 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9430 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9431 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9432 // CHECK13: omp.inner.for.body:
9433 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9434 // CHECK13-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
9435 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9436 // CHECK13-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
9437 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8
9438 // CHECK13-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
9439 // CHECK13-NEXT: store i32 [[TMP12]], i32* [[CONV2]], align 4
9440 // CHECK13-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
9441 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
9442 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9443 // CHECK13: omp.inner.for.inc:
9444 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9445 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9446 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
9447 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9448 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
9449 // CHECK13: omp.inner.for.end:
9450 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9451 // CHECK13: omp.loop.exit:
9452 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
9453 // CHECK13-NEXT: ret void
9454 //
9455 //
9456 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..22
9457 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
9458 // CHECK13-NEXT: entry:
9459 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9460 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9461 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9462 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9463 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9464 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
9465 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9466 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9467 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9468 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9469 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9470 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9471 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9472 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9473 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9474 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9475 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9476 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9477 // CHECK13-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
9478 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9479 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
9480 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9481 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
9482 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9483 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9484 // CHECK13-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9485 // CHECK13-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
9486 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
9487 // CHECK13-NEXT: store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
9488 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9489 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9490 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
9491 // CHECK13-NEXT: [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9492 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
9493 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
9494 // CHECK13-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
9495 // CHECK13: omp.dispatch.cond:
9496 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9497 // CHECK13-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9498 // CHECK13-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
9499 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
9500 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9501 // CHECK13: cond.true:
9502 // CHECK13-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9503 // CHECK13-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
9504 // CHECK13-NEXT: br label [[COND_END:%.*]]
9505 // CHECK13: cond.false:
9506 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9507 // CHECK13-NEXT: br label [[COND_END]]
9508 // CHECK13: cond.end:
9509 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
9510 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9511 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9512 // CHECK13-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
9513 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9514 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9515 // CHECK13-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
9516 // CHECK13-NEXT: br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
9517 // CHECK13: omp.dispatch.body:
9518 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9519 // CHECK13: omp.inner.for.cond:
9520 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9521 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9522 // CHECK13-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
9523 // CHECK13-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9524 // CHECK13: omp.inner.for.body:
9525 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9526 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
9527 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9528 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9529 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4
9530 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
9531 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
9532 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
9533 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9534 // CHECK13: omp.body.continue:
9535 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9536 // CHECK13: omp.inner.for.inc:
9537 // CHECK13-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9538 // CHECK13-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
9539 // CHECK13-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
9540 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
9541 // CHECK13: omp.inner.for.end:
9542 // CHECK13-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
9543 // CHECK13: omp.dispatch.inc:
9544 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9545 // CHECK13-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9546 // CHECK13-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
9547 // CHECK13-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
9548 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9549 // CHECK13-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9550 // CHECK13-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
9551 // CHECK13-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
9552 // CHECK13-NEXT: br label [[OMP_DISPATCH_COND]]
9553 // CHECK13: omp.dispatch.end:
9554 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
9555 // CHECK13-NEXT: ret void
9556 //
9557 //
9558 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
9559 // CHECK13-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9560 // CHECK13-NEXT: entry:
9561 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9562 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9563 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9564 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
9565 // CHECK13-NEXT: ret void
9566 //
9567 //
9568 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..25
9569 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9570 // CHECK13-NEXT: entry:
9571 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9572 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9573 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9574 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9575 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9576 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9577 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9578 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9579 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9580 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9581 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9582 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9583 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9584 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9585 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9586 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
9587 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9588 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9589 // CHECK13-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9590 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9591 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9592 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9593 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
9594 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9595 // CHECK13: cond.true:
9596 // CHECK13-NEXT: br label [[COND_END:%.*]]
9597 // CHECK13: cond.false:
9598 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9599 // CHECK13-NEXT: br label [[COND_END]]
9600 // CHECK13: cond.end:
9601 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9602 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9603 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9604 // CHECK13-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
9605 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9606 // CHECK13: omp.inner.for.cond:
9607 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9608 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9609 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9610 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9611 // CHECK13: omp.inner.for.body:
9612 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9613 // CHECK13-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
9614 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9615 // CHECK13-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
9616 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
9617 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9618 // CHECK13: omp.inner.for.inc:
9619 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9620 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9621 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
9622 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9623 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
9624 // CHECK13: omp.inner.for.end:
9625 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9626 // CHECK13: omp.loop.exit:
9627 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
9628 // CHECK13-NEXT: ret void
9629 //
9630 //
9631 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..26
9632 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9633 // CHECK13-NEXT: entry:
9634 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9635 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9636 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9637 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9638 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9639 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9640 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9641 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9642 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9643 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9644 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9645 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9646 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9647 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9648 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9649 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9650 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9651 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9652 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9653 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
9654 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9655 // CHECK13-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
9656 // CHECK13-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9657 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
9658 // CHECK13-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9659 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9660 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9661 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9662 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9663 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9664 // CHECK13-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9665 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
9666 // CHECK13-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
9667 // CHECK13-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
9668 // CHECK13: omp.dispatch.cond:
9669 // CHECK13-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
9670 // CHECK13-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
9671 // CHECK13-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
9672 // CHECK13: omp.dispatch.body:
9673 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9674 // CHECK13-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
9675 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9676 // CHECK13: omp.inner.for.cond:
9677 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
9678 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
9679 // CHECK13-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
9680 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9681 // CHECK13: omp.inner.for.body:
9682 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
9683 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
9684 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9685 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !21
9686 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !21
9687 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
9688 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
9689 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
9690 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9691 // CHECK13: omp.body.continue:
9692 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9693 // CHECK13: omp.inner.for.inc:
9694 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
9695 // CHECK13-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
9696 // CHECK13-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
9697 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
9698 // CHECK13: omp.inner.for.end:
9699 // CHECK13-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
9700 // CHECK13: omp.dispatch.inc:
9701 // CHECK13-NEXT: br label [[OMP_DISPATCH_COND]]
9702 // CHECK13: omp.dispatch.end:
9703 // CHECK13-NEXT: ret void
9704 //
9705 //
9706 // CHECK13-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136
9707 // CHECK13-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9708 // CHECK13-NEXT: entry:
9709 // CHECK13-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
9710 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9711 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
9712 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
9713 // CHECK13-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
9714 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9715 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
9716 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9717 // CHECK13-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
9718 // CHECK13-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
9719 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9720 // CHECK13-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
9721 // CHECK13-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
9722 // CHECK13-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
9723 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
9724 // CHECK13-NEXT: ret void
9725 //
9726 //
9727 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..29
9728 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
9729 // CHECK13-NEXT: entry:
9730 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9731 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9732 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9733 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
9734 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9735 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9736 // CHECK13-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9737 // CHECK13-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9738 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9739 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9740 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9741 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
9742 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9743 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9744 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9745 // CHECK13-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
9746 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9747 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
9748 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9749 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
9750 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9751 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9752 // CHECK13-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9753 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9754 // CHECK13-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
9755 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9756 // CHECK13-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
9757 // CHECK13-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9758 // CHECK13: cond.true:
9759 // CHECK13-NEXT: br label [[COND_END:%.*]]
9760 // CHECK13: cond.false:
9761 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9762 // CHECK13-NEXT: br label [[COND_END]]
9763 // CHECK13: cond.end:
9764 // CHECK13-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9765 // CHECK13-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9766 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9767 // CHECK13-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
9768 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9769 // CHECK13: omp.inner.for.cond:
9770 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9771 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9772 // CHECK13-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9773 // CHECK13-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9774 // CHECK13: omp.inner.for.body:
9775 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9776 // CHECK13-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
9777 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9778 // CHECK13-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
9779 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8
9780 // CHECK13-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
9781 // CHECK13-NEXT: store i32 [[TMP12]], i32* [[CONV2]], align 4
9782 // CHECK13-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
9783 // CHECK13-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
9784 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9785 // CHECK13: omp.inner.for.inc:
9786 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9787 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9788 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
9789 // CHECK13-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9790 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]]
9791 // CHECK13: omp.inner.for.end:
9792 // CHECK13-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9793 // CHECK13: omp.loop.exit:
9794 // CHECK13-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
9795 // CHECK13-NEXT: ret void
9796 //
9797 //
9798 // CHECK13-LABEL: define {{[^@]+}}@.omp_outlined..30
9799 // CHECK13-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
9800 // CHECK13-NEXT: entry:
9801 // CHECK13-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9802 // CHECK13-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9803 // CHECK13-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9804 // CHECK13-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9805 // CHECK13-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
9806 // CHECK13-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
9807 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9808 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9809 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9810 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9811 // CHECK13-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9812 // CHECK13-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9813 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9814 // CHECK13-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9815 // CHECK13-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9816 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9817 // CHECK13-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9818 // CHECK13-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
9819 // CHECK13-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
9820 // CHECK13-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
9821 // CHECK13-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
9822 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9823 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
9824 // CHECK13-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9825 // CHECK13-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9826 // CHECK13-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9827 // CHECK13-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
9828 // CHECK13-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
9829 // CHECK13-NEXT: store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
9830 // CHECK13-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9831 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9832 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
9833 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9834 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9835 // CHECK13-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9836 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
9837 // CHECK13-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 1073741859, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
9838 // CHECK13-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
9839 // CHECK13: omp.dispatch.cond:
9840 // CHECK13-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
9841 // CHECK13-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
9842 // CHECK13-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
9843 // CHECK13: omp.dispatch.body:
9844 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9845 // CHECK13-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
9846 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9847 // CHECK13: omp.inner.for.cond:
9848 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
9849 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !24
9850 // CHECK13-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
9851 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9852 // CHECK13: omp.inner.for.body:
9853 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
9854 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
9855 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9856 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !24
9857 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !24
9858 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
9859 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
9860 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !24
9861 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9862 // CHECK13: omp.body.continue:
9863 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9864 // CHECK13: omp.inner.for.inc:
9865 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
9866 // CHECK13-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
9867 // CHECK13-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
9868 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
9869 // CHECK13: omp.inner.for.end:
9870 // CHECK13-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
9871 // CHECK13: omp.dispatch.inc:
9872 // CHECK13-NEXT: br label [[OMP_DISPATCH_COND]]
9873 // CHECK13: omp.dispatch.end:
9874 // CHECK13-NEXT: ret void
9875 //
9876 //
9877 // CHECK13-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
9878 // CHECK13-SAME: () #[[ATTR5:[0-9]+]] {
9879 // CHECK13-NEXT: entry:
9880 // CHECK13-NEXT: call void @__tgt_register_requires(i64 1)
9881 // CHECK13-NEXT: ret void
9882 //
9883 //
9884 // CHECK14-LABEL: define {{[^@]+}}@main
9885 // CHECK14-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
9886 // CHECK14-NEXT: entry:
9887 // CHECK14-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
9888 // CHECK14-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
9889 // CHECK14-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
9890 // CHECK14-NEXT: [[N:%.*]] = alloca i32, align 4
9891 // CHECK14-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
9892 // CHECK14-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
9893 // CHECK14-NEXT: [[M:%.*]] = alloca i32, align 4
9894 // CHECK14-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
9895 // CHECK14-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
9896 // CHECK14-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
9897 // CHECK14-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
9898 // CHECK14-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
9899 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9900 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
9901 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
9902 // CHECK14-NEXT: [[N_CASTED3:%.*]] = alloca i64, align 8
9903 // CHECK14-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
9904 // CHECK14-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
9905 // CHECK14-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
9906 // CHECK14-NEXT: [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
9907 // CHECK14-NEXT: [[_TMP9:%.*]] = alloca i32, align 4
9908 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
9909 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
9910 // CHECK14-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
9911 // CHECK14-NEXT: [[N_CASTED19:%.*]] = alloca i64, align 8
9912 // CHECK14-NEXT: [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [4 x i8*], align 8
9913 // CHECK14-NEXT: [[DOTOFFLOAD_PTRS22:%.*]] = alloca [4 x i8*], align 8
9914 // CHECK14-NEXT: [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [4 x i8*], align 8
9915 // CHECK14-NEXT: [[DOTOFFLOAD_SIZES24:%.*]] = alloca [4 x i64], align 8
9916 // CHECK14-NEXT: [[_TMP25:%.*]] = alloca i32, align 4
9917 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
9918 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
9919 // CHECK14-NEXT: [[N_CASTED34:%.*]] = alloca i64, align 8
9920 // CHECK14-NEXT: [[DOTOFFLOAD_BASEPTRS36:%.*]] = alloca [3 x i8*], align 8
9921 // CHECK14-NEXT: [[DOTOFFLOAD_PTRS37:%.*]] = alloca [3 x i8*], align 8
9922 // CHECK14-NEXT: [[DOTOFFLOAD_MAPPERS38:%.*]] = alloca [3 x i8*], align 8
9923 // CHECK14-NEXT: [[DOTOFFLOAD_SIZES39:%.*]] = alloca [3 x i64], align 8
9924 // CHECK14-NEXT: [[_TMP40:%.*]] = alloca i32, align 4
9925 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
9926 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
9927 // CHECK14-NEXT: [[M_CASTED49:%.*]] = alloca i64, align 8
9928 // CHECK14-NEXT: [[N_CASTED51:%.*]] = alloca i64, align 8
9929 // CHECK14-NEXT: [[DOTOFFLOAD_BASEPTRS53:%.*]] = alloca [4 x i8*], align 8
9930 // CHECK14-NEXT: [[DOTOFFLOAD_PTRS54:%.*]] = alloca [4 x i8*], align 8
9931 // CHECK14-NEXT: [[DOTOFFLOAD_MAPPERS55:%.*]] = alloca [4 x i8*], align 8
9932 // CHECK14-NEXT: [[DOTOFFLOAD_SIZES56:%.*]] = alloca [4 x i64], align 8
9933 // CHECK14-NEXT: [[_TMP57:%.*]] = alloca i32, align 4
9934 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_58:%.*]] = alloca i32, align 4
9935 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_59:%.*]] = alloca i32, align 4
9936 // CHECK14-NEXT: store i32 0, i32* [[RETVAL]], align 4
9937 // CHECK14-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
9938 // CHECK14-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
9939 // CHECK14-NEXT: store i32 100, i32* [[N]], align 4
9940 // CHECK14-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
9941 // CHECK14-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
9942 // CHECK14-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
9943 // CHECK14-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
9944 // CHECK14-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
9945 // CHECK14-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
9946 // CHECK14-NEXT: store i32 10, i32* [[M]], align 4
9947 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
9948 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
9949 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
9950 // CHECK14-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
9951 // CHECK14-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
9952 // CHECK14-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9953 // CHECK14-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
9954 // CHECK14-NEXT: store i64 [[TMP4]], i64* [[TMP7]], align 8
9955 // CHECK14-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9956 // CHECK14-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
9957 // CHECK14-NEXT: store i64 [[TMP4]], i64* [[TMP9]], align 8
9958 // CHECK14-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
9959 // CHECK14-NEXT: store i64 4, i64* [[TMP10]], align 8
9960 // CHECK14-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
9961 // CHECK14-NEXT: store i8* null, i8** [[TMP11]], align 8
9962 // CHECK14-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
9963 // CHECK14-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
9964 // CHECK14-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
9965 // CHECK14-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
9966 // CHECK14-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
9967 // CHECK14-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
9968 // CHECK14-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
9969 // CHECK14-NEXT: store i64 8, i64* [[TMP16]], align 8
9970 // CHECK14-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
9971 // CHECK14-NEXT: store i8* null, i8** [[TMP17]], align 8
9972 // CHECK14-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
9973 // CHECK14-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
9974 // CHECK14-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 8
9975 // CHECK14-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
9976 // CHECK14-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
9977 // CHECK14-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 8
9978 // CHECK14-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
9979 // CHECK14-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 8
9980 // CHECK14-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
9981 // CHECK14-NEXT: store i8* null, i8** [[TMP23]], align 8
9982 // CHECK14-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9983 // CHECK14-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9984 // CHECK14-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
9985 // CHECK14-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
9986 // CHECK14-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
9987 // CHECK14-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9988 // CHECK14-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
9989 // CHECK14-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
9990 // CHECK14-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
9991 // CHECK14-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
9992 // CHECK14-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
9993 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
9994 // CHECK14-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
9995 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
9996 // CHECK14-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
9997 // CHECK14-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
9998 // CHECK14-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
9999 // CHECK14: omp_offload.failed:
10000 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
10001 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT]]
10002 // CHECK14: omp_offload.cont:
10003 // CHECK14-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
10004 // CHECK14-NEXT: [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
10005 // CHECK14-NEXT: store i32 [[TMP33]], i32* [[CONV4]], align 4
10006 // CHECK14-NEXT: [[TMP34:%.*]] = load i64, i64* [[N_CASTED3]], align 8
10007 // CHECK14-NEXT: [[TMP35:%.*]] = mul nuw i64 [[TMP1]], 4
10008 // CHECK14-NEXT: [[TMP36:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
10009 // CHECK14-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
10010 // CHECK14-NEXT: store i64 [[TMP34]], i64* [[TMP37]], align 8
10011 // CHECK14-NEXT: [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
10012 // CHECK14-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
10013 // CHECK14-NEXT: store i64 [[TMP34]], i64* [[TMP39]], align 8
10014 // CHECK14-NEXT: [[TMP40:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
10015 // CHECK14-NEXT: store i64 4, i64* [[TMP40]], align 8
10016 // CHECK14-NEXT: [[TMP41:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
10017 // CHECK14-NEXT: store i8* null, i8** [[TMP41]], align 8
10018 // CHECK14-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
10019 // CHECK14-NEXT: [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
10020 // CHECK14-NEXT: store i64 [[TMP1]], i64* [[TMP43]], align 8
10021 // CHECK14-NEXT: [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
10022 // CHECK14-NEXT: [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i64*
10023 // CHECK14-NEXT: store i64 [[TMP1]], i64* [[TMP45]], align 8
10024 // CHECK14-NEXT: [[TMP46:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 1
10025 // CHECK14-NEXT: store i64 8, i64* [[TMP46]], align 8
10026 // CHECK14-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
10027 // CHECK14-NEXT: store i8* null, i8** [[TMP47]], align 8
10028 // CHECK14-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
10029 // CHECK14-NEXT: [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32**
10030 // CHECK14-NEXT: store i32* [[VLA]], i32** [[TMP49]], align 8
10031 // CHECK14-NEXT: [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
10032 // CHECK14-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i32**
10033 // CHECK14-NEXT: store i32* [[VLA]], i32** [[TMP51]], align 8
10034 // CHECK14-NEXT: [[TMP52:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
10035 // CHECK14-NEXT: store i64 [[TMP35]], i64* [[TMP52]], align 8
10036 // CHECK14-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
10037 // CHECK14-NEXT: store i8* null, i8** [[TMP53]], align 8
10038 // CHECK14-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
10039 // CHECK14-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
10040 // CHECK14-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
10041 // CHECK14-NEXT: [[TMP57:%.*]] = load i32, i32* [[N]], align 4
10042 // CHECK14-NEXT: store i32 [[TMP57]], i32* [[DOTCAPTURE_EXPR_10]], align 4
10043 // CHECK14-NEXT: [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
10044 // CHECK14-NEXT: [[SUB12:%.*]] = sub nsw i32 [[TMP58]], 0
10045 // CHECK14-NEXT: [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
10046 // CHECK14-NEXT: [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
10047 // CHECK14-NEXT: store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
10048 // CHECK14-NEXT: [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
10049 // CHECK14-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP59]], 1
10050 // CHECK14-NEXT: [[TMP60:%.*]] = zext i32 [[ADD15]] to i64
10051 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP60]])
10052 // CHECK14-NEXT: [[TMP61:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153.region_id, i32 3, i8** [[TMP54]], i8** [[TMP55]], i64* [[TMP56]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10053 // CHECK14-NEXT: [[TMP62:%.*]] = icmp ne i32 [[TMP61]], 0
10054 // CHECK14-NEXT: br i1 [[TMP62]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
10055 // CHECK14: omp_offload.failed16:
10056 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153(i64 [[TMP34]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
10057 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT17]]
10058 // CHECK14: omp_offload.cont17:
10059 // CHECK14-NEXT: [[TMP63:%.*]] = load i32, i32* [[M]], align 4
10060 // CHECK14-NEXT: [[CONV18:%.*]] = bitcast i64* [[M_CASTED]] to i32*
10061 // CHECK14-NEXT: store i32 [[TMP63]], i32* [[CONV18]], align 4
10062 // CHECK14-NEXT: [[TMP64:%.*]] = load i64, i64* [[M_CASTED]], align 8
10063 // CHECK14-NEXT: [[TMP65:%.*]] = load i32, i32* [[N]], align 4
10064 // CHECK14-NEXT: [[CONV20:%.*]] = bitcast i64* [[N_CASTED19]] to i32*
10065 // CHECK14-NEXT: store i32 [[TMP65]], i32* [[CONV20]], align 4
10066 // CHECK14-NEXT: [[TMP66:%.*]] = load i64, i64* [[N_CASTED19]], align 8
10067 // CHECK14-NEXT: [[TMP67:%.*]] = mul nuw i64 [[TMP1]], 4
10068 // CHECK14-NEXT: [[TMP68:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
10069 // CHECK14-NEXT: [[TMP69:%.*]] = bitcast i8** [[TMP68]] to i64*
10070 // CHECK14-NEXT: store i64 [[TMP64]], i64* [[TMP69]], align 8
10071 // CHECK14-NEXT: [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
10072 // CHECK14-NEXT: [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i64*
10073 // CHECK14-NEXT: store i64 [[TMP64]], i64* [[TMP71]], align 8
10074 // CHECK14-NEXT: [[TMP72:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
10075 // CHECK14-NEXT: store i64 4, i64* [[TMP72]], align 8
10076 // CHECK14-NEXT: [[TMP73:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
10077 // CHECK14-NEXT: store i8* null, i8** [[TMP73]], align 8
10078 // CHECK14-NEXT: [[TMP74:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
10079 // CHECK14-NEXT: [[TMP75:%.*]] = bitcast i8** [[TMP74]] to i64*
10080 // CHECK14-NEXT: store i64 [[TMP66]], i64* [[TMP75]], align 8
10081 // CHECK14-NEXT: [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
10082 // CHECK14-NEXT: [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i64*
10083 // CHECK14-NEXT: store i64 [[TMP66]], i64* [[TMP77]], align 8
10084 // CHECK14-NEXT: [[TMP78:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 1
10085 // CHECK14-NEXT: store i64 4, i64* [[TMP78]], align 8
10086 // CHECK14-NEXT: [[TMP79:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
10087 // CHECK14-NEXT: store i8* null, i8** [[TMP79]], align 8
10088 // CHECK14-NEXT: [[TMP80:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 2
10089 // CHECK14-NEXT: [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
10090 // CHECK14-NEXT: store i64 [[TMP1]], i64* [[TMP81]], align 8
10091 // CHECK14-NEXT: [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 2
10092 // CHECK14-NEXT: [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
10093 // CHECK14-NEXT: store i64 [[TMP1]], i64* [[TMP83]], align 8
10094 // CHECK14-NEXT: [[TMP84:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 2
10095 // CHECK14-NEXT: store i64 8, i64* [[TMP84]], align 8
10096 // CHECK14-NEXT: [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 2
10097 // CHECK14-NEXT: store i8* null, i8** [[TMP85]], align 8
10098 // CHECK14-NEXT: [[TMP86:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 3
10099 // CHECK14-NEXT: [[TMP87:%.*]] = bitcast i8** [[TMP86]] to i32**
10100 // CHECK14-NEXT: store i32* [[VLA]], i32** [[TMP87]], align 8
10101 // CHECK14-NEXT: [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 3
10102 // CHECK14-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
10103 // CHECK14-NEXT: store i32* [[VLA]], i32** [[TMP89]], align 8
10104 // CHECK14-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 3
10105 // CHECK14-NEXT: store i64 [[TMP67]], i64* [[TMP90]], align 8
10106 // CHECK14-NEXT: [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 3
10107 // CHECK14-NEXT: store i8* null, i8** [[TMP91]], align 8
10108 // CHECK14-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
10109 // CHECK14-NEXT: [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
10110 // CHECK14-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
10111 // CHECK14-NEXT: [[TMP95:%.*]] = load i32, i32* [[N]], align 4
10112 // CHECK14-NEXT: store i32 [[TMP95]], i32* [[DOTCAPTURE_EXPR_26]], align 4
10113 // CHECK14-NEXT: [[TMP96:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_26]], align 4
10114 // CHECK14-NEXT: [[SUB28:%.*]] = sub nsw i32 [[TMP96]], 0
10115 // CHECK14-NEXT: [[DIV29:%.*]] = sdiv i32 [[SUB28]], 1
10116 // CHECK14-NEXT: [[SUB30:%.*]] = sub nsw i32 [[DIV29]], 1
10117 // CHECK14-NEXT: store i32 [[SUB30]], i32* [[DOTCAPTURE_EXPR_27]], align 4
10118 // CHECK14-NEXT: [[TMP97:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
10119 // CHECK14-NEXT: [[ADD31:%.*]] = add nsw i32 [[TMP97]], 1
10120 // CHECK14-NEXT: [[TMP98:%.*]] = zext i32 [[ADD31]] to i64
10121 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP98]])
10122 // CHECK14-NEXT: [[TMP99:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158.region_id, i32 4, i8** [[TMP92]], i8** [[TMP93]], i64* [[TMP94]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10123 // CHECK14-NEXT: [[TMP100:%.*]] = icmp ne i32 [[TMP99]], 0
10124 // CHECK14-NEXT: br i1 [[TMP100]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
10125 // CHECK14: omp_offload.failed32:
10126 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158(i64 [[TMP64]], i64 [[TMP66]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
10127 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT33]]
10128 // CHECK14: omp_offload.cont33:
10129 // CHECK14-NEXT: [[TMP101:%.*]] = load i32, i32* [[N]], align 4
10130 // CHECK14-NEXT: [[CONV35:%.*]] = bitcast i64* [[N_CASTED34]] to i32*
10131 // CHECK14-NEXT: store i32 [[TMP101]], i32* [[CONV35]], align 4
10132 // CHECK14-NEXT: [[TMP102:%.*]] = load i64, i64* [[N_CASTED34]], align 8
10133 // CHECK14-NEXT: [[TMP103:%.*]] = mul nuw i64 [[TMP1]], 4
10134 // CHECK14-NEXT: [[TMP104:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
10135 // CHECK14-NEXT: [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i64*
10136 // CHECK14-NEXT: store i64 [[TMP102]], i64* [[TMP105]], align 8
10137 // CHECK14-NEXT: [[TMP106:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
10138 // CHECK14-NEXT: [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i64*
10139 // CHECK14-NEXT: store i64 [[TMP102]], i64* [[TMP107]], align 8
10140 // CHECK14-NEXT: [[TMP108:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
10141 // CHECK14-NEXT: store i64 4, i64* [[TMP108]], align 8
10142 // CHECK14-NEXT: [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 0
10143 // CHECK14-NEXT: store i8* null, i8** [[TMP109]], align 8
10144 // CHECK14-NEXT: [[TMP110:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 1
10145 // CHECK14-NEXT: [[TMP111:%.*]] = bitcast i8** [[TMP110]] to i64*
10146 // CHECK14-NEXT: store i64 [[TMP1]], i64* [[TMP111]], align 8
10147 // CHECK14-NEXT: [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 1
10148 // CHECK14-NEXT: [[TMP113:%.*]] = bitcast i8** [[TMP112]] to i64*
10149 // CHECK14-NEXT: store i64 [[TMP1]], i64* [[TMP113]], align 8
10150 // CHECK14-NEXT: [[TMP114:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 1
10151 // CHECK14-NEXT: store i64 8, i64* [[TMP114]], align 8
10152 // CHECK14-NEXT: [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 1
10153 // CHECK14-NEXT: store i8* null, i8** [[TMP115]], align 8
10154 // CHECK14-NEXT: [[TMP116:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 2
10155 // CHECK14-NEXT: [[TMP117:%.*]] = bitcast i8** [[TMP116]] to i32**
10156 // CHECK14-NEXT: store i32* [[VLA]], i32** [[TMP117]], align 8
10157 // CHECK14-NEXT: [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 2
10158 // CHECK14-NEXT: [[TMP119:%.*]] = bitcast i8** [[TMP118]] to i32**
10159 // CHECK14-NEXT: store i32* [[VLA]], i32** [[TMP119]], align 8
10160 // CHECK14-NEXT: [[TMP120:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 2
10161 // CHECK14-NEXT: store i64 [[TMP103]], i64* [[TMP120]], align 8
10162 // CHECK14-NEXT: [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 2
10163 // CHECK14-NEXT: store i8* null, i8** [[TMP121]], align 8
10164 // CHECK14-NEXT: [[TMP122:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
10165 // CHECK14-NEXT: [[TMP123:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
10166 // CHECK14-NEXT: [[TMP124:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
10167 // CHECK14-NEXT: [[TMP125:%.*]] = load i32, i32* [[N]], align 4
10168 // CHECK14-NEXT: store i32 [[TMP125]], i32* [[DOTCAPTURE_EXPR_41]], align 4
10169 // CHECK14-NEXT: [[TMP126:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
10170 // CHECK14-NEXT: [[SUB43:%.*]] = sub nsw i32 [[TMP126]], 0
10171 // CHECK14-NEXT: [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
10172 // CHECK14-NEXT: [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
10173 // CHECK14-NEXT: store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
10174 // CHECK14-NEXT: [[TMP127:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
10175 // CHECK14-NEXT: [[ADD46:%.*]] = add nsw i32 [[TMP127]], 1
10176 // CHECK14-NEXT: [[TMP128:%.*]] = zext i32 [[ADD46]] to i64
10177 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP128]])
10178 // CHECK14-NEXT: [[TMP129:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163.region_id, i32 3, i8** [[TMP122]], i8** [[TMP123]], i64* [[TMP124]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10179 // CHECK14-NEXT: [[TMP130:%.*]] = icmp ne i32 [[TMP129]], 0
10180 // CHECK14-NEXT: br i1 [[TMP130]], label [[OMP_OFFLOAD_FAILED47:%.*]], label [[OMP_OFFLOAD_CONT48:%.*]]
10181 // CHECK14: omp_offload.failed47:
10182 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163(i64 [[TMP102]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
10183 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT48]]
10184 // CHECK14: omp_offload.cont48:
10185 // CHECK14-NEXT: [[TMP131:%.*]] = load i32, i32* [[M]], align 4
10186 // CHECK14-NEXT: [[CONV50:%.*]] = bitcast i64* [[M_CASTED49]] to i32*
10187 // CHECK14-NEXT: store i32 [[TMP131]], i32* [[CONV50]], align 4
10188 // CHECK14-NEXT: [[TMP132:%.*]] = load i64, i64* [[M_CASTED49]], align 8
10189 // CHECK14-NEXT: [[TMP133:%.*]] = load i32, i32* [[N]], align 4
10190 // CHECK14-NEXT: [[CONV52:%.*]] = bitcast i64* [[N_CASTED51]] to i32*
10191 // CHECK14-NEXT: store i32 [[TMP133]], i32* [[CONV52]], align 4
10192 // CHECK14-NEXT: [[TMP134:%.*]] = load i64, i64* [[N_CASTED51]], align 8
10193 // CHECK14-NEXT: [[TMP135:%.*]] = mul nuw i64 [[TMP1]], 4
10194 // CHECK14-NEXT: [[TMP136:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
10195 // CHECK14-NEXT: [[TMP137:%.*]] = bitcast i8** [[TMP136]] to i64*
10196 // CHECK14-NEXT: store i64 [[TMP132]], i64* [[TMP137]], align 8
10197 // CHECK14-NEXT: [[TMP138:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
10198 // CHECK14-NEXT: [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i64*
10199 // CHECK14-NEXT: store i64 [[TMP132]], i64* [[TMP139]], align 8
10200 // CHECK14-NEXT: [[TMP140:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
10201 // CHECK14-NEXT: store i64 4, i64* [[TMP140]], align 8
10202 // CHECK14-NEXT: [[TMP141:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 0
10203 // CHECK14-NEXT: store i8* null, i8** [[TMP141]], align 8
10204 // CHECK14-NEXT: [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 1
10205 // CHECK14-NEXT: [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i64*
10206 // CHECK14-NEXT: store i64 [[TMP134]], i64* [[TMP143]], align 8
10207 // CHECK14-NEXT: [[TMP144:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 1
10208 // CHECK14-NEXT: [[TMP145:%.*]] = bitcast i8** [[TMP144]] to i64*
10209 // CHECK14-NEXT: store i64 [[TMP134]], i64* [[TMP145]], align 8
10210 // CHECK14-NEXT: [[TMP146:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 1
10211 // CHECK14-NEXT: store i64 4, i64* [[TMP146]], align 8
10212 // CHECK14-NEXT: [[TMP147:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 1
10213 // CHECK14-NEXT: store i8* null, i8** [[TMP147]], align 8
10214 // CHECK14-NEXT: [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 2
10215 // CHECK14-NEXT: [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
10216 // CHECK14-NEXT: store i64 [[TMP1]], i64* [[TMP149]], align 8
10217 // CHECK14-NEXT: [[TMP150:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 2
10218 // CHECK14-NEXT: [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
10219 // CHECK14-NEXT: store i64 [[TMP1]], i64* [[TMP151]], align 8
10220 // CHECK14-NEXT: [[TMP152:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 2
10221 // CHECK14-NEXT: store i64 8, i64* [[TMP152]], align 8
10222 // CHECK14-NEXT: [[TMP153:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 2
10223 // CHECK14-NEXT: store i8* null, i8** [[TMP153]], align 8
10224 // CHECK14-NEXT: [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 3
10225 // CHECK14-NEXT: [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32**
10226 // CHECK14-NEXT: store i32* [[VLA]], i32** [[TMP155]], align 8
10227 // CHECK14-NEXT: [[TMP156:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 3
10228 // CHECK14-NEXT: [[TMP157:%.*]] = bitcast i8** [[TMP156]] to i32**
10229 // CHECK14-NEXT: store i32* [[VLA]], i32** [[TMP157]], align 8
10230 // CHECK14-NEXT: [[TMP158:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 3
10231 // CHECK14-NEXT: store i64 [[TMP135]], i64* [[TMP158]], align 8
10232 // CHECK14-NEXT: [[TMP159:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 3
10233 // CHECK14-NEXT: store i8* null, i8** [[TMP159]], align 8
10234 // CHECK14-NEXT: [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
10235 // CHECK14-NEXT: [[TMP161:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
10236 // CHECK14-NEXT: [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
10237 // CHECK14-NEXT: [[TMP163:%.*]] = load i32, i32* [[N]], align 4
10238 // CHECK14-NEXT: store i32 [[TMP163]], i32* [[DOTCAPTURE_EXPR_58]], align 4
10239 // CHECK14-NEXT: [[TMP164:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_58]], align 4
10240 // CHECK14-NEXT: [[SUB60:%.*]] = sub nsw i32 [[TMP164]], 0
10241 // CHECK14-NEXT: [[DIV61:%.*]] = sdiv i32 [[SUB60]], 1
10242 // CHECK14-NEXT: [[SUB62:%.*]] = sub nsw i32 [[DIV61]], 1
10243 // CHECK14-NEXT: store i32 [[SUB62]], i32* [[DOTCAPTURE_EXPR_59]], align 4
10244 // CHECK14-NEXT: [[TMP165:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_59]], align 4
10245 // CHECK14-NEXT: [[ADD63:%.*]] = add nsw i32 [[TMP165]], 1
10246 // CHECK14-NEXT: [[TMP166:%.*]] = zext i32 [[ADD63]] to i64
10247 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP166]])
10248 // CHECK14-NEXT: [[TMP167:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168.region_id, i32 4, i8** [[TMP160]], i8** [[TMP161]], i64* [[TMP162]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
10249 // CHECK14-NEXT: [[TMP168:%.*]] = icmp ne i32 [[TMP167]], 0
10250 // CHECK14-NEXT: br i1 [[TMP168]], label [[OMP_OFFLOAD_FAILED64:%.*]], label [[OMP_OFFLOAD_CONT65:%.*]]
10251 // CHECK14: omp_offload.failed64:
10252 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168(i64 [[TMP132]], i64 [[TMP134]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
10253 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT65]]
10254 // CHECK14: omp_offload.cont65:
10255 // CHECK14-NEXT: [[TMP169:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
10256 // CHECK14-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP169]])
10257 // CHECK14-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
10258 // CHECK14-NEXT: [[TMP170:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
10259 // CHECK14-NEXT: call void @llvm.stackrestore(i8* [[TMP170]])
10260 // CHECK14-NEXT: [[TMP171:%.*]] = load i32, i32* [[RETVAL]], align 4
10261 // CHECK14-NEXT: ret i32 [[TMP171]]
10262 //
10263 //
10264 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148
10265 // CHECK14-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
10266 // CHECK14-NEXT: entry:
10267 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
10268 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
10269 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
10270 // CHECK14-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
10271 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10272 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
10273 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
10274 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10275 // CHECK14-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
10276 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
10277 // CHECK14-NEXT: ret void
10278 //
10279 //
10280 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined.
10281 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
10282 // CHECK14-NEXT: entry:
10283 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10284 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10285 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
10286 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
10287 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
10288 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10289 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10290 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10291 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
10292 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10293 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10294 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10295 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10296 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10297 // CHECK14-NEXT: [[I3:%.*]] = alloca i32, align 4
10298 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10299 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10300 // CHECK14-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
10301 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10302 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
10303 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
10304 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10305 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
10306 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
10307 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
10308 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10309 // CHECK14-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
10310 // CHECK14-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
10311 // CHECK14-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
10312 // CHECK14-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
10313 // CHECK14-NEXT: store i32 0, i32* [[I]], align 4
10314 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10315 // CHECK14-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
10316 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
10317 // CHECK14: omp.precond.then:
10318 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
10319 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10320 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
10321 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10322 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10323 // CHECK14-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10324 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
10325 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10326 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10327 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10328 // CHECK14-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
10329 // CHECK14-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10330 // CHECK14: cond.true:
10331 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10332 // CHECK14-NEXT: br label [[COND_END:%.*]]
10333 // CHECK14: cond.false:
10334 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10335 // CHECK14-NEXT: br label [[COND_END]]
10336 // CHECK14: cond.end:
10337 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
10338 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
10339 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10340 // CHECK14-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
10341 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10342 // CHECK14: omp.inner.for.cond:
10343 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10344 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10345 // CHECK14-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
10346 // CHECK14-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10347 // CHECK14: omp.inner.for.body:
10348 // CHECK14-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10349 // CHECK14-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
10350 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10351 // CHECK14-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
10352 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
10353 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10354 // CHECK14: omp.inner.for.inc:
10355 // CHECK14-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10356 // CHECK14-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10357 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
10358 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
10359 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
10360 // CHECK14: omp.inner.for.end:
10361 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
10362 // CHECK14: omp.loop.exit:
10363 // CHECK14-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10364 // CHECK14-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
10365 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
10366 // CHECK14-NEXT: br label [[OMP_PRECOND_END]]
10367 // CHECK14: omp.precond.end:
10368 // CHECK14-NEXT: ret void
10369 //
10370 //
10371 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..1
10372 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
10373 // CHECK14-NEXT: entry:
10374 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10375 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10376 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
10377 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
10378 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
10379 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
10380 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
10381 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10382 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10383 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10384 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
10385 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10386 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
10387 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
10388 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10389 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10390 // CHECK14-NEXT: [[I4:%.*]] = alloca i32, align 4
10391 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10392 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10393 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10394 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10395 // CHECK14-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
10396 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10397 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
10398 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
10399 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10400 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
10401 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
10402 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
10403 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10404 // CHECK14-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
10405 // CHECK14-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
10406 // CHECK14-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
10407 // CHECK14-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
10408 // CHECK14-NEXT: store i32 0, i32* [[I]], align 4
10409 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10410 // CHECK14-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
10411 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
10412 // CHECK14: omp.precond.then:
10413 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
10414 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10415 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
10416 // CHECK14-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10417 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
10418 // CHECK14-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10419 // CHECK14-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
10420 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
10421 // CHECK14-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
10422 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10423 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10424 // CHECK14-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10425 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
10426 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10427 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10428 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10429 // CHECK14-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
10430 // CHECK14-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10431 // CHECK14: cond.true:
10432 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10433 // CHECK14-NEXT: br label [[COND_END:%.*]]
10434 // CHECK14: cond.false:
10435 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10436 // CHECK14-NEXT: br label [[COND_END]]
10437 // CHECK14: cond.end:
10438 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
10439 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10440 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10441 // CHECK14-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
10442 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10443 // CHECK14: omp.inner.for.cond:
10444 // CHECK14-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10445 // CHECK14-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10446 // CHECK14-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
10447 // CHECK14-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10448 // CHECK14: omp.inner.for.body:
10449 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10450 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
10451 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10452 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
10453 // CHECK14-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
10454 // CHECK14-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
10455 // CHECK14-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
10456 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
10457 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
10458 // CHECK14: omp.body.continue:
10459 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10460 // CHECK14: omp.inner.for.inc:
10461 // CHECK14-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10462 // CHECK14-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
10463 // CHECK14-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
10464 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
10465 // CHECK14: omp.inner.for.end:
10466 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
10467 // CHECK14: omp.loop.exit:
10468 // CHECK14-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10469 // CHECK14-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
10470 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
10471 // CHECK14-NEXT: br label [[OMP_PRECOND_END]]
10472 // CHECK14: omp.precond.end:
10473 // CHECK14-NEXT: ret void
10474 //
10475 //
10476 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153
10477 // CHECK14-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
10478 // CHECK14-NEXT: entry:
10479 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
10480 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
10481 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
10482 // CHECK14-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
10483 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10484 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
10485 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
10486 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10487 // CHECK14-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
10488 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
10489 // CHECK14-NEXT: ret void
10490 //
10491 //
10492 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..2
10493 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
10494 // CHECK14-NEXT: entry:
10495 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10496 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10497 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
10498 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
10499 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
10500 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10501 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10502 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10503 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
10504 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10505 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10506 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10507 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10508 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10509 // CHECK14-NEXT: [[I3:%.*]] = alloca i32, align 4
10510 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10511 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10512 // CHECK14-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
10513 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10514 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
10515 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
10516 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10517 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
10518 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
10519 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
10520 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10521 // CHECK14-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
10522 // CHECK14-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
10523 // CHECK14-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
10524 // CHECK14-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
10525 // CHECK14-NEXT: store i32 0, i32* [[I]], align 4
10526 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10527 // CHECK14-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
10528 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
10529 // CHECK14: omp.precond.then:
10530 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
10531 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10532 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
10533 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10534 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10535 // CHECK14-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10536 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
10537 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10538 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10539 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10540 // CHECK14-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
10541 // CHECK14-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10542 // CHECK14: cond.true:
10543 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10544 // CHECK14-NEXT: br label [[COND_END:%.*]]
10545 // CHECK14: cond.false:
10546 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10547 // CHECK14-NEXT: br label [[COND_END]]
10548 // CHECK14: cond.end:
10549 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
10550 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
10551 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10552 // CHECK14-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
10553 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10554 // CHECK14: omp.inner.for.cond:
10555 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10556 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10557 // CHECK14-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
10558 // CHECK14-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10559 // CHECK14: omp.inner.for.body:
10560 // CHECK14-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10561 // CHECK14-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
10562 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10563 // CHECK14-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
10564 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
10565 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10566 // CHECK14: omp.inner.for.inc:
10567 // CHECK14-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10568 // CHECK14-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10569 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
10570 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
10571 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
10572 // CHECK14: omp.inner.for.end:
10573 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
10574 // CHECK14: omp.loop.exit:
10575 // CHECK14-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10576 // CHECK14-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
10577 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
10578 // CHECK14-NEXT: br label [[OMP_PRECOND_END]]
10579 // CHECK14: omp.precond.end:
10580 // CHECK14-NEXT: ret void
10581 //
10582 //
10583 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..3
10584 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
10585 // CHECK14-NEXT: entry:
10586 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10587 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10588 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
10589 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
10590 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
10591 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
10592 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
10593 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10594 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10595 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10596 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
10597 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10598 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
10599 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
10600 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10601 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10602 // CHECK14-NEXT: [[I4:%.*]] = alloca i32, align 4
10603 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10604 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10605 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10606 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10607 // CHECK14-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
10608 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10609 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
10610 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
10611 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10612 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
10613 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
10614 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
10615 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10616 // CHECK14-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
10617 // CHECK14-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
10618 // CHECK14-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
10619 // CHECK14-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
10620 // CHECK14-NEXT: store i32 0, i32* [[I]], align 4
10621 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10622 // CHECK14-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
10623 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
10624 // CHECK14: omp.precond.then:
10625 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
10626 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10627 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
10628 // CHECK14-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10629 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
10630 // CHECK14-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10631 // CHECK14-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
10632 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
10633 // CHECK14-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
10634 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10635 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10636 // CHECK14-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10637 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
10638 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10639 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10640 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10641 // CHECK14-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
10642 // CHECK14-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10643 // CHECK14: cond.true:
10644 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10645 // CHECK14-NEXT: br label [[COND_END:%.*]]
10646 // CHECK14: cond.false:
10647 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10648 // CHECK14-NEXT: br label [[COND_END]]
10649 // CHECK14: cond.end:
10650 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
10651 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10652 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10653 // CHECK14-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
10654 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10655 // CHECK14: omp.inner.for.cond:
10656 // CHECK14-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10657 // CHECK14-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10658 // CHECK14-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
10659 // CHECK14-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10660 // CHECK14: omp.inner.for.body:
10661 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10662 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
10663 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10664 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
10665 // CHECK14-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
10666 // CHECK14-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
10667 // CHECK14-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
10668 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
10669 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
10670 // CHECK14: omp.body.continue:
10671 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10672 // CHECK14: omp.inner.for.inc:
10673 // CHECK14-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10674 // CHECK14-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
10675 // CHECK14-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
10676 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
10677 // CHECK14: omp.inner.for.end:
10678 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
10679 // CHECK14: omp.loop.exit:
10680 // CHECK14-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10681 // CHECK14-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
10682 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
10683 // CHECK14-NEXT: br label [[OMP_PRECOND_END]]
10684 // CHECK14: omp.precond.end:
10685 // CHECK14-NEXT: ret void
10686 //
10687 //
10688 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158
10689 // CHECK14-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
10690 // CHECK14-NEXT: entry:
10691 // CHECK14-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
10692 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
10693 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
10694 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
10695 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10696 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
10697 // CHECK14-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
10698 // CHECK14-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
10699 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10700 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
10701 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
10702 // CHECK14-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
10703 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10704 // CHECK14-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
10705 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
10706 // CHECK14-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
10707 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10708 // CHECK14-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
10709 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
10710 // CHECK14-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
10711 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
10712 // CHECK14-NEXT: ret void
10713 //
10714 //
10715 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..5
10716 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
10717 // CHECK14-NEXT: entry:
10718 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10719 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10720 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
10721 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
10722 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
10723 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
10724 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10725 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10726 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
10727 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
10728 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10729 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10730 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10731 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10732 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10733 // CHECK14-NEXT: [[I4:%.*]] = alloca i32, align 4
10734 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
10735 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10736 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10737 // CHECK14-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
10738 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10739 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
10740 // CHECK14-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
10741 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
10742 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10743 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
10744 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
10745 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
10746 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
10747 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10748 // CHECK14-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
10749 // CHECK14-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
10750 // CHECK14-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
10751 // CHECK14-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
10752 // CHECK14-NEXT: store i32 0, i32* [[I]], align 4
10753 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10754 // CHECK14-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
10755 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
10756 // CHECK14: omp.precond.then:
10757 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
10758 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10759 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
10760 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10761 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10762 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[CONV]], align 8
10763 // CHECK14-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10764 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
10765 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
10766 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10767 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10768 // CHECK14-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
10769 // CHECK14-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10770 // CHECK14: cond.true:
10771 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10772 // CHECK14-NEXT: br label [[COND_END:%.*]]
10773 // CHECK14: cond.false:
10774 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10775 // CHECK14-NEXT: br label [[COND_END]]
10776 // CHECK14: cond.end:
10777 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
10778 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
10779 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10780 // CHECK14-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
10781 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10782 // CHECK14: omp.inner.for.cond:
10783 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10784 // CHECK14-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10785 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
10786 // CHECK14-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
10787 // CHECK14-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10788 // CHECK14: omp.inner.for.body:
10789 // CHECK14-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10790 // CHECK14-NEXT: [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
10791 // CHECK14-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10792 // CHECK14-NEXT: [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
10793 // CHECK14-NEXT: [[TMP21:%.*]] = load i32, i32* [[CONV]], align 8
10794 // CHECK14-NEXT: [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
10795 // CHECK14-NEXT: store i32 [[TMP21]], i32* [[CONV7]], align 4
10796 // CHECK14-NEXT: [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
10797 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]])
10798 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10799 // CHECK14: omp.inner.for.inc:
10800 // CHECK14-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10801 // CHECK14-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10802 // CHECK14-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
10803 // CHECK14-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
10804 // CHECK14-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10805 // CHECK14-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10806 // CHECK14-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
10807 // CHECK14-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4
10808 // CHECK14-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10809 // CHECK14-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10810 // CHECK14-NEXT: [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
10811 // CHECK14-NEXT: store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4
10812 // CHECK14-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10813 // CHECK14-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10814 // CHECK14-NEXT: [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
10815 // CHECK14-NEXT: br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
10816 // CHECK14: cond.true12:
10817 // CHECK14-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10818 // CHECK14-NEXT: br label [[COND_END14:%.*]]
10819 // CHECK14: cond.false13:
10820 // CHECK14-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10821 // CHECK14-NEXT: br label [[COND_END14]]
10822 // CHECK14: cond.end14:
10823 // CHECK14-NEXT: [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
10824 // CHECK14-NEXT: store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4
10825 // CHECK14-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10826 // CHECK14-NEXT: store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4
10827 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
10828 // CHECK14: omp.inner.for.end:
10829 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
10830 // CHECK14: omp.loop.exit:
10831 // CHECK14-NEXT: [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10832 // CHECK14-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
10833 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
10834 // CHECK14-NEXT: br label [[OMP_PRECOND_END]]
10835 // CHECK14: omp.precond.end:
10836 // CHECK14-NEXT: ret void
10837 //
10838 //
10839 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..6
10840 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
10841 // CHECK14-NEXT: entry:
10842 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10843 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10844 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
10845 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
10846 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
10847 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
10848 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
10849 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
10850 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10851 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10852 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
10853 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
10854 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10855 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
10856 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
10857 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10858 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10859 // CHECK14-NEXT: [[I6:%.*]] = alloca i32, align 4
10860 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10861 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10862 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10863 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10864 // CHECK14-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
10865 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10866 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
10867 // CHECK14-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
10868 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
10869 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10870 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
10871 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
10872 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
10873 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
10874 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10875 // CHECK14-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
10876 // CHECK14-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
10877 // CHECK14-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
10878 // CHECK14-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
10879 // CHECK14-NEXT: store i32 0, i32* [[I]], align 4
10880 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10881 // CHECK14-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
10882 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
10883 // CHECK14: omp.precond.then:
10884 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
10885 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10886 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
10887 // CHECK14-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10888 // CHECK14-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
10889 // CHECK14-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10890 // CHECK14-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
10891 // CHECK14-NEXT: store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
10892 // CHECK14-NEXT: store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
10893 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10894 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10895 // CHECK14-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10896 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
10897 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
10898 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10899 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10900 // CHECK14-NEXT: [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
10901 // CHECK14-NEXT: br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10902 // CHECK14: cond.true:
10903 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
10904 // CHECK14-NEXT: br label [[COND_END:%.*]]
10905 // CHECK14: cond.false:
10906 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10907 // CHECK14-NEXT: br label [[COND_END]]
10908 // CHECK14: cond.end:
10909 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
10910 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10911 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10912 // CHECK14-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
10913 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10914 // CHECK14: omp.inner.for.cond:
10915 // CHECK14-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10916 // CHECK14-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10917 // CHECK14-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
10918 // CHECK14-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10919 // CHECK14: omp.inner.for.body:
10920 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10921 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
10922 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10923 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I6]], align 4
10924 // CHECK14-NEXT: [[TMP19:%.*]] = load i32, i32* [[I6]], align 4
10925 // CHECK14-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
10926 // CHECK14-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
10927 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
10928 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
10929 // CHECK14: omp.body.continue:
10930 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10931 // CHECK14: omp.inner.for.inc:
10932 // CHECK14-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10933 // CHECK14-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
10934 // CHECK14-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
10935 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
10936 // CHECK14: omp.inner.for.end:
10937 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
10938 // CHECK14: omp.loop.exit:
10939 // CHECK14-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10940 // CHECK14-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
10941 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
10942 // CHECK14-NEXT: br label [[OMP_PRECOND_END]]
10943 // CHECK14: omp.precond.end:
10944 // CHECK14-NEXT: ret void
10945 //
10946 //
10947 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163
10948 // CHECK14-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
10949 // CHECK14-NEXT: entry:
10950 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
10951 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
10952 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
10953 // CHECK14-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
10954 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10955 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
10956 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
10957 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10958 // CHECK14-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
10959 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
10960 // CHECK14-NEXT: ret void
10961 //
10962 //
10963 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..8
10964 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
10965 // CHECK14-NEXT: entry:
10966 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10967 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10968 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
10969 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
10970 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
10971 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10972 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
10973 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10974 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
10975 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
10976 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10977 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10978 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10979 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10980 // CHECK14-NEXT: [[I3:%.*]] = alloca i32, align 4
10981 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10982 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10983 // CHECK14-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
10984 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
10985 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
10986 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
10987 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
10988 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
10989 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
10990 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
10991 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10992 // CHECK14-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
10993 // CHECK14-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
10994 // CHECK14-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
10995 // CHECK14-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
10996 // CHECK14-NEXT: store i32 0, i32* [[I]], align 4
10997 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10998 // CHECK14-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
10999 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
11000 // CHECK14: omp.precond.then:
11001 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
11002 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
11003 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
11004 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11005 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11006 // CHECK14-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11007 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
11008 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11009 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11010 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
11011 // CHECK14-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
11012 // CHECK14-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11013 // CHECK14: cond.true:
11014 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
11015 // CHECK14-NEXT: br label [[COND_END:%.*]]
11016 // CHECK14: cond.false:
11017 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11018 // CHECK14-NEXT: br label [[COND_END]]
11019 // CHECK14: cond.end:
11020 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
11021 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
11022 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11023 // CHECK14-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
11024 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
11025 // CHECK14: omp.inner.for.cond:
11026 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11027 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11028 // CHECK14-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
11029 // CHECK14-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11030 // CHECK14: omp.inner.for.body:
11031 // CHECK14-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11032 // CHECK14-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
11033 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11034 // CHECK14-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
11035 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
11036 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
11037 // CHECK14: omp.inner.for.inc:
11038 // CHECK14-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11039 // CHECK14-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
11040 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
11041 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
11042 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
11043 // CHECK14: omp.inner.for.end:
11044 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
11045 // CHECK14: omp.loop.exit:
11046 // CHECK14-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11047 // CHECK14-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
11048 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
11049 // CHECK14-NEXT: br label [[OMP_PRECOND_END]]
11050 // CHECK14: omp.precond.end:
11051 // CHECK14-NEXT: ret void
11052 //
11053 //
11054 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..9
11055 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
11056 // CHECK14-NEXT: entry:
11057 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11058 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11059 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
11060 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
11061 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
11062 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
11063 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
11064 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
11065 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
11066 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
11067 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
11068 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
11069 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
11070 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
11071 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11072 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11073 // CHECK14-NEXT: [[I4:%.*]] = alloca i32, align 4
11074 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11075 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11076 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11077 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11078 // CHECK14-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
11079 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
11080 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
11081 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
11082 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
11083 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
11084 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
11085 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
11086 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
11087 // CHECK14-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
11088 // CHECK14-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
11089 // CHECK14-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
11090 // CHECK14-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
11091 // CHECK14-NEXT: store i32 0, i32* [[I]], align 4
11092 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
11093 // CHECK14-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
11094 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
11095 // CHECK14: omp.precond.then:
11096 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
11097 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
11098 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
11099 // CHECK14-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11100 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
11101 // CHECK14-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11102 // CHECK14-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
11103 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
11104 // CHECK14-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
11105 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11106 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11107 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11108 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11109 // CHECK14-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11110 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
11111 // CHECK14-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 1073741859, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
11112 // CHECK14-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
11113 // CHECK14: omp.dispatch.cond:
11114 // CHECK14-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11115 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
11116 // CHECK14-NEXT: [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
11117 // CHECK14-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
11118 // CHECK14-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
11119 // CHECK14: omp.dispatch.body:
11120 // CHECK14-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11121 // CHECK14-NEXT: store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
11122 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
11123 // CHECK14: omp.inner.for.cond:
11124 // CHECK14-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
11125 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !15
11126 // CHECK14-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
11127 // CHECK14-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11128 // CHECK14: omp.inner.for.body:
11129 // CHECK14-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
11130 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
11131 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11132 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !15
11133 // CHECK14-NEXT: [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !15
11134 // CHECK14-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
11135 // CHECK14-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
11136 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !15
11137 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
11138 // CHECK14: omp.body.continue:
11139 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
11140 // CHECK14: omp.inner.for.inc:
11141 // CHECK14-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
11142 // CHECK14-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP21]], 1
11143 // CHECK14-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
11144 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
11145 // CHECK14: omp.inner.for.end:
11146 // CHECK14-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
11147 // CHECK14: omp.dispatch.inc:
11148 // CHECK14-NEXT: br label [[OMP_DISPATCH_COND]]
11149 // CHECK14: omp.dispatch.end:
11150 // CHECK14-NEXT: br label [[OMP_PRECOND_END]]
11151 // CHECK14: omp.precond.end:
11152 // CHECK14-NEXT: ret void
11153 //
11154 //
11155 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168
11156 // CHECK14-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
11157 // CHECK14-NEXT: entry:
11158 // CHECK14-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
11159 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
11160 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
11161 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
11162 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
11163 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
11164 // CHECK14-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
11165 // CHECK14-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
11166 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
11167 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
11168 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
11169 // CHECK14-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
11170 // CHECK14-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
11171 // CHECK14-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
11172 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
11173 // CHECK14-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
11174 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
11175 // CHECK14-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
11176 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
11177 // CHECK14-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
11178 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
11179 // CHECK14-NEXT: ret void
11180 //
11181 //
11182 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..11
11183 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
11184 // CHECK14-NEXT: entry:
11185 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11186 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11187 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
11188 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
11189 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
11190 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
11191 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
11192 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
11193 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
11194 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
11195 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
11196 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
11197 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
11198 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11199 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11200 // CHECK14-NEXT: [[I4:%.*]] = alloca i32, align 4
11201 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
11202 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11203 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11204 // CHECK14-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
11205 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
11206 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
11207 // CHECK14-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
11208 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
11209 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
11210 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
11211 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
11212 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
11213 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
11214 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
11215 // CHECK14-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
11216 // CHECK14-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
11217 // CHECK14-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
11218 // CHECK14-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
11219 // CHECK14-NEXT: store i32 0, i32* [[I]], align 4
11220 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
11221 // CHECK14-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
11222 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
11223 // CHECK14: omp.precond.then:
11224 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
11225 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
11226 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
11227 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11228 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11229 // CHECK14-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11230 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
11231 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11232 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11233 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
11234 // CHECK14-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
11235 // CHECK14-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11236 // CHECK14: cond.true:
11237 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
11238 // CHECK14-NEXT: br label [[COND_END:%.*]]
11239 // CHECK14: cond.false:
11240 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11241 // CHECK14-NEXT: br label [[COND_END]]
11242 // CHECK14: cond.end:
11243 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
11244 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
11245 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11246 // CHECK14-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
11247 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
11248 // CHECK14: omp.inner.for.cond:
11249 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11250 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11251 // CHECK14-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
11252 // CHECK14-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11253 // CHECK14: omp.inner.for.body:
11254 // CHECK14-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11255 // CHECK14-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
11256 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11257 // CHECK14-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
11258 // CHECK14-NEXT: [[TMP20:%.*]] = load i32, i32* [[CONV]], align 8
11259 // CHECK14-NEXT: [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
11260 // CHECK14-NEXT: store i32 [[TMP20]], i32* [[CONV7]], align 4
11261 // CHECK14-NEXT: [[TMP21:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
11262 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP21]])
11263 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
11264 // CHECK14: omp.inner.for.inc:
11265 // CHECK14-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11266 // CHECK14-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
11267 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
11268 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
11269 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
11270 // CHECK14: omp.inner.for.end:
11271 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
11272 // CHECK14: omp.loop.exit:
11273 // CHECK14-NEXT: [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11274 // CHECK14-NEXT: [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
11275 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
11276 // CHECK14-NEXT: br label [[OMP_PRECOND_END]]
11277 // CHECK14: omp.precond.end:
11278 // CHECK14-NEXT: ret void
11279 //
11280 //
11281 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..12
11282 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
11283 // CHECK14-NEXT: entry:
11284 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11285 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11286 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
11287 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
11288 // CHECK14-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
11289 // CHECK14-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
11290 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
11291 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
11292 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
11293 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
11294 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
11295 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
11296 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
11297 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
11298 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
11299 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11300 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11301 // CHECK14-NEXT: [[I6:%.*]] = alloca i32, align 4
11302 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11303 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11304 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11305 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11306 // CHECK14-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
11307 // CHECK14-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
11308 // CHECK14-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
11309 // CHECK14-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
11310 // CHECK14-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
11311 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
11312 // CHECK14-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
11313 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
11314 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
11315 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
11316 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
11317 // CHECK14-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
11318 // CHECK14-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
11319 // CHECK14-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
11320 // CHECK14-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
11321 // CHECK14-NEXT: store i32 0, i32* [[I]], align 4
11322 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
11323 // CHECK14-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
11324 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
11325 // CHECK14: omp.precond.then:
11326 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
11327 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
11328 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
11329 // CHECK14-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11330 // CHECK14-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
11331 // CHECK14-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11332 // CHECK14-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
11333 // CHECK14-NEXT: store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
11334 // CHECK14-NEXT: store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
11335 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11336 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11337 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
11338 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11339 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11340 // CHECK14-NEXT: [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11341 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
11342 // CHECK14-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 1073741859, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
11343 // CHECK14-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
11344 // CHECK14: omp.dispatch.cond:
11345 // CHECK14-NEXT: [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11346 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
11347 // CHECK14-NEXT: [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
11348 // CHECK14-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
11349 // CHECK14-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
11350 // CHECK14: omp.dispatch.body:
11351 // CHECK14-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11352 // CHECK14-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
11353 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
11354 // CHECK14: omp.inner.for.cond:
11355 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
11356 // CHECK14-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
11357 // CHECK14-NEXT: [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
11358 // CHECK14-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11359 // CHECK14: omp.inner.for.body:
11360 // CHECK14-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
11361 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
11362 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11363 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !18
11364 // CHECK14-NEXT: [[TMP21:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !18
11365 // CHECK14-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
11366 // CHECK14-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
11367 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
11368 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
11369 // CHECK14: omp.body.continue:
11370 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
11371 // CHECK14: omp.inner.for.inc:
11372 // CHECK14-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
11373 // CHECK14-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
11374 // CHECK14-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
11375 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
11376 // CHECK14: omp.inner.for.end:
11377 // CHECK14-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
11378 // CHECK14: omp.dispatch.inc:
11379 // CHECK14-NEXT: br label [[OMP_DISPATCH_COND]]
11380 // CHECK14: omp.dispatch.end:
11381 // CHECK14-NEXT: br label [[OMP_PRECOND_END]]
11382 // CHECK14: omp.precond.end:
11383 // CHECK14-NEXT: ret void
11384 //
11385 //
11386 // CHECK14-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
11387 // CHECK14-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
11388 // CHECK14-NEXT: entry:
11389 // CHECK14-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
11390 // CHECK14-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
11391 // CHECK14-NEXT: [[M:%.*]] = alloca i32, align 4
11392 // CHECK14-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
11393 // CHECK14-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
11394 // CHECK14-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
11395 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
11396 // CHECK14-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
11397 // CHECK14-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
11398 // CHECK14-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
11399 // CHECK14-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
11400 // CHECK14-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
11401 // CHECK14-NEXT: [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
11402 // CHECK14-NEXT: [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
11403 // CHECK14-NEXT: [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
11404 // CHECK14-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
11405 // CHECK14-NEXT: [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 8
11406 // CHECK14-NEXT: [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 8
11407 // CHECK14-NEXT: [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 8
11408 // CHECK14-NEXT: [[_TMP16:%.*]] = alloca i32, align 4
11409 // CHECK14-NEXT: [[M_CASTED19:%.*]] = alloca i64, align 8
11410 // CHECK14-NEXT: [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [2 x i8*], align 8
11411 // CHECK14-NEXT: [[DOTOFFLOAD_PTRS22:%.*]] = alloca [2 x i8*], align 8
11412 // CHECK14-NEXT: [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [2 x i8*], align 8
11413 // CHECK14-NEXT: [[_TMP24:%.*]] = alloca i32, align 4
11414 // CHECK14-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
11415 // CHECK14-NEXT: store i32 10, i32* [[M]], align 4
11416 // CHECK14-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11417 // CHECK14-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
11418 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
11419 // CHECK14-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11420 // CHECK14-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
11421 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
11422 // CHECK14-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
11423 // CHECK14-NEXT: store i8* null, i8** [[TMP4]], align 8
11424 // CHECK14-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11425 // CHECK14-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11426 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
11427 // CHECK14-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
11428 // CHECK14-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
11429 // CHECK14-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
11430 // CHECK14: omp_offload.failed:
11431 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116([10 x i32]* [[A]]) #[[ATTR3]]
11432 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT]]
11433 // CHECK14: omp_offload.cont:
11434 // CHECK14-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
11435 // CHECK14-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
11436 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 8
11437 // CHECK14-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
11438 // CHECK14-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
11439 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 8
11440 // CHECK14-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
11441 // CHECK14-NEXT: store i8* null, i8** [[TMP13]], align 8
11442 // CHECK14-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
11443 // CHECK14-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
11444 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
11445 // CHECK14-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
11446 // CHECK14-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
11447 // CHECK14-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
11448 // CHECK14: omp_offload.failed5:
11449 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121([10 x i32]* [[A]]) #[[ATTR3]]
11450 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT6]]
11451 // CHECK14: omp_offload.cont6:
11452 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[M]], align 4
11453 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
11454 // CHECK14-NEXT: store i32 [[TMP18]], i32* [[CONV]], align 4
11455 // CHECK14-NEXT: [[TMP19:%.*]] = load i64, i64* [[M_CASTED]], align 8
11456 // CHECK14-NEXT: [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
11457 // CHECK14-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
11458 // CHECK14-NEXT: store i64 [[TMP19]], i64* [[TMP21]], align 8
11459 // CHECK14-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
11460 // CHECK14-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
11461 // CHECK14-NEXT: store i64 [[TMP19]], i64* [[TMP23]], align 8
11462 // CHECK14-NEXT: [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
11463 // CHECK14-NEXT: store i8* null, i8** [[TMP24]], align 8
11464 // CHECK14-NEXT: [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
11465 // CHECK14-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
11466 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 8
11467 // CHECK14-NEXT: [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
11468 // CHECK14-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
11469 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 8
11470 // CHECK14-NEXT: [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
11471 // CHECK14-NEXT: store i8* null, i8** [[TMP29]], align 8
11472 // CHECK14-NEXT: [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
11473 // CHECK14-NEXT: [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
11474 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
11475 // CHECK14-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
11476 // CHECK14-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
11477 // CHECK14-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
11478 // CHECK14: omp_offload.failed11:
11479 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126(i64 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
11480 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT12]]
11481 // CHECK14: omp_offload.cont12:
11482 // CHECK14-NEXT: [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
11483 // CHECK14-NEXT: [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
11484 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 8
11485 // CHECK14-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
11486 // CHECK14-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
11487 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 8
11488 // CHECK14-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i64 0, i64 0
11489 // CHECK14-NEXT: store i8* null, i8** [[TMP38]], align 8
11490 // CHECK14-NEXT: [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
11491 // CHECK14-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
11492 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
11493 // CHECK14-NEXT: [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
11494 // CHECK14-NEXT: [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
11495 // CHECK14-NEXT: br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
11496 // CHECK14: omp_offload.failed17:
11497 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131([10 x i32]* [[A]]) #[[ATTR3]]
11498 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT18]]
11499 // CHECK14: omp_offload.cont18:
11500 // CHECK14-NEXT: [[TMP43:%.*]] = load i32, i32* [[M]], align 4
11501 // CHECK14-NEXT: [[CONV20:%.*]] = bitcast i64* [[M_CASTED19]] to i32*
11502 // CHECK14-NEXT: store i32 [[TMP43]], i32* [[CONV20]], align 4
11503 // CHECK14-NEXT: [[TMP44:%.*]] = load i64, i64* [[M_CASTED19]], align 8
11504 // CHECK14-NEXT: [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
11505 // CHECK14-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
11506 // CHECK14-NEXT: store i64 [[TMP44]], i64* [[TMP46]], align 8
11507 // CHECK14-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
11508 // CHECK14-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
11509 // CHECK14-NEXT: store i64 [[TMP44]], i64* [[TMP48]], align 8
11510 // CHECK14-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
11511 // CHECK14-NEXT: store i8* null, i8** [[TMP49]], align 8
11512 // CHECK14-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
11513 // CHECK14-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
11514 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 8
11515 // CHECK14-NEXT: [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
11516 // CHECK14-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
11517 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 8
11518 // CHECK14-NEXT: [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
11519 // CHECK14-NEXT: store i8* null, i8** [[TMP54]], align 8
11520 // CHECK14-NEXT: [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
11521 // CHECK14-NEXT: [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
11522 // CHECK14-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
11523 // CHECK14-NEXT: [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
11524 // CHECK14-NEXT: [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
11525 // CHECK14-NEXT: br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED25:%.*]], label [[OMP_OFFLOAD_CONT26:%.*]]
11526 // CHECK14: omp_offload.failed25:
11527 // CHECK14-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136(i64 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
11528 // CHECK14-NEXT: br label [[OMP_OFFLOAD_CONT26]]
11529 // CHECK14: omp_offload.cont26:
11530 // CHECK14-NEXT: ret i32 0
11531 //
11532 //
11533 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116
11534 // CHECK14-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
11535 // CHECK14-NEXT: entry:
11536 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
11537 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
11538 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
11539 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
11540 // CHECK14-NEXT: ret void
11541 //
11542 //
11543 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..14
11544 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
11545 // CHECK14-NEXT: entry:
11546 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11547 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11548 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
11549 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
11550 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
11551 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
11552 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
11553 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11554 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11555 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
11556 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11557 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11558 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
11559 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
11560 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
11561 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
11562 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11563 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11564 // CHECK14-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11565 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
11566 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11567 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11568 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
11569 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11570 // CHECK14: cond.true:
11571 // CHECK14-NEXT: br label [[COND_END:%.*]]
11572 // CHECK14: cond.false:
11573 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11574 // CHECK14-NEXT: br label [[COND_END]]
11575 // CHECK14: cond.end:
11576 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
11577 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
11578 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11579 // CHECK14-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
11580 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
11581 // CHECK14: omp.inner.for.cond:
11582 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11583 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11584 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
11585 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11586 // CHECK14: omp.inner.for.body:
11587 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11588 // CHECK14-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
11589 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11590 // CHECK14-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
11591 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
11592 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
11593 // CHECK14: omp.inner.for.inc:
11594 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11595 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
11596 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
11597 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
11598 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
11599 // CHECK14: omp.inner.for.end:
11600 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
11601 // CHECK14: omp.loop.exit:
11602 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
11603 // CHECK14-NEXT: ret void
11604 //
11605 //
11606 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..15
11607 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
11608 // CHECK14-NEXT: entry:
11609 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11610 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11611 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
11612 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
11613 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
11614 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
11615 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
11616 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
11617 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
11618 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11619 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11620 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
11621 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11622 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11623 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11624 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11625 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
11626 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
11627 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
11628 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
11629 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11630 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
11631 // CHECK14-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11632 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
11633 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
11634 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
11635 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11636 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11637 // CHECK14-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11638 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
11639 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11640 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11641 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
11642 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11643 // CHECK14: cond.true:
11644 // CHECK14-NEXT: br label [[COND_END:%.*]]
11645 // CHECK14: cond.false:
11646 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11647 // CHECK14-NEXT: br label [[COND_END]]
11648 // CHECK14: cond.end:
11649 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
11650 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11651 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11652 // CHECK14-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
11653 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
11654 // CHECK14: omp.inner.for.cond:
11655 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11656 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11657 // CHECK14-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
11658 // CHECK14-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11659 // CHECK14: omp.inner.for.body:
11660 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11661 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
11662 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11663 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4
11664 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
11665 // CHECK14-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
11666 // CHECK14-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
11667 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
11668 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
11669 // CHECK14: omp.body.continue:
11670 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
11671 // CHECK14: omp.inner.for.inc:
11672 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11673 // CHECK14-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
11674 // CHECK14-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
11675 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
11676 // CHECK14: omp.inner.for.end:
11677 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
11678 // CHECK14: omp.loop.exit:
11679 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
11680 // CHECK14-NEXT: ret void
11681 //
11682 //
11683 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121
11684 // CHECK14-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
11685 // CHECK14-NEXT: entry:
11686 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
11687 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
11688 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
11689 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
11690 // CHECK14-NEXT: ret void
11691 //
11692 //
11693 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..17
11694 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
11695 // CHECK14-NEXT: entry:
11696 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11697 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11698 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
11699 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
11700 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
11701 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
11702 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
11703 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11704 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11705 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
11706 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11707 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11708 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
11709 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
11710 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
11711 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
11712 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11713 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11714 // CHECK14-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11715 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
11716 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11717 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11718 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
11719 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11720 // CHECK14: cond.true:
11721 // CHECK14-NEXT: br label [[COND_END:%.*]]
11722 // CHECK14: cond.false:
11723 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11724 // CHECK14-NEXT: br label [[COND_END]]
11725 // CHECK14: cond.end:
11726 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
11727 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
11728 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11729 // CHECK14-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
11730 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
11731 // CHECK14: omp.inner.for.cond:
11732 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11733 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11734 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
11735 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11736 // CHECK14: omp.inner.for.body:
11737 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11738 // CHECK14-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
11739 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11740 // CHECK14-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
11741 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
11742 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
11743 // CHECK14: omp.inner.for.inc:
11744 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11745 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
11746 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
11747 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
11748 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
11749 // CHECK14: omp.inner.for.end:
11750 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
11751 // CHECK14: omp.loop.exit:
11752 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
11753 // CHECK14-NEXT: ret void
11754 //
11755 //
11756 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..18
11757 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
11758 // CHECK14-NEXT: entry:
11759 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11760 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11761 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
11762 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
11763 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
11764 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
11765 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
11766 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
11767 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
11768 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11769 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11770 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
11771 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11772 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11773 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11774 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11775 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
11776 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
11777 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
11778 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
11779 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11780 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
11781 // CHECK14-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11782 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
11783 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
11784 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
11785 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11786 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11787 // CHECK14-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11788 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
11789 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11790 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11791 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
11792 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11793 // CHECK14: cond.true:
11794 // CHECK14-NEXT: br label [[COND_END:%.*]]
11795 // CHECK14: cond.false:
11796 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11797 // CHECK14-NEXT: br label [[COND_END]]
11798 // CHECK14: cond.end:
11799 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
11800 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11801 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11802 // CHECK14-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
11803 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
11804 // CHECK14: omp.inner.for.cond:
11805 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11806 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11807 // CHECK14-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
11808 // CHECK14-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11809 // CHECK14: omp.inner.for.body:
11810 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11811 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
11812 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11813 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4
11814 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
11815 // CHECK14-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
11816 // CHECK14-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
11817 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
11818 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
11819 // CHECK14: omp.body.continue:
11820 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
11821 // CHECK14: omp.inner.for.inc:
11822 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11823 // CHECK14-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
11824 // CHECK14-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
11825 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
11826 // CHECK14: omp.inner.for.end:
11827 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
11828 // CHECK14: omp.loop.exit:
11829 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
11830 // CHECK14-NEXT: ret void
11831 //
11832 //
11833 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126
11834 // CHECK14-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
11835 // CHECK14-NEXT: entry:
11836 // CHECK14-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
11837 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
11838 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
11839 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
11840 // CHECK14-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
11841 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
11842 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
11843 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
11844 // CHECK14-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
11845 // CHECK14-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
11846 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
11847 // CHECK14-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
11848 // CHECK14-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
11849 // CHECK14-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
11850 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
11851 // CHECK14-NEXT: ret void
11852 //
11853 //
11854 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..21
11855 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
11856 // CHECK14-NEXT: entry:
11857 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11858 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11859 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
11860 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
11861 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
11862 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
11863 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
11864 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
11865 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11866 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11867 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
11868 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
11869 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11870 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11871 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
11872 // CHECK14-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
11873 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
11874 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
11875 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
11876 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
11877 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11878 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11879 // CHECK14-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11880 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
11881 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
11882 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11883 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
11884 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11885 // CHECK14: cond.true:
11886 // CHECK14-NEXT: br label [[COND_END:%.*]]
11887 // CHECK14: cond.false:
11888 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11889 // CHECK14-NEXT: br label [[COND_END]]
11890 // CHECK14: cond.end:
11891 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
11892 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
11893 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11894 // CHECK14-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
11895 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
11896 // CHECK14: omp.inner.for.cond:
11897 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11898 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11899 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
11900 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11901 // CHECK14: omp.inner.for.body:
11902 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11903 // CHECK14-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
11904 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11905 // CHECK14-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
11906 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8
11907 // CHECK14-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
11908 // CHECK14-NEXT: store i32 [[TMP12]], i32* [[CONV2]], align 4
11909 // CHECK14-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
11910 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
11911 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
11912 // CHECK14: omp.inner.for.inc:
11913 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11914 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
11915 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
11916 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
11917 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
11918 // CHECK14: omp.inner.for.end:
11919 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
11920 // CHECK14: omp.loop.exit:
11921 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
11922 // CHECK14-NEXT: ret void
11923 //
11924 //
11925 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..22
11926 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
11927 // CHECK14-NEXT: entry:
11928 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11929 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11930 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
11931 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
11932 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
11933 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
11934 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
11935 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
11936 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
11937 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
11938 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11939 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11940 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
11941 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11942 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11943 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11944 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11945 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
11946 // CHECK14-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
11947 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
11948 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
11949 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
11950 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
11951 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11952 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
11953 // CHECK14-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11954 // CHECK14-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
11955 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
11956 // CHECK14-NEXT: store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
11957 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11958 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11959 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
11960 // CHECK14-NEXT: [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11961 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
11962 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
11963 // CHECK14-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
11964 // CHECK14: omp.dispatch.cond:
11965 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11966 // CHECK14-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11967 // CHECK14-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
11968 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
11969 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11970 // CHECK14: cond.true:
11971 // CHECK14-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11972 // CHECK14-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
11973 // CHECK14-NEXT: br label [[COND_END:%.*]]
11974 // CHECK14: cond.false:
11975 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11976 // CHECK14-NEXT: br label [[COND_END]]
11977 // CHECK14: cond.end:
11978 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
11979 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11980 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11981 // CHECK14-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
11982 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11983 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11984 // CHECK14-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
11985 // CHECK14-NEXT: br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
11986 // CHECK14: omp.dispatch.body:
11987 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
11988 // CHECK14: omp.inner.for.cond:
11989 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11990 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11991 // CHECK14-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
11992 // CHECK14-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11993 // CHECK14: omp.inner.for.body:
11994 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11995 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
11996 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11997 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4
11998 // CHECK14-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4
11999 // CHECK14-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
12000 // CHECK14-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
12001 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
12002 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
12003 // CHECK14: omp.body.continue:
12004 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
12005 // CHECK14: omp.inner.for.inc:
12006 // CHECK14-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12007 // CHECK14-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
12008 // CHECK14-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
12009 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
12010 // CHECK14: omp.inner.for.end:
12011 // CHECK14-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
12012 // CHECK14: omp.dispatch.inc:
12013 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12014 // CHECK14-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
12015 // CHECK14-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
12016 // CHECK14-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
12017 // CHECK14-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12018 // CHECK14-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
12019 // CHECK14-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
12020 // CHECK14-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
12021 // CHECK14-NEXT: br label [[OMP_DISPATCH_COND]]
12022 // CHECK14: omp.dispatch.end:
12023 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
12024 // CHECK14-NEXT: ret void
12025 //
12026 //
12027 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
12028 // CHECK14-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
12029 // CHECK14-NEXT: entry:
12030 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
12031 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
12032 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
12033 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
12034 // CHECK14-NEXT: ret void
12035 //
12036 //
12037 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..25
12038 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
12039 // CHECK14-NEXT: entry:
12040 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12041 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12042 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
12043 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
12044 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
12045 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
12046 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
12047 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12048 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12049 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
12050 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12051 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12052 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
12053 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
12054 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
12055 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
12056 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12057 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12058 // CHECK14-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12059 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
12060 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
12061 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12062 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
12063 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12064 // CHECK14: cond.true:
12065 // CHECK14-NEXT: br label [[COND_END:%.*]]
12066 // CHECK14: cond.false:
12067 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12068 // CHECK14-NEXT: br label [[COND_END]]
12069 // CHECK14: cond.end:
12070 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
12071 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
12072 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
12073 // CHECK14-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
12074 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
12075 // CHECK14: omp.inner.for.cond:
12076 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12077 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12078 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
12079 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12080 // CHECK14: omp.inner.for.body:
12081 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
12082 // CHECK14-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
12083 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12084 // CHECK14-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
12085 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
12086 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
12087 // CHECK14: omp.inner.for.inc:
12088 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12089 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
12090 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
12091 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
12092 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
12093 // CHECK14: omp.inner.for.end:
12094 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
12095 // CHECK14: omp.loop.exit:
12096 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
12097 // CHECK14-NEXT: ret void
12098 //
12099 //
12100 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..26
12101 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
12102 // CHECK14-NEXT: entry:
12103 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12104 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12105 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
12106 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
12107 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
12108 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
12109 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
12110 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
12111 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
12112 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12113 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12114 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
12115 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12116 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12117 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12118 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12119 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
12120 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
12121 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
12122 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
12123 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12124 // CHECK14-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
12125 // CHECK14-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12126 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
12127 // CHECK14-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
12128 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
12129 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12130 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12131 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12132 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12133 // CHECK14-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12134 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
12135 // CHECK14-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
12136 // CHECK14-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
12137 // CHECK14: omp.dispatch.cond:
12138 // CHECK14-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
12139 // CHECK14-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
12140 // CHECK14-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
12141 // CHECK14: omp.dispatch.body:
12142 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12143 // CHECK14-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
12144 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
12145 // CHECK14: omp.inner.for.cond:
12146 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
12147 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
12148 // CHECK14-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
12149 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12150 // CHECK14: omp.inner.for.body:
12151 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
12152 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
12153 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12154 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !21
12155 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !21
12156 // CHECK14-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
12157 // CHECK14-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
12158 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
12159 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
12160 // CHECK14: omp.body.continue:
12161 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
12162 // CHECK14: omp.inner.for.inc:
12163 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
12164 // CHECK14-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
12165 // CHECK14-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
12166 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
12167 // CHECK14: omp.inner.for.end:
12168 // CHECK14-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
12169 // CHECK14: omp.dispatch.inc:
12170 // CHECK14-NEXT: br label [[OMP_DISPATCH_COND]]
12171 // CHECK14: omp.dispatch.end:
12172 // CHECK14-NEXT: ret void
12173 //
12174 //
12175 // CHECK14-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136
12176 // CHECK14-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
12177 // CHECK14-NEXT: entry:
12178 // CHECK14-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
12179 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
12180 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
12181 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
12182 // CHECK14-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
12183 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
12184 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
12185 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
12186 // CHECK14-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
12187 // CHECK14-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
12188 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
12189 // CHECK14-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
12190 // CHECK14-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
12191 // CHECK14-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
12192 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
12193 // CHECK14-NEXT: ret void
12194 //
12195 //
12196 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..29
12197 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
12198 // CHECK14-NEXT: entry:
12199 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12200 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12201 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
12202 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
12203 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
12204 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
12205 // CHECK14-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
12206 // CHECK14-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
12207 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12208 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12209 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
12210 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
12211 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12212 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12213 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
12214 // CHECK14-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
12215 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
12216 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
12217 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
12218 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
12219 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12220 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12221 // CHECK14-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12222 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
12223 // CHECK14-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
12224 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12225 // CHECK14-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
12226 // CHECK14-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12227 // CHECK14: cond.true:
12228 // CHECK14-NEXT: br label [[COND_END:%.*]]
12229 // CHECK14: cond.false:
12230 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12231 // CHECK14-NEXT: br label [[COND_END]]
12232 // CHECK14: cond.end:
12233 // CHECK14-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
12234 // CHECK14-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
12235 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
12236 // CHECK14-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
12237 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
12238 // CHECK14: omp.inner.for.cond:
12239 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12240 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12241 // CHECK14-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
12242 // CHECK14-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12243 // CHECK14: omp.inner.for.body:
12244 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
12245 // CHECK14-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
12246 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12247 // CHECK14-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
12248 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8
12249 // CHECK14-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
12250 // CHECK14-NEXT: store i32 [[TMP12]], i32* [[CONV2]], align 4
12251 // CHECK14-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
12252 // CHECK14-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
12253 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
12254 // CHECK14: omp.inner.for.inc:
12255 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12256 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
12257 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
12258 // CHECK14-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
12259 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]]
12260 // CHECK14: omp.inner.for.end:
12261 // CHECK14-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
12262 // CHECK14: omp.loop.exit:
12263 // CHECK14-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
12264 // CHECK14-NEXT: ret void
12265 //
12266 //
12267 // CHECK14-LABEL: define {{[^@]+}}@.omp_outlined..30
12268 // CHECK14-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
12269 // CHECK14-NEXT: entry:
12270 // CHECK14-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12271 // CHECK14-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12272 // CHECK14-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
12273 // CHECK14-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
12274 // CHECK14-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
12275 // CHECK14-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
12276 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
12277 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
12278 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
12279 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
12280 // CHECK14-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12281 // CHECK14-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12282 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
12283 // CHECK14-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12284 // CHECK14-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12285 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12286 // CHECK14-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12287 // CHECK14-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
12288 // CHECK14-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
12289 // CHECK14-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
12290 // CHECK14-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
12291 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
12292 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
12293 // CHECK14-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12294 // CHECK14-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
12295 // CHECK14-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12296 // CHECK14-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
12297 // CHECK14-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
12298 // CHECK14-NEXT: store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
12299 // CHECK14-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12300 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12301 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
12302 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12303 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12304 // CHECK14-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12305 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
12306 // CHECK14-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 1073741859, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
12307 // CHECK14-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
12308 // CHECK14: omp.dispatch.cond:
12309 // CHECK14-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
12310 // CHECK14-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
12311 // CHECK14-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
12312 // CHECK14: omp.dispatch.body:
12313 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12314 // CHECK14-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
12315 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
12316 // CHECK14: omp.inner.for.cond:
12317 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
12318 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !24
12319 // CHECK14-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
12320 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12321 // CHECK14: omp.inner.for.body:
12322 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
12323 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
12324 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12325 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !24
12326 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !24
12327 // CHECK14-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
12328 // CHECK14-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
12329 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !24
12330 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
12331 // CHECK14: omp.body.continue:
12332 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
12333 // CHECK14: omp.inner.for.inc:
12334 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
12335 // CHECK14-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
12336 // CHECK14-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
12337 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
12338 // CHECK14: omp.inner.for.end:
12339 // CHECK14-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
12340 // CHECK14: omp.dispatch.inc:
12341 // CHECK14-NEXT: br label [[OMP_DISPATCH_COND]]
12342 // CHECK14: omp.dispatch.end:
12343 // CHECK14-NEXT: ret void
12344 //
12345 //
12346 // CHECK14-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
12347 // CHECK14-SAME: () #[[ATTR5:[0-9]+]] {
12348 // CHECK14-NEXT: entry:
12349 // CHECK14-NEXT: call void @__tgt_register_requires(i64 1)
12350 // CHECK14-NEXT: ret void
12351 //
12352 //
12353 // CHECK15-LABEL: define {{[^@]+}}@main
12354 // CHECK15-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
12355 // CHECK15-NEXT: entry:
12356 // CHECK15-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
12357 // CHECK15-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
12358 // CHECK15-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
12359 // CHECK15-NEXT: [[N:%.*]] = alloca i32, align 4
12360 // CHECK15-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
12361 // CHECK15-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
12362 // CHECK15-NEXT: [[M:%.*]] = alloca i32, align 4
12363 // CHECK15-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
12364 // CHECK15-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
12365 // CHECK15-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
12366 // CHECK15-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
12367 // CHECK15-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
12368 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
12369 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
12370 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
12371 // CHECK15-NEXT: [[N_CASTED3:%.*]] = alloca i32, align 4
12372 // CHECK15-NEXT: [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
12373 // CHECK15-NEXT: [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
12374 // CHECK15-NEXT: [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
12375 // CHECK15-NEXT: [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
12376 // CHECK15-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
12377 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
12378 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
12379 // CHECK15-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
12380 // CHECK15-NEXT: [[N_CASTED17:%.*]] = alloca i32, align 4
12381 // CHECK15-NEXT: [[DOTOFFLOAD_BASEPTRS18:%.*]] = alloca [4 x i8*], align 4
12382 // CHECK15-NEXT: [[DOTOFFLOAD_PTRS19:%.*]] = alloca [4 x i8*], align 4
12383 // CHECK15-NEXT: [[DOTOFFLOAD_MAPPERS20:%.*]] = alloca [4 x i8*], align 4
12384 // CHECK15-NEXT: [[DOTOFFLOAD_SIZES21:%.*]] = alloca [4 x i64], align 4
12385 // CHECK15-NEXT: [[_TMP22:%.*]] = alloca i32, align 4
12386 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_23:%.*]] = alloca i32, align 4
12387 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
12388 // CHECK15-NEXT: [[N_CASTED31:%.*]] = alloca i32, align 4
12389 // CHECK15-NEXT: [[DOTOFFLOAD_BASEPTRS32:%.*]] = alloca [3 x i8*], align 4
12390 // CHECK15-NEXT: [[DOTOFFLOAD_PTRS33:%.*]] = alloca [3 x i8*], align 4
12391 // CHECK15-NEXT: [[DOTOFFLOAD_MAPPERS34:%.*]] = alloca [3 x i8*], align 4
12392 // CHECK15-NEXT: [[DOTOFFLOAD_SIZES35:%.*]] = alloca [3 x i64], align 4
12393 // CHECK15-NEXT: [[_TMP36:%.*]] = alloca i32, align 4
12394 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_37:%.*]] = alloca i32, align 4
12395 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
12396 // CHECK15-NEXT: [[M_CASTED45:%.*]] = alloca i32, align 4
12397 // CHECK15-NEXT: [[N_CASTED46:%.*]] = alloca i32, align 4
12398 // CHECK15-NEXT: [[DOTOFFLOAD_BASEPTRS47:%.*]] = alloca [4 x i8*], align 4
12399 // CHECK15-NEXT: [[DOTOFFLOAD_PTRS48:%.*]] = alloca [4 x i8*], align 4
12400 // CHECK15-NEXT: [[DOTOFFLOAD_MAPPERS49:%.*]] = alloca [4 x i8*], align 4
12401 // CHECK15-NEXT: [[DOTOFFLOAD_SIZES50:%.*]] = alloca [4 x i64], align 4
12402 // CHECK15-NEXT: [[_TMP51:%.*]] = alloca i32, align 4
12403 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_52:%.*]] = alloca i32, align 4
12404 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_53:%.*]] = alloca i32, align 4
12405 // CHECK15-NEXT: store i32 0, i32* [[RETVAL]], align 4
12406 // CHECK15-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
12407 // CHECK15-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
12408 // CHECK15-NEXT: store i32 100, i32* [[N]], align 4
12409 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
12410 // CHECK15-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
12411 // CHECK15-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
12412 // CHECK15-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
12413 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
12414 // CHECK15-NEXT: store i32 10, i32* [[M]], align 4
12415 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
12416 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
12417 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
12418 // CHECK15-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
12419 // CHECK15-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
12420 // CHECK15-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12421 // CHECK15-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
12422 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[TMP7]], align 4
12423 // CHECK15-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12424 // CHECK15-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
12425 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[TMP9]], align 4
12426 // CHECK15-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
12427 // CHECK15-NEXT: store i64 4, i64* [[TMP10]], align 4
12428 // CHECK15-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
12429 // CHECK15-NEXT: store i8* null, i8** [[TMP11]], align 4
12430 // CHECK15-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
12431 // CHECK15-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
12432 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
12433 // CHECK15-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
12434 // CHECK15-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
12435 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
12436 // CHECK15-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
12437 // CHECK15-NEXT: store i64 4, i64* [[TMP16]], align 4
12438 // CHECK15-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
12439 // CHECK15-NEXT: store i8* null, i8** [[TMP17]], align 4
12440 // CHECK15-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
12441 // CHECK15-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
12442 // CHECK15-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 4
12443 // CHECK15-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
12444 // CHECK15-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
12445 // CHECK15-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 4
12446 // CHECK15-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
12447 // CHECK15-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 4
12448 // CHECK15-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
12449 // CHECK15-NEXT: store i8* null, i8** [[TMP23]], align 4
12450 // CHECK15-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12451 // CHECK15-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12452 // CHECK15-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
12453 // CHECK15-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
12454 // CHECK15-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
12455 // CHECK15-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
12456 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
12457 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
12458 // CHECK15-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
12459 // CHECK15-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
12460 // CHECK15-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
12461 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
12462 // CHECK15-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
12463 // CHECK15-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
12464 // CHECK15-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12465 // CHECK15-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
12466 // CHECK15-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
12467 // CHECK15: omp_offload.failed:
12468 // CHECK15-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
12469 // CHECK15-NEXT: br label [[OMP_OFFLOAD_CONT]]
12470 // CHECK15: omp_offload.cont:
12471 // CHECK15-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
12472 // CHECK15-NEXT: store i32 [[TMP33]], i32* [[N_CASTED3]], align 4
12473 // CHECK15-NEXT: [[TMP34:%.*]] = load i32, i32* [[N_CASTED3]], align 4
12474 // CHECK15-NEXT: [[TMP35:%.*]] = mul nuw i32 [[TMP0]], 4
12475 // CHECK15-NEXT: [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
12476 // CHECK15-NEXT: [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
12477 // CHECK15-NEXT: [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
12478 // CHECK15-NEXT: store i32 [[TMP34]], i32* [[TMP38]], align 4
12479 // CHECK15-NEXT: [[TMP39:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
12480 // CHECK15-NEXT: [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
12481 // CHECK15-NEXT: store i32 [[TMP34]], i32* [[TMP40]], align 4
12482 // CHECK15-NEXT: [[TMP41:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
12483 // CHECK15-NEXT: store i64 4, i64* [[TMP41]], align 4
12484 // CHECK15-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
12485 // CHECK15-NEXT: store i8* null, i8** [[TMP42]], align 4
12486 // CHECK15-NEXT: [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
12487 // CHECK15-NEXT: [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i32*
12488 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[TMP44]], align 4
12489 // CHECK15-NEXT: [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
12490 // CHECK15-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
12491 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[TMP46]], align 4
12492 // CHECK15-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 1
12493 // CHECK15-NEXT: store i64 4, i64* [[TMP47]], align 4
12494 // CHECK15-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
12495 // CHECK15-NEXT: store i8* null, i8** [[TMP48]], align 4
12496 // CHECK15-NEXT: [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
12497 // CHECK15-NEXT: [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32**
12498 // CHECK15-NEXT: store i32* [[VLA]], i32** [[TMP50]], align 4
12499 // CHECK15-NEXT: [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
12500 // CHECK15-NEXT: [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32**
12501 // CHECK15-NEXT: store i32* [[VLA]], i32** [[TMP52]], align 4
12502 // CHECK15-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
12503 // CHECK15-NEXT: store i64 [[TMP36]], i64* [[TMP53]], align 4
12504 // CHECK15-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
12505 // CHECK15-NEXT: store i8* null, i8** [[TMP54]], align 4
12506 // CHECK15-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
12507 // CHECK15-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
12508 // CHECK15-NEXT: [[TMP57:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
12509 // CHECK15-NEXT: [[TMP58:%.*]] = load i32, i32* [[N]], align 4
12510 // CHECK15-NEXT: store i32 [[TMP58]], i32* [[DOTCAPTURE_EXPR_9]], align 4
12511 // CHECK15-NEXT: [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
12512 // CHECK15-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP59]], 0
12513 // CHECK15-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
12514 // CHECK15-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
12515 // CHECK15-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
12516 // CHECK15-NEXT: [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
12517 // CHECK15-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP60]], 1
12518 // CHECK15-NEXT: [[TMP61:%.*]] = zext i32 [[ADD14]] to i64
12519 // CHECK15-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP61]])
12520 // CHECK15-NEXT: [[TMP62:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153.region_id, i32 3, i8** [[TMP55]], i8** [[TMP56]], i64* [[TMP57]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12521 // CHECK15-NEXT: [[TMP63:%.*]] = icmp ne i32 [[TMP62]], 0
12522 // CHECK15-NEXT: br i1 [[TMP63]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
12523 // CHECK15: omp_offload.failed15:
12524 // CHECK15-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153(i32 [[TMP34]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
12525 // CHECK15-NEXT: br label [[OMP_OFFLOAD_CONT16]]
12526 // CHECK15: omp_offload.cont16:
12527 // CHECK15-NEXT: [[TMP64:%.*]] = load i32, i32* [[M]], align 4
12528 // CHECK15-NEXT: store i32 [[TMP64]], i32* [[M_CASTED]], align 4
12529 // CHECK15-NEXT: [[TMP65:%.*]] = load i32, i32* [[M_CASTED]], align 4
12530 // CHECK15-NEXT: [[TMP66:%.*]] = load i32, i32* [[N]], align 4
12531 // CHECK15-NEXT: store i32 [[TMP66]], i32* [[N_CASTED17]], align 4
12532 // CHECK15-NEXT: [[TMP67:%.*]] = load i32, i32* [[N_CASTED17]], align 4
12533 // CHECK15-NEXT: [[TMP68:%.*]] = mul nuw i32 [[TMP0]], 4
12534 // CHECK15-NEXT: [[TMP69:%.*]] = sext i32 [[TMP68]] to i64
12535 // CHECK15-NEXT: [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
12536 // CHECK15-NEXT: [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i32*
12537 // CHECK15-NEXT: store i32 [[TMP65]], i32* [[TMP71]], align 4
12538 // CHECK15-NEXT: [[TMP72:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
12539 // CHECK15-NEXT: [[TMP73:%.*]] = bitcast i8** [[TMP72]] to i32*
12540 // CHECK15-NEXT: store i32 [[TMP65]], i32* [[TMP73]], align 4
12541 // CHECK15-NEXT: [[TMP74:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
12542 // CHECK15-NEXT: store i64 4, i64* [[TMP74]], align 4
12543 // CHECK15-NEXT: [[TMP75:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 0
12544 // CHECK15-NEXT: store i8* null, i8** [[TMP75]], align 4
12545 // CHECK15-NEXT: [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 1
12546 // CHECK15-NEXT: [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i32*
12547 // CHECK15-NEXT: store i32 [[TMP67]], i32* [[TMP77]], align 4
12548 // CHECK15-NEXT: [[TMP78:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 1
12549 // CHECK15-NEXT: [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
12550 // CHECK15-NEXT: store i32 [[TMP67]], i32* [[TMP79]], align 4
12551 // CHECK15-NEXT: [[TMP80:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 1
12552 // CHECK15-NEXT: store i64 4, i64* [[TMP80]], align 4
12553 // CHECK15-NEXT: [[TMP81:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 1
12554 // CHECK15-NEXT: store i8* null, i8** [[TMP81]], align 4
12555 // CHECK15-NEXT: [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 2
12556 // CHECK15-NEXT: [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i32*
12557 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[TMP83]], align 4
12558 // CHECK15-NEXT: [[TMP84:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 2
12559 // CHECK15-NEXT: [[TMP85:%.*]] = bitcast i8** [[TMP84]] to i32*
12560 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[TMP85]], align 4
12561 // CHECK15-NEXT: [[TMP86:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 2
12562 // CHECK15-NEXT: store i64 4, i64* [[TMP86]], align 4
12563 // CHECK15-NEXT: [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 2
12564 // CHECK15-NEXT: store i8* null, i8** [[TMP87]], align 4
12565 // CHECK15-NEXT: [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 3
12566 // CHECK15-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
12567 // CHECK15-NEXT: store i32* [[VLA]], i32** [[TMP89]], align 4
12568 // CHECK15-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 3
12569 // CHECK15-NEXT: [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i32**
12570 // CHECK15-NEXT: store i32* [[VLA]], i32** [[TMP91]], align 4
12571 // CHECK15-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 3
12572 // CHECK15-NEXT: store i64 [[TMP69]], i64* [[TMP92]], align 4
12573 // CHECK15-NEXT: [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 3
12574 // CHECK15-NEXT: store i8* null, i8** [[TMP93]], align 4
12575 // CHECK15-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
12576 // CHECK15-NEXT: [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
12577 // CHECK15-NEXT: [[TMP96:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
12578 // CHECK15-NEXT: [[TMP97:%.*]] = load i32, i32* [[N]], align 4
12579 // CHECK15-NEXT: store i32 [[TMP97]], i32* [[DOTCAPTURE_EXPR_23]], align 4
12580 // CHECK15-NEXT: [[TMP98:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_23]], align 4
12581 // CHECK15-NEXT: [[SUB25:%.*]] = sub nsw i32 [[TMP98]], 0
12582 // CHECK15-NEXT: [[DIV26:%.*]] = sdiv i32 [[SUB25]], 1
12583 // CHECK15-NEXT: [[SUB27:%.*]] = sub nsw i32 [[DIV26]], 1
12584 // CHECK15-NEXT: store i32 [[SUB27]], i32* [[DOTCAPTURE_EXPR_24]], align 4
12585 // CHECK15-NEXT: [[TMP99:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
12586 // CHECK15-NEXT: [[ADD28:%.*]] = add nsw i32 [[TMP99]], 1
12587 // CHECK15-NEXT: [[TMP100:%.*]] = zext i32 [[ADD28]] to i64
12588 // CHECK15-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP100]])
12589 // CHECK15-NEXT: [[TMP101:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158.region_id, i32 4, i8** [[TMP94]], i8** [[TMP95]], i64* [[TMP96]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12590 // CHECK15-NEXT: [[TMP102:%.*]] = icmp ne i32 [[TMP101]], 0
12591 // CHECK15-NEXT: br i1 [[TMP102]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
12592 // CHECK15: omp_offload.failed29:
12593 // CHECK15-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158(i32 [[TMP65]], i32 [[TMP67]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
12594 // CHECK15-NEXT: br label [[OMP_OFFLOAD_CONT30]]
12595 // CHECK15: omp_offload.cont30:
12596 // CHECK15-NEXT: [[TMP103:%.*]] = load i32, i32* [[N]], align 4
12597 // CHECK15-NEXT: store i32 [[TMP103]], i32* [[N_CASTED31]], align 4
12598 // CHECK15-NEXT: [[TMP104:%.*]] = load i32, i32* [[N_CASTED31]], align 4
12599 // CHECK15-NEXT: [[TMP105:%.*]] = mul nuw i32 [[TMP0]], 4
12600 // CHECK15-NEXT: [[TMP106:%.*]] = sext i32 [[TMP105]] to i64
12601 // CHECK15-NEXT: [[TMP107:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
12602 // CHECK15-NEXT: [[TMP108:%.*]] = bitcast i8** [[TMP107]] to i32*
12603 // CHECK15-NEXT: store i32 [[TMP104]], i32* [[TMP108]], align 4
12604 // CHECK15-NEXT: [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
12605 // CHECK15-NEXT: [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i32*
12606 // CHECK15-NEXT: store i32 [[TMP104]], i32* [[TMP110]], align 4
12607 // CHECK15-NEXT: [[TMP111:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
12608 // CHECK15-NEXT: store i64 4, i64* [[TMP111]], align 4
12609 // CHECK15-NEXT: [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 0
12610 // CHECK15-NEXT: store i8* null, i8** [[TMP112]], align 4
12611 // CHECK15-NEXT: [[TMP113:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 1
12612 // CHECK15-NEXT: [[TMP114:%.*]] = bitcast i8** [[TMP113]] to i32*
12613 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[TMP114]], align 4
12614 // CHECK15-NEXT: [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 1
12615 // CHECK15-NEXT: [[TMP116:%.*]] = bitcast i8** [[TMP115]] to i32*
12616 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[TMP116]], align 4
12617 // CHECK15-NEXT: [[TMP117:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 1
12618 // CHECK15-NEXT: store i64 4, i64* [[TMP117]], align 4
12619 // CHECK15-NEXT: [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 1
12620 // CHECK15-NEXT: store i8* null, i8** [[TMP118]], align 4
12621 // CHECK15-NEXT: [[TMP119:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 2
12622 // CHECK15-NEXT: [[TMP120:%.*]] = bitcast i8** [[TMP119]] to i32**
12623 // CHECK15-NEXT: store i32* [[VLA]], i32** [[TMP120]], align 4
12624 // CHECK15-NEXT: [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 2
12625 // CHECK15-NEXT: [[TMP122:%.*]] = bitcast i8** [[TMP121]] to i32**
12626 // CHECK15-NEXT: store i32* [[VLA]], i32** [[TMP122]], align 4
12627 // CHECK15-NEXT: [[TMP123:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 2
12628 // CHECK15-NEXT: store i64 [[TMP106]], i64* [[TMP123]], align 4
12629 // CHECK15-NEXT: [[TMP124:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 2
12630 // CHECK15-NEXT: store i8* null, i8** [[TMP124]], align 4
12631 // CHECK15-NEXT: [[TMP125:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
12632 // CHECK15-NEXT: [[TMP126:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
12633 // CHECK15-NEXT: [[TMP127:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
12634 // CHECK15-NEXT: [[TMP128:%.*]] = load i32, i32* [[N]], align 4
12635 // CHECK15-NEXT: store i32 [[TMP128]], i32* [[DOTCAPTURE_EXPR_37]], align 4
12636 // CHECK15-NEXT: [[TMP129:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_37]], align 4
12637 // CHECK15-NEXT: [[SUB39:%.*]] = sub nsw i32 [[TMP129]], 0
12638 // CHECK15-NEXT: [[DIV40:%.*]] = sdiv i32 [[SUB39]], 1
12639 // CHECK15-NEXT: [[SUB41:%.*]] = sub nsw i32 [[DIV40]], 1
12640 // CHECK15-NEXT: store i32 [[SUB41]], i32* [[DOTCAPTURE_EXPR_38]], align 4
12641 // CHECK15-NEXT: [[TMP130:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_38]], align 4
12642 // CHECK15-NEXT: [[ADD42:%.*]] = add nsw i32 [[TMP130]], 1
12643 // CHECK15-NEXT: [[TMP131:%.*]] = zext i32 [[ADD42]] to i64
12644 // CHECK15-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP131]])
12645 // CHECK15-NEXT: [[TMP132:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163.region_id, i32 3, i8** [[TMP125]], i8** [[TMP126]], i64* [[TMP127]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12646 // CHECK15-NEXT: [[TMP133:%.*]] = icmp ne i32 [[TMP132]], 0
12647 // CHECK15-NEXT: br i1 [[TMP133]], label [[OMP_OFFLOAD_FAILED43:%.*]], label [[OMP_OFFLOAD_CONT44:%.*]]
12648 // CHECK15: omp_offload.failed43:
12649 // CHECK15-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163(i32 [[TMP104]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
12650 // CHECK15-NEXT: br label [[OMP_OFFLOAD_CONT44]]
12651 // CHECK15: omp_offload.cont44:
12652 // CHECK15-NEXT: [[TMP134:%.*]] = load i32, i32* [[M]], align 4
12653 // CHECK15-NEXT: store i32 [[TMP134]], i32* [[M_CASTED45]], align 4
12654 // CHECK15-NEXT: [[TMP135:%.*]] = load i32, i32* [[M_CASTED45]], align 4
12655 // CHECK15-NEXT: [[TMP136:%.*]] = load i32, i32* [[N]], align 4
12656 // CHECK15-NEXT: store i32 [[TMP136]], i32* [[N_CASTED46]], align 4
12657 // CHECK15-NEXT: [[TMP137:%.*]] = load i32, i32* [[N_CASTED46]], align 4
12658 // CHECK15-NEXT: [[TMP138:%.*]] = mul nuw i32 [[TMP0]], 4
12659 // CHECK15-NEXT: [[TMP139:%.*]] = sext i32 [[TMP138]] to i64
12660 // CHECK15-NEXT: [[TMP140:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
12661 // CHECK15-NEXT: [[TMP141:%.*]] = bitcast i8** [[TMP140]] to i32*
12662 // CHECK15-NEXT: store i32 [[TMP135]], i32* [[TMP141]], align 4
12663 // CHECK15-NEXT: [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
12664 // CHECK15-NEXT: [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i32*
12665 // CHECK15-NEXT: store i32 [[TMP135]], i32* [[TMP143]], align 4
12666 // CHECK15-NEXT: [[TMP144:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
12667 // CHECK15-NEXT: store i64 4, i64* [[TMP144]], align 4
12668 // CHECK15-NEXT: [[TMP145:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 0
12669 // CHECK15-NEXT: store i8* null, i8** [[TMP145]], align 4
12670 // CHECK15-NEXT: [[TMP146:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 1
12671 // CHECK15-NEXT: [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i32*
12672 // CHECK15-NEXT: store i32 [[TMP137]], i32* [[TMP147]], align 4
12673 // CHECK15-NEXT: [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 1
12674 // CHECK15-NEXT: [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
12675 // CHECK15-NEXT: store i32 [[TMP137]], i32* [[TMP149]], align 4
12676 // CHECK15-NEXT: [[TMP150:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 1
12677 // CHECK15-NEXT: store i64 4, i64* [[TMP150]], align 4
12678 // CHECK15-NEXT: [[TMP151:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 1
12679 // CHECK15-NEXT: store i8* null, i8** [[TMP151]], align 4
12680 // CHECK15-NEXT: [[TMP152:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 2
12681 // CHECK15-NEXT: [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i32*
12682 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[TMP153]], align 4
12683 // CHECK15-NEXT: [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 2
12684 // CHECK15-NEXT: [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32*
12685 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[TMP155]], align 4
12686 // CHECK15-NEXT: [[TMP156:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 2
12687 // CHECK15-NEXT: store i64 4, i64* [[TMP156]], align 4
12688 // CHECK15-NEXT: [[TMP157:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 2
12689 // CHECK15-NEXT: store i8* null, i8** [[TMP157]], align 4
12690 // CHECK15-NEXT: [[TMP158:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 3
12691 // CHECK15-NEXT: [[TMP159:%.*]] = bitcast i8** [[TMP158]] to i32**
12692 // CHECK15-NEXT: store i32* [[VLA]], i32** [[TMP159]], align 4
12693 // CHECK15-NEXT: [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 3
12694 // CHECK15-NEXT: [[TMP161:%.*]] = bitcast i8** [[TMP160]] to i32**
12695 // CHECK15-NEXT: store i32* [[VLA]], i32** [[TMP161]], align 4
12696 // CHECK15-NEXT: [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 3
12697 // CHECK15-NEXT: store i64 [[TMP139]], i64* [[TMP162]], align 4
12698 // CHECK15-NEXT: [[TMP163:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 3
12699 // CHECK15-NEXT: store i8* null, i8** [[TMP163]], align 4
12700 // CHECK15-NEXT: [[TMP164:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
12701 // CHECK15-NEXT: [[TMP165:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
12702 // CHECK15-NEXT: [[TMP166:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
12703 // CHECK15-NEXT: [[TMP167:%.*]] = load i32, i32* [[N]], align 4
12704 // CHECK15-NEXT: store i32 [[TMP167]], i32* [[DOTCAPTURE_EXPR_52]], align 4
12705 // CHECK15-NEXT: [[TMP168:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_52]], align 4
12706 // CHECK15-NEXT: [[SUB54:%.*]] = sub nsw i32 [[TMP168]], 0
12707 // CHECK15-NEXT: [[DIV55:%.*]] = sdiv i32 [[SUB54]], 1
12708 // CHECK15-NEXT: [[SUB56:%.*]] = sub nsw i32 [[DIV55]], 1
12709 // CHECK15-NEXT: store i32 [[SUB56]], i32* [[DOTCAPTURE_EXPR_53]], align 4
12710 // CHECK15-NEXT: [[TMP169:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_53]], align 4
12711 // CHECK15-NEXT: [[ADD57:%.*]] = add nsw i32 [[TMP169]], 1
12712 // CHECK15-NEXT: [[TMP170:%.*]] = zext i32 [[ADD57]] to i64
12713 // CHECK15-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP170]])
12714 // CHECK15-NEXT: [[TMP171:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168.region_id, i32 4, i8** [[TMP164]], i8** [[TMP165]], i64* [[TMP166]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
12715 // CHECK15-NEXT: [[TMP172:%.*]] = icmp ne i32 [[TMP171]], 0
12716 // CHECK15-NEXT: br i1 [[TMP172]], label [[OMP_OFFLOAD_FAILED58:%.*]], label [[OMP_OFFLOAD_CONT59:%.*]]
12717 // CHECK15: omp_offload.failed58:
12718 // CHECK15-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168(i32 [[TMP135]], i32 [[TMP137]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
12719 // CHECK15-NEXT: br label [[OMP_OFFLOAD_CONT59]]
12720 // CHECK15: omp_offload.cont59:
12721 // CHECK15-NEXT: [[TMP173:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
12722 // CHECK15-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP173]])
12723 // CHECK15-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
12724 // CHECK15-NEXT: [[TMP174:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
12725 // CHECK15-NEXT: call void @llvm.stackrestore(i8* [[TMP174]])
12726 // CHECK15-NEXT: [[TMP175:%.*]] = load i32, i32* [[RETVAL]], align 4
12727 // CHECK15-NEXT: ret i32 [[TMP175]]
12728 //
12729 //
12730 // CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148
12731 // CHECK15-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
12732 // CHECK15-NEXT: entry:
12733 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
12734 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
12735 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
12736 // CHECK15-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
12737 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
12738 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
12739 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
12740 // CHECK15-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
12741 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
12742 // CHECK15-NEXT: ret void
12743 //
12744 //
12745 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined.
12746 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
12747 // CHECK15-NEXT: entry:
12748 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12749 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12750 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
12751 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
12752 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
12753 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
12754 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
12755 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
12756 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
12757 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
12758 // CHECK15-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
12759 // CHECK15-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
12760 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12761 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12762 // CHECK15-NEXT: [[I3:%.*]] = alloca i32, align 4
12763 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12764 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12765 // CHECK15-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
12766 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
12767 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
12768 // CHECK15-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
12769 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
12770 // CHECK15-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
12771 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
12772 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
12773 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
12774 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
12775 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
12776 // CHECK15-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
12777 // CHECK15-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
12778 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
12779 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
12780 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
12781 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
12782 // CHECK15: omp.precond.then:
12783 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
12784 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
12785 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
12786 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12787 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12788 // CHECK15-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
12789 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
12790 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
12791 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12792 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
12793 // CHECK15-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
12794 // CHECK15-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12795 // CHECK15: cond.true:
12796 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
12797 // CHECK15-NEXT: br label [[COND_END:%.*]]
12798 // CHECK15: cond.false:
12799 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12800 // CHECK15-NEXT: br label [[COND_END]]
12801 // CHECK15: cond.end:
12802 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
12803 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
12804 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
12805 // CHECK15-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
12806 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
12807 // CHECK15: omp.inner.for.cond:
12808 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12809 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12810 // CHECK15-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
12811 // CHECK15-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12812 // CHECK15: omp.inner.for.body:
12813 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
12814 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12815 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
12816 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
12817 // CHECK15: omp.inner.for.inc:
12818 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12819 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
12820 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
12821 // CHECK15-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
12822 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
12823 // CHECK15: omp.inner.for.end:
12824 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
12825 // CHECK15: omp.loop.exit:
12826 // CHECK15-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
12827 // CHECK15-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
12828 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
12829 // CHECK15-NEXT: br label [[OMP_PRECOND_END]]
12830 // CHECK15: omp.precond.end:
12831 // CHECK15-NEXT: ret void
12832 //
12833 //
12834 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..1
12835 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
12836 // CHECK15-NEXT: entry:
12837 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12838 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12839 // CHECK15-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
12840 // CHECK15-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
12841 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
12842 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
12843 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
12844 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
12845 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
12846 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
12847 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
12848 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
12849 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
12850 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
12851 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12852 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12853 // CHECK15-NEXT: [[I3:%.*]] = alloca i32, align 4
12854 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12855 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12856 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
12857 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
12858 // CHECK15-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
12859 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
12860 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
12861 // CHECK15-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
12862 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
12863 // CHECK15-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
12864 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
12865 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
12866 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
12867 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
12868 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
12869 // CHECK15-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
12870 // CHECK15-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
12871 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
12872 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
12873 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
12874 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
12875 // CHECK15: omp.precond.then:
12876 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
12877 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
12878 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
12879 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
12880 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
12881 // CHECK15-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
12882 // CHECK15-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
12883 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12884 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12885 // CHECK15-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
12886 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
12887 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
12888 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12889 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
12890 // CHECK15-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
12891 // CHECK15-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12892 // CHECK15: cond.true:
12893 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
12894 // CHECK15-NEXT: br label [[COND_END:%.*]]
12895 // CHECK15: cond.false:
12896 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12897 // CHECK15-NEXT: br label [[COND_END]]
12898 // CHECK15: cond.end:
12899 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
12900 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
12901 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12902 // CHECK15-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
12903 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
12904 // CHECK15: omp.inner.for.cond:
12905 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12906 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12907 // CHECK15-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
12908 // CHECK15-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12909 // CHECK15: omp.inner.for.body:
12910 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12911 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
12912 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12913 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
12914 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
12915 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
12916 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
12917 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
12918 // CHECK15: omp.body.continue:
12919 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
12920 // CHECK15: omp.inner.for.inc:
12921 // CHECK15-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12922 // CHECK15-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
12923 // CHECK15-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
12924 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
12925 // CHECK15: omp.inner.for.end:
12926 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
12927 // CHECK15: omp.loop.exit:
12928 // CHECK15-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
12929 // CHECK15-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
12930 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
12931 // CHECK15-NEXT: br label [[OMP_PRECOND_END]]
12932 // CHECK15: omp.precond.end:
12933 // CHECK15-NEXT: ret void
12934 //
12935 //
12936 // CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153
12937 // CHECK15-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
12938 // CHECK15-NEXT: entry:
12939 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
12940 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
12941 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
12942 // CHECK15-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
12943 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
12944 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
12945 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
12946 // CHECK15-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
12947 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
12948 // CHECK15-NEXT: ret void
12949 //
12950 //
12951 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..2
12952 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
12953 // CHECK15-NEXT: entry:
12954 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
12955 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
12956 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
12957 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
12958 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
12959 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
12960 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
12961 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
12962 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
12963 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
12964 // CHECK15-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
12965 // CHECK15-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
12966 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12967 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12968 // CHECK15-NEXT: [[I3:%.*]] = alloca i32, align 4
12969 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
12970 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
12971 // CHECK15-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
12972 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
12973 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
12974 // CHECK15-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
12975 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
12976 // CHECK15-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
12977 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
12978 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
12979 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
12980 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
12981 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
12982 // CHECK15-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
12983 // CHECK15-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
12984 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
12985 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
12986 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
12987 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
12988 // CHECK15: omp.precond.then:
12989 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
12990 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
12991 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
12992 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12993 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12994 // CHECK15-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
12995 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
12996 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
12997 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12998 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
12999 // CHECK15-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
13000 // CHECK15-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13001 // CHECK15: cond.true:
13002 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13003 // CHECK15-NEXT: br label [[COND_END:%.*]]
13004 // CHECK15: cond.false:
13005 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13006 // CHECK15-NEXT: br label [[COND_END]]
13007 // CHECK15: cond.end:
13008 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
13009 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
13010 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13011 // CHECK15-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
13012 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
13013 // CHECK15: omp.inner.for.cond:
13014 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13015 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13016 // CHECK15-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
13017 // CHECK15-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13018 // CHECK15: omp.inner.for.body:
13019 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13020 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13021 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
13022 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
13023 // CHECK15: omp.inner.for.inc:
13024 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13025 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
13026 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
13027 // CHECK15-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
13028 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
13029 // CHECK15: omp.inner.for.end:
13030 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
13031 // CHECK15: omp.loop.exit:
13032 // CHECK15-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13033 // CHECK15-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
13034 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
13035 // CHECK15-NEXT: br label [[OMP_PRECOND_END]]
13036 // CHECK15: omp.precond.end:
13037 // CHECK15-NEXT: ret void
13038 //
13039 //
13040 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..3
13041 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
13042 // CHECK15-NEXT: entry:
13043 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
13044 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
13045 // CHECK15-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
13046 // CHECK15-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
13047 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
13048 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
13049 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
13050 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
13051 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
13052 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
13053 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
13054 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
13055 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
13056 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
13057 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13058 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13059 // CHECK15-NEXT: [[I3:%.*]] = alloca i32, align 4
13060 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
13061 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
13062 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
13063 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
13064 // CHECK15-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
13065 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
13066 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
13067 // CHECK15-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
13068 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
13069 // CHECK15-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
13070 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
13071 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
13072 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13073 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
13074 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
13075 // CHECK15-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
13076 // CHECK15-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
13077 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
13078 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13079 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
13080 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
13081 // CHECK15: omp.precond.then:
13082 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
13083 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13084 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
13085 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
13086 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
13087 // CHECK15-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
13088 // CHECK15-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
13089 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13090 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13091 // CHECK15-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13092 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
13093 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
13094 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13095 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13096 // CHECK15-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
13097 // CHECK15-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13098 // CHECK15: cond.true:
13099 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13100 // CHECK15-NEXT: br label [[COND_END:%.*]]
13101 // CHECK15: cond.false:
13102 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13103 // CHECK15-NEXT: br label [[COND_END]]
13104 // CHECK15: cond.end:
13105 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
13106 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
13107 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13108 // CHECK15-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
13109 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
13110 // CHECK15: omp.inner.for.cond:
13111 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13112 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13113 // CHECK15-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
13114 // CHECK15-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13115 // CHECK15: omp.inner.for.body:
13116 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13117 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
13118 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13119 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
13120 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
13121 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
13122 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
13123 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
13124 // CHECK15: omp.body.continue:
13125 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
13126 // CHECK15: omp.inner.for.inc:
13127 // CHECK15-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13128 // CHECK15-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
13129 // CHECK15-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
13130 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
13131 // CHECK15: omp.inner.for.end:
13132 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
13133 // CHECK15: omp.loop.exit:
13134 // CHECK15-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13135 // CHECK15-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
13136 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
13137 // CHECK15-NEXT: br label [[OMP_PRECOND_END]]
13138 // CHECK15: omp.precond.end:
13139 // CHECK15-NEXT: ret void
13140 //
13141 //
13142 // CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158
13143 // CHECK15-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
13144 // CHECK15-NEXT: entry:
13145 // CHECK15-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
13146 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
13147 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
13148 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
13149 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
13150 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
13151 // CHECK15-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
13152 // CHECK15-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
13153 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
13154 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
13155 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
13156 // CHECK15-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
13157 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
13158 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
13159 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13160 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
13161 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
13162 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
13163 // CHECK15-NEXT: ret void
13164 //
13165 //
13166 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..5
13167 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
13168 // CHECK15-NEXT: entry:
13169 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
13170 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
13171 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
13172 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
13173 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
13174 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
13175 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
13176 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
13177 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
13178 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
13179 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
13180 // CHECK15-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
13181 // CHECK15-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
13182 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13183 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13184 // CHECK15-NEXT: [[I4:%.*]] = alloca i32, align 4
13185 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
13186 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
13187 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
13188 // CHECK15-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
13189 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
13190 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
13191 // CHECK15-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
13192 // CHECK15-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
13193 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
13194 // CHECK15-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
13195 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
13196 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
13197 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13198 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
13199 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
13200 // CHECK15-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
13201 // CHECK15-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
13202 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
13203 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13204 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
13205 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
13206 // CHECK15: omp.precond.then:
13207 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
13208 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13209 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
13210 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13211 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13212 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
13213 // CHECK15-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13214 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
13215 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
13216 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13217 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13218 // CHECK15-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
13219 // CHECK15-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13220 // CHECK15: cond.true:
13221 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13222 // CHECK15-NEXT: br label [[COND_END:%.*]]
13223 // CHECK15: cond.false:
13224 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13225 // CHECK15-NEXT: br label [[COND_END]]
13226 // CHECK15: cond.end:
13227 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
13228 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
13229 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13230 // CHECK15-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
13231 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
13232 // CHECK15: omp.inner.for.cond:
13233 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13234 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13235 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
13236 // CHECK15-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
13237 // CHECK15-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13238 // CHECK15: omp.inner.for.body:
13239 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13240 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13241 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
13242 // CHECK15-NEXT: store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
13243 // CHECK15-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
13244 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]])
13245 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
13246 // CHECK15: omp.inner.for.inc:
13247 // CHECK15-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13248 // CHECK15-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
13249 // CHECK15-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
13250 // CHECK15-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
13251 // CHECK15-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13252 // CHECK15-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
13253 // CHECK15-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
13254 // CHECK15-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4
13255 // CHECK15-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13256 // CHECK15-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
13257 // CHECK15-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
13258 // CHECK15-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4
13259 // CHECK15-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13260 // CHECK15-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13261 // CHECK15-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
13262 // CHECK15-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
13263 // CHECK15: cond.true11:
13264 // CHECK15-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13265 // CHECK15-NEXT: br label [[COND_END13:%.*]]
13266 // CHECK15: cond.false12:
13267 // CHECK15-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13268 // CHECK15-NEXT: br label [[COND_END13]]
13269 // CHECK15: cond.end13:
13270 // CHECK15-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
13271 // CHECK15-NEXT: store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4
13272 // CHECK15-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13273 // CHECK15-NEXT: store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
13274 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
13275 // CHECK15: omp.inner.for.end:
13276 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
13277 // CHECK15: omp.loop.exit:
13278 // CHECK15-NEXT: [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13279 // CHECK15-NEXT: [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
13280 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
13281 // CHECK15-NEXT: br label [[OMP_PRECOND_END]]
13282 // CHECK15: omp.precond.end:
13283 // CHECK15-NEXT: ret void
13284 //
13285 //
13286 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..6
13287 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
13288 // CHECK15-NEXT: entry:
13289 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
13290 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
13291 // CHECK15-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
13292 // CHECK15-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
13293 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
13294 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
13295 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
13296 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
13297 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
13298 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
13299 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
13300 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
13301 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
13302 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
13303 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
13304 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13305 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13306 // CHECK15-NEXT: [[I4:%.*]] = alloca i32, align 4
13307 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
13308 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
13309 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
13310 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
13311 // CHECK15-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
13312 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
13313 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
13314 // CHECK15-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
13315 // CHECK15-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
13316 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
13317 // CHECK15-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
13318 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
13319 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
13320 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13321 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
13322 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
13323 // CHECK15-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
13324 // CHECK15-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
13325 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
13326 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13327 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
13328 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
13329 // CHECK15: omp.precond.then:
13330 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
13331 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13332 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
13333 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
13334 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
13335 // CHECK15-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
13336 // CHECK15-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
13337 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13338 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13339 // CHECK15-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13340 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
13341 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
13342 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13343 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13344 // CHECK15-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
13345 // CHECK15-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13346 // CHECK15: cond.true:
13347 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13348 // CHECK15-NEXT: br label [[COND_END:%.*]]
13349 // CHECK15: cond.false:
13350 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13351 // CHECK15-NEXT: br label [[COND_END]]
13352 // CHECK15: cond.end:
13353 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
13354 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
13355 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13356 // CHECK15-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
13357 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
13358 // CHECK15: omp.inner.for.cond:
13359 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13360 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13361 // CHECK15-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
13362 // CHECK15-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13363 // CHECK15: omp.inner.for.body:
13364 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13365 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
13366 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13367 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
13368 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
13369 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
13370 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
13371 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
13372 // CHECK15: omp.body.continue:
13373 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
13374 // CHECK15: omp.inner.for.inc:
13375 // CHECK15-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13376 // CHECK15-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
13377 // CHECK15-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
13378 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
13379 // CHECK15: omp.inner.for.end:
13380 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
13381 // CHECK15: omp.loop.exit:
13382 // CHECK15-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13383 // CHECK15-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
13384 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
13385 // CHECK15-NEXT: br label [[OMP_PRECOND_END]]
13386 // CHECK15: omp.precond.end:
13387 // CHECK15-NEXT: ret void
13388 //
13389 //
13390 // CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163
13391 // CHECK15-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
13392 // CHECK15-NEXT: entry:
13393 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
13394 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
13395 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
13396 // CHECK15-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
13397 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
13398 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
13399 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
13400 // CHECK15-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
13401 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
13402 // CHECK15-NEXT: ret void
13403 //
13404 //
13405 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..8
13406 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
13407 // CHECK15-NEXT: entry:
13408 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
13409 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
13410 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
13411 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
13412 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
13413 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
13414 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
13415 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
13416 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
13417 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
13418 // CHECK15-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
13419 // CHECK15-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
13420 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13421 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13422 // CHECK15-NEXT: [[I3:%.*]] = alloca i32, align 4
13423 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
13424 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
13425 // CHECK15-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
13426 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
13427 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
13428 // CHECK15-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
13429 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
13430 // CHECK15-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
13431 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
13432 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
13433 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13434 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
13435 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
13436 // CHECK15-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
13437 // CHECK15-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
13438 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
13439 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13440 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
13441 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
13442 // CHECK15: omp.precond.then:
13443 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
13444 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13445 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
13446 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13447 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13448 // CHECK15-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13449 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
13450 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
13451 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13452 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13453 // CHECK15-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
13454 // CHECK15-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13455 // CHECK15: cond.true:
13456 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13457 // CHECK15-NEXT: br label [[COND_END:%.*]]
13458 // CHECK15: cond.false:
13459 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13460 // CHECK15-NEXT: br label [[COND_END]]
13461 // CHECK15: cond.end:
13462 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
13463 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
13464 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13465 // CHECK15-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
13466 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
13467 // CHECK15: omp.inner.for.cond:
13468 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13469 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13470 // CHECK15-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
13471 // CHECK15-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13472 // CHECK15: omp.inner.for.body:
13473 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13474 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13475 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
13476 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
13477 // CHECK15: omp.inner.for.inc:
13478 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13479 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
13480 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
13481 // CHECK15-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
13482 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
13483 // CHECK15: omp.inner.for.end:
13484 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
13485 // CHECK15: omp.loop.exit:
13486 // CHECK15-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13487 // CHECK15-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
13488 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
13489 // CHECK15-NEXT: br label [[OMP_PRECOND_END]]
13490 // CHECK15: omp.precond.end:
13491 // CHECK15-NEXT: ret void
13492 //
13493 //
13494 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..9
13495 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
13496 // CHECK15-NEXT: entry:
13497 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
13498 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
13499 // CHECK15-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
13500 // CHECK15-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
13501 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
13502 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
13503 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
13504 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
13505 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
13506 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
13507 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
13508 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
13509 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
13510 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
13511 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13512 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13513 // CHECK15-NEXT: [[I3:%.*]] = alloca i32, align 4
13514 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
13515 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
13516 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
13517 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
13518 // CHECK15-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
13519 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
13520 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
13521 // CHECK15-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
13522 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
13523 // CHECK15-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
13524 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
13525 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
13526 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13527 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
13528 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
13529 // CHECK15-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
13530 // CHECK15-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
13531 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
13532 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13533 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
13534 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
13535 // CHECK15: omp.precond.then:
13536 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
13537 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13538 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
13539 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
13540 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
13541 // CHECK15-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
13542 // CHECK15-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
13543 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13544 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13545 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13546 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13547 // CHECK15-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13548 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
13549 // CHECK15-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 1073741859, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
13550 // CHECK15-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
13551 // CHECK15: omp.dispatch.cond:
13552 // CHECK15-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13553 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
13554 // CHECK15-NEXT: [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
13555 // CHECK15-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
13556 // CHECK15-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
13557 // CHECK15: omp.dispatch.body:
13558 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13559 // CHECK15-NEXT: store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
13560 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
13561 // CHECK15: omp.inner.for.cond:
13562 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
13563 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !16
13564 // CHECK15-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
13565 // CHECK15-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13566 // CHECK15: omp.inner.for.body:
13567 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
13568 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
13569 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13570 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !16
13571 // CHECK15-NEXT: [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !16
13572 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
13573 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !16
13574 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
13575 // CHECK15: omp.body.continue:
13576 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
13577 // CHECK15: omp.inner.for.inc:
13578 // CHECK15-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
13579 // CHECK15-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP21]], 1
13580 // CHECK15-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
13581 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
13582 // CHECK15: omp.inner.for.end:
13583 // CHECK15-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
13584 // CHECK15: omp.dispatch.inc:
13585 // CHECK15-NEXT: br label [[OMP_DISPATCH_COND]]
13586 // CHECK15: omp.dispatch.end:
13587 // CHECK15-NEXT: br label [[OMP_PRECOND_END]]
13588 // CHECK15: omp.precond.end:
13589 // CHECK15-NEXT: ret void
13590 //
13591 //
13592 // CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168
13593 // CHECK15-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
13594 // CHECK15-NEXT: entry:
13595 // CHECK15-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
13596 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
13597 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
13598 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
13599 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
13600 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
13601 // CHECK15-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
13602 // CHECK15-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
13603 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
13604 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
13605 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
13606 // CHECK15-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
13607 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
13608 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
13609 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
13610 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
13611 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
13612 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
13613 // CHECK15-NEXT: ret void
13614 //
13615 //
13616 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..11
13617 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
13618 // CHECK15-NEXT: entry:
13619 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
13620 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
13621 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
13622 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
13623 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
13624 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
13625 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
13626 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
13627 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
13628 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
13629 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
13630 // CHECK15-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
13631 // CHECK15-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
13632 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13633 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13634 // CHECK15-NEXT: [[I4:%.*]] = alloca i32, align 4
13635 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
13636 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
13637 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
13638 // CHECK15-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
13639 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
13640 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
13641 // CHECK15-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
13642 // CHECK15-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
13643 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
13644 // CHECK15-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
13645 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
13646 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
13647 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13648 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
13649 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
13650 // CHECK15-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
13651 // CHECK15-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
13652 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
13653 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13654 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
13655 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
13656 // CHECK15: omp.precond.then:
13657 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
13658 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13659 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
13660 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13661 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13662 // CHECK15-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13663 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
13664 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
13665 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13666 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13667 // CHECK15-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
13668 // CHECK15-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13669 // CHECK15: cond.true:
13670 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13671 // CHECK15-NEXT: br label [[COND_END:%.*]]
13672 // CHECK15: cond.false:
13673 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13674 // CHECK15-NEXT: br label [[COND_END]]
13675 // CHECK15: cond.end:
13676 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
13677 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
13678 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13679 // CHECK15-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
13680 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
13681 // CHECK15: omp.inner.for.cond:
13682 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13683 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13684 // CHECK15-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
13685 // CHECK15-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13686 // CHECK15: omp.inner.for.body:
13687 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13688 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13689 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
13690 // CHECK15-NEXT: store i32 [[TMP18]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
13691 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
13692 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP19]])
13693 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
13694 // CHECK15: omp.inner.for.inc:
13695 // CHECK15-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13696 // CHECK15-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
13697 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
13698 // CHECK15-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
13699 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
13700 // CHECK15: omp.inner.for.end:
13701 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
13702 // CHECK15: omp.loop.exit:
13703 // CHECK15-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13704 // CHECK15-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
13705 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
13706 // CHECK15-NEXT: br label [[OMP_PRECOND_END]]
13707 // CHECK15: omp.precond.end:
13708 // CHECK15-NEXT: ret void
13709 //
13710 //
13711 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..12
13712 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
13713 // CHECK15-NEXT: entry:
13714 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
13715 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
13716 // CHECK15-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
13717 // CHECK15-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
13718 // CHECK15-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
13719 // CHECK15-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
13720 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
13721 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
13722 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
13723 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
13724 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
13725 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
13726 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
13727 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
13728 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
13729 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13730 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13731 // CHECK15-NEXT: [[I4:%.*]] = alloca i32, align 4
13732 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
13733 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
13734 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
13735 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
13736 // CHECK15-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
13737 // CHECK15-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
13738 // CHECK15-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
13739 // CHECK15-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
13740 // CHECK15-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
13741 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
13742 // CHECK15-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
13743 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
13744 // CHECK15-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
13745 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13746 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
13747 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
13748 // CHECK15-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
13749 // CHECK15-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
13750 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
13751 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
13752 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
13753 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
13754 // CHECK15: omp.precond.then:
13755 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
13756 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
13757 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
13758 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
13759 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
13760 // CHECK15-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
13761 // CHECK15-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
13762 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13763 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13764 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
13765 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13766 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13767 // CHECK15-NEXT: [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13768 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
13769 // CHECK15-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 1073741859, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
13770 // CHECK15-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
13771 // CHECK15: omp.dispatch.cond:
13772 // CHECK15-NEXT: [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13773 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
13774 // CHECK15-NEXT: [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
13775 // CHECK15-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
13776 // CHECK15-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
13777 // CHECK15: omp.dispatch.body:
13778 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13779 // CHECK15-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
13780 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
13781 // CHECK15: omp.inner.for.cond:
13782 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
13783 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
13784 // CHECK15-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
13785 // CHECK15-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13786 // CHECK15: omp.inner.for.body:
13787 // CHECK15-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
13788 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
13789 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13790 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !19
13791 // CHECK15-NEXT: [[TMP21:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !19
13792 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP21]]
13793 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !19
13794 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
13795 // CHECK15: omp.body.continue:
13796 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
13797 // CHECK15: omp.inner.for.inc:
13798 // CHECK15-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
13799 // CHECK15-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
13800 // CHECK15-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
13801 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
13802 // CHECK15: omp.inner.for.end:
13803 // CHECK15-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
13804 // CHECK15: omp.dispatch.inc:
13805 // CHECK15-NEXT: br label [[OMP_DISPATCH_COND]]
13806 // CHECK15: omp.dispatch.end:
13807 // CHECK15-NEXT: br label [[OMP_PRECOND_END]]
13808 // CHECK15: omp.precond.end:
13809 // CHECK15-NEXT: ret void
13810 //
13811 //
13812 // CHECK15-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
13813 // CHECK15-SAME: (i32 [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
13814 // CHECK15-NEXT: entry:
13815 // CHECK15-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
13816 // CHECK15-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
13817 // CHECK15-NEXT: [[M:%.*]] = alloca i32, align 4
13818 // CHECK15-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
13819 // CHECK15-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
13820 // CHECK15-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
13821 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
13822 // CHECK15-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
13823 // CHECK15-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
13824 // CHECK15-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
13825 // CHECK15-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
13826 // CHECK15-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
13827 // CHECK15-NEXT: [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 4
13828 // CHECK15-NEXT: [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 4
13829 // CHECK15-NEXT: [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 4
13830 // CHECK15-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
13831 // CHECK15-NEXT: [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 4
13832 // CHECK15-NEXT: [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 4
13833 // CHECK15-NEXT: [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 4
13834 // CHECK15-NEXT: [[_TMP16:%.*]] = alloca i32, align 4
13835 // CHECK15-NEXT: [[M_CASTED19:%.*]] = alloca i32, align 4
13836 // CHECK15-NEXT: [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [2 x i8*], align 4
13837 // CHECK15-NEXT: [[DOTOFFLOAD_PTRS21:%.*]] = alloca [2 x i8*], align 4
13838 // CHECK15-NEXT: [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [2 x i8*], align 4
13839 // CHECK15-NEXT: [[_TMP23:%.*]] = alloca i32, align 4
13840 // CHECK15-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
13841 // CHECK15-NEXT: store i32 10, i32* [[M]], align 4
13842 // CHECK15-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
13843 // CHECK15-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
13844 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
13845 // CHECK15-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
13846 // CHECK15-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
13847 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
13848 // CHECK15-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
13849 // CHECK15-NEXT: store i8* null, i8** [[TMP4]], align 4
13850 // CHECK15-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
13851 // CHECK15-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
13852 // CHECK15-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
13853 // CHECK15-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
13854 // CHECK15-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
13855 // CHECK15-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
13856 // CHECK15: omp_offload.failed:
13857 // CHECK15-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116([10 x i32]* [[A]]) #[[ATTR3]]
13858 // CHECK15-NEXT: br label [[OMP_OFFLOAD_CONT]]
13859 // CHECK15: omp_offload.cont:
13860 // CHECK15-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
13861 // CHECK15-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
13862 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 4
13863 // CHECK15-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
13864 // CHECK15-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
13865 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 4
13866 // CHECK15-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
13867 // CHECK15-NEXT: store i8* null, i8** [[TMP13]], align 4
13868 // CHECK15-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
13869 // CHECK15-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
13870 // CHECK15-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
13871 // CHECK15-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
13872 // CHECK15-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
13873 // CHECK15-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
13874 // CHECK15: omp_offload.failed5:
13875 // CHECK15-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121([10 x i32]* [[A]]) #[[ATTR3]]
13876 // CHECK15-NEXT: br label [[OMP_OFFLOAD_CONT6]]
13877 // CHECK15: omp_offload.cont6:
13878 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[M]], align 4
13879 // CHECK15-NEXT: store i32 [[TMP18]], i32* [[M_CASTED]], align 4
13880 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[M_CASTED]], align 4
13881 // CHECK15-NEXT: [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
13882 // CHECK15-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
13883 // CHECK15-NEXT: store i32 [[TMP19]], i32* [[TMP21]], align 4
13884 // CHECK15-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
13885 // CHECK15-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
13886 // CHECK15-NEXT: store i32 [[TMP19]], i32* [[TMP23]], align 4
13887 // CHECK15-NEXT: [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 0
13888 // CHECK15-NEXT: store i8* null, i8** [[TMP24]], align 4
13889 // CHECK15-NEXT: [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
13890 // CHECK15-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
13891 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 4
13892 // CHECK15-NEXT: [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
13893 // CHECK15-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
13894 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 4
13895 // CHECK15-NEXT: [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 1
13896 // CHECK15-NEXT: store i8* null, i8** [[TMP29]], align 4
13897 // CHECK15-NEXT: [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
13898 // CHECK15-NEXT: [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
13899 // CHECK15-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
13900 // CHECK15-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
13901 // CHECK15-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
13902 // CHECK15-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
13903 // CHECK15: omp_offload.failed11:
13904 // CHECK15-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126(i32 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
13905 // CHECK15-NEXT: br label [[OMP_OFFLOAD_CONT12]]
13906 // CHECK15: omp_offload.cont12:
13907 // CHECK15-NEXT: [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
13908 // CHECK15-NEXT: [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
13909 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 4
13910 // CHECK15-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
13911 // CHECK15-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
13912 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 4
13913 // CHECK15-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
13914 // CHECK15-NEXT: store i8* null, i8** [[TMP38]], align 4
13915 // CHECK15-NEXT: [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
13916 // CHECK15-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
13917 // CHECK15-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
13918 // CHECK15-NEXT: [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
13919 // CHECK15-NEXT: [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
13920 // CHECK15-NEXT: br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
13921 // CHECK15: omp_offload.failed17:
13922 // CHECK15-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131([10 x i32]* [[A]]) #[[ATTR3]]
13923 // CHECK15-NEXT: br label [[OMP_OFFLOAD_CONT18]]
13924 // CHECK15: omp_offload.cont18:
13925 // CHECK15-NEXT: [[TMP43:%.*]] = load i32, i32* [[M]], align 4
13926 // CHECK15-NEXT: store i32 [[TMP43]], i32* [[M_CASTED19]], align 4
13927 // CHECK15-NEXT: [[TMP44:%.*]] = load i32, i32* [[M_CASTED19]], align 4
13928 // CHECK15-NEXT: [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
13929 // CHECK15-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
13930 // CHECK15-NEXT: store i32 [[TMP44]], i32* [[TMP46]], align 4
13931 // CHECK15-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
13932 // CHECK15-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i32*
13933 // CHECK15-NEXT: store i32 [[TMP44]], i32* [[TMP48]], align 4
13934 // CHECK15-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
13935 // CHECK15-NEXT: store i8* null, i8** [[TMP49]], align 4
13936 // CHECK15-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
13937 // CHECK15-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
13938 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 4
13939 // CHECK15-NEXT: [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
13940 // CHECK15-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
13941 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 4
13942 // CHECK15-NEXT: [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
13943 // CHECK15-NEXT: store i8* null, i8** [[TMP54]], align 4
13944 // CHECK15-NEXT: [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
13945 // CHECK15-NEXT: [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
13946 // CHECK15-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
13947 // CHECK15-NEXT: [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
13948 // CHECK15-NEXT: [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
13949 // CHECK15-NEXT: br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
13950 // CHECK15: omp_offload.failed24:
13951 // CHECK15-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136(i32 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
13952 // CHECK15-NEXT: br label [[OMP_OFFLOAD_CONT25]]
13953 // CHECK15: omp_offload.cont25:
13954 // CHECK15-NEXT: ret i32 0
13955 //
13956 //
13957 // CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116
13958 // CHECK15-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
13959 // CHECK15-NEXT: entry:
13960 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
13961 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
13962 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
13963 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
13964 // CHECK15-NEXT: ret void
13965 //
13966 //
13967 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..14
13968 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
13969 // CHECK15-NEXT: entry:
13970 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
13971 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
13972 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
13973 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
13974 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
13975 // CHECK15-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
13976 // CHECK15-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
13977 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13978 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13979 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
13980 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
13981 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
13982 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
13983 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
13984 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
13985 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
13986 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13987 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13988 // CHECK15-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
13989 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
13990 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
13991 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13992 // CHECK15-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
13993 // CHECK15-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13994 // CHECK15: cond.true:
13995 // CHECK15-NEXT: br label [[COND_END:%.*]]
13996 // CHECK15: cond.false:
13997 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13998 // CHECK15-NEXT: br label [[COND_END]]
13999 // CHECK15: cond.end:
14000 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
14001 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
14002 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
14003 // CHECK15-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
14004 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
14005 // CHECK15: omp.inner.for.cond:
14006 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14007 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14008 // CHECK15-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
14009 // CHECK15-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14010 // CHECK15: omp.inner.for.body:
14011 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
14012 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14013 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
14014 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
14015 // CHECK15: omp.inner.for.inc:
14016 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14017 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
14018 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
14019 // CHECK15-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
14020 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
14021 // CHECK15: omp.inner.for.end:
14022 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
14023 // CHECK15: omp.loop.exit:
14024 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
14025 // CHECK15-NEXT: ret void
14026 //
14027 //
14028 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..15
14029 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
14030 // CHECK15-NEXT: entry:
14031 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14032 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14033 // CHECK15-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
14034 // CHECK15-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
14035 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14036 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
14037 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
14038 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
14039 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
14040 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14041 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14042 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
14043 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14044 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14045 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
14046 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
14047 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14048 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14049 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
14050 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
14051 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
14052 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
14053 // CHECK15-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
14054 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
14055 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14056 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14057 // CHECK15-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14058 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
14059 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
14060 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14061 // CHECK15-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
14062 // CHECK15-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
14063 // CHECK15: cond.true:
14064 // CHECK15-NEXT: br label [[COND_END:%.*]]
14065 // CHECK15: cond.false:
14066 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14067 // CHECK15-NEXT: br label [[COND_END]]
14068 // CHECK15: cond.end:
14069 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
14070 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
14071 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14072 // CHECK15-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
14073 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
14074 // CHECK15: omp.inner.for.cond:
14075 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14076 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14077 // CHECK15-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
14078 // CHECK15-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14079 // CHECK15: omp.inner.for.body:
14080 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14081 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
14082 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14083 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I]], align 4
14084 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
14085 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
14086 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
14087 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
14088 // CHECK15: omp.body.continue:
14089 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
14090 // CHECK15: omp.inner.for.inc:
14091 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14092 // CHECK15-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
14093 // CHECK15-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
14094 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
14095 // CHECK15: omp.inner.for.end:
14096 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
14097 // CHECK15: omp.loop.exit:
14098 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
14099 // CHECK15-NEXT: ret void
14100 //
14101 //
14102 // CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121
14103 // CHECK15-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
14104 // CHECK15-NEXT: entry:
14105 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14106 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14107 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14108 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
14109 // CHECK15-NEXT: ret void
14110 //
14111 //
14112 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..17
14113 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
14114 // CHECK15-NEXT: entry:
14115 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14116 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14117 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14118 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
14119 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
14120 // CHECK15-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
14121 // CHECK15-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
14122 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14123 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14124 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
14125 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14126 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14127 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14128 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14129 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
14130 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
14131 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14132 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14133 // CHECK15-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14134 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
14135 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
14136 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14137 // CHECK15-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
14138 // CHECK15-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
14139 // CHECK15: cond.true:
14140 // CHECK15-NEXT: br label [[COND_END:%.*]]
14141 // CHECK15: cond.false:
14142 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14143 // CHECK15-NEXT: br label [[COND_END]]
14144 // CHECK15: cond.end:
14145 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
14146 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
14147 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
14148 // CHECK15-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
14149 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
14150 // CHECK15: omp.inner.for.cond:
14151 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14152 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14153 // CHECK15-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
14154 // CHECK15-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14155 // CHECK15: omp.inner.for.body:
14156 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
14157 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14158 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
14159 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
14160 // CHECK15: omp.inner.for.inc:
14161 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14162 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
14163 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
14164 // CHECK15-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
14165 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
14166 // CHECK15: omp.inner.for.end:
14167 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
14168 // CHECK15: omp.loop.exit:
14169 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
14170 // CHECK15-NEXT: ret void
14171 //
14172 //
14173 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..18
14174 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
14175 // CHECK15-NEXT: entry:
14176 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14177 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14178 // CHECK15-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
14179 // CHECK15-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
14180 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14181 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
14182 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
14183 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
14184 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
14185 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14186 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14187 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
14188 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14189 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14190 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
14191 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
14192 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14193 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14194 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
14195 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
14196 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
14197 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
14198 // CHECK15-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
14199 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
14200 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14201 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14202 // CHECK15-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14203 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
14204 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
14205 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14206 // CHECK15-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
14207 // CHECK15-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
14208 // CHECK15: cond.true:
14209 // CHECK15-NEXT: br label [[COND_END:%.*]]
14210 // CHECK15: cond.false:
14211 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14212 // CHECK15-NEXT: br label [[COND_END]]
14213 // CHECK15: cond.end:
14214 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
14215 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
14216 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14217 // CHECK15-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
14218 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
14219 // CHECK15: omp.inner.for.cond:
14220 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14221 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14222 // CHECK15-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
14223 // CHECK15-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14224 // CHECK15: omp.inner.for.body:
14225 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14226 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
14227 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14228 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I]], align 4
14229 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
14230 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
14231 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
14232 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
14233 // CHECK15: omp.body.continue:
14234 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
14235 // CHECK15: omp.inner.for.inc:
14236 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14237 // CHECK15-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
14238 // CHECK15-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
14239 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
14240 // CHECK15: omp.inner.for.end:
14241 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
14242 // CHECK15: omp.loop.exit:
14243 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
14244 // CHECK15-NEXT: ret void
14245 //
14246 //
14247 // CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126
14248 // CHECK15-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
14249 // CHECK15-NEXT: entry:
14250 // CHECK15-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
14251 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14252 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
14253 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
14254 // CHECK15-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
14255 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14256 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14257 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
14258 // CHECK15-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
14259 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
14260 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
14261 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
14262 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
14263 // CHECK15-NEXT: ret void
14264 //
14265 //
14266 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..21
14267 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
14268 // CHECK15-NEXT: entry:
14269 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14270 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14271 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14272 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
14273 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
14274 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
14275 // CHECK15-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
14276 // CHECK15-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
14277 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14278 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14279 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
14280 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
14281 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14282 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14283 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14284 // CHECK15-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14285 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14286 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
14287 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
14288 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14289 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14290 // CHECK15-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14291 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
14292 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
14293 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14294 // CHECK15-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
14295 // CHECK15-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
14296 // CHECK15: cond.true:
14297 // CHECK15-NEXT: br label [[COND_END:%.*]]
14298 // CHECK15: cond.false:
14299 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14300 // CHECK15-NEXT: br label [[COND_END]]
14301 // CHECK15: cond.end:
14302 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
14303 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
14304 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
14305 // CHECK15-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
14306 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
14307 // CHECK15: omp.inner.for.cond:
14308 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14309 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14310 // CHECK15-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
14311 // CHECK15-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14312 // CHECK15: omp.inner.for.body:
14313 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
14314 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14315 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14316 // CHECK15-NEXT: store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
14317 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
14318 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
14319 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
14320 // CHECK15: omp.inner.for.inc:
14321 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14322 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
14323 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
14324 // CHECK15-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
14325 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
14326 // CHECK15: omp.inner.for.end:
14327 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
14328 // CHECK15: omp.loop.exit:
14329 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
14330 // CHECK15-NEXT: ret void
14331 //
14332 //
14333 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..22
14334 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
14335 // CHECK15-NEXT: entry:
14336 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14337 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14338 // CHECK15-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
14339 // CHECK15-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
14340 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14341 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
14342 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
14343 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
14344 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
14345 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
14346 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14347 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14348 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
14349 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14350 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14351 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
14352 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
14353 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14354 // CHECK15-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14355 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14356 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
14357 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
14358 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
14359 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
14360 // CHECK15-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
14361 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
14362 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14363 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14364 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14365 // CHECK15-NEXT: [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14366 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
14367 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
14368 // CHECK15-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
14369 // CHECK15: omp.dispatch.cond:
14370 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14371 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
14372 // CHECK15-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
14373 // CHECK15-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
14374 // CHECK15: cond.true:
14375 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
14376 // CHECK15-NEXT: br label [[COND_END:%.*]]
14377 // CHECK15: cond.false:
14378 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14379 // CHECK15-NEXT: br label [[COND_END]]
14380 // CHECK15: cond.end:
14381 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
14382 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
14383 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14384 // CHECK15-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
14385 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14386 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14387 // CHECK15-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
14388 // CHECK15-NEXT: br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
14389 // CHECK15: omp.dispatch.body:
14390 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
14391 // CHECK15: omp.inner.for.cond:
14392 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14393 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14394 // CHECK15-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
14395 // CHECK15-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14396 // CHECK15: omp.inner.for.body:
14397 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14398 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
14399 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14400 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I]], align 4
14401 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4
14402 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP16]]
14403 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
14404 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
14405 // CHECK15: omp.body.continue:
14406 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
14407 // CHECK15: omp.inner.for.inc:
14408 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14409 // CHECK15-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
14410 // CHECK15-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
14411 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
14412 // CHECK15: omp.inner.for.end:
14413 // CHECK15-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
14414 // CHECK15: omp.dispatch.inc:
14415 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14416 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
14417 // CHECK15-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
14418 // CHECK15-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
14419 // CHECK15-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14420 // CHECK15-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
14421 // CHECK15-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
14422 // CHECK15-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
14423 // CHECK15-NEXT: br label [[OMP_DISPATCH_COND]]
14424 // CHECK15: omp.dispatch.end:
14425 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
14426 // CHECK15-NEXT: ret void
14427 //
14428 //
14429 // CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
14430 // CHECK15-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
14431 // CHECK15-NEXT: entry:
14432 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14433 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14434 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14435 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
14436 // CHECK15-NEXT: ret void
14437 //
14438 //
14439 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..25
14440 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
14441 // CHECK15-NEXT: entry:
14442 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14443 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14444 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14445 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
14446 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
14447 // CHECK15-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
14448 // CHECK15-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
14449 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14450 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14451 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
14452 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14453 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14454 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14455 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14456 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
14457 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
14458 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14459 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14460 // CHECK15-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14461 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
14462 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
14463 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14464 // CHECK15-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
14465 // CHECK15-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
14466 // CHECK15: cond.true:
14467 // CHECK15-NEXT: br label [[COND_END:%.*]]
14468 // CHECK15: cond.false:
14469 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14470 // CHECK15-NEXT: br label [[COND_END]]
14471 // CHECK15: cond.end:
14472 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
14473 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
14474 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
14475 // CHECK15-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
14476 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
14477 // CHECK15: omp.inner.for.cond:
14478 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14479 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14480 // CHECK15-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
14481 // CHECK15-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14482 // CHECK15: omp.inner.for.body:
14483 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
14484 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14485 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
14486 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
14487 // CHECK15: omp.inner.for.inc:
14488 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14489 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
14490 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
14491 // CHECK15-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
14492 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
14493 // CHECK15: omp.inner.for.end:
14494 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
14495 // CHECK15: omp.loop.exit:
14496 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
14497 // CHECK15-NEXT: ret void
14498 //
14499 //
14500 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..26
14501 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
14502 // CHECK15-NEXT: entry:
14503 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14504 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14505 // CHECK15-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
14506 // CHECK15-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
14507 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14508 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
14509 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
14510 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
14511 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
14512 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14513 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14514 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
14515 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14516 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14517 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
14518 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
14519 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14520 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14521 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
14522 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
14523 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
14524 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
14525 // CHECK15-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
14526 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
14527 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14528 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14529 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14530 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14531 // CHECK15-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14532 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
14533 // CHECK15-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
14534 // CHECK15-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
14535 // CHECK15: omp.dispatch.cond:
14536 // CHECK15-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
14537 // CHECK15-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
14538 // CHECK15-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
14539 // CHECK15: omp.dispatch.body:
14540 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14541 // CHECK15-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
14542 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
14543 // CHECK15: omp.inner.for.cond:
14544 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
14545 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
14546 // CHECK15-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
14547 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14548 // CHECK15: omp.inner.for.body:
14549 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
14550 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
14551 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14552 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
14553 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !22
14554 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP12]]
14555 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !22
14556 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
14557 // CHECK15: omp.body.continue:
14558 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
14559 // CHECK15: omp.inner.for.inc:
14560 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
14561 // CHECK15-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
14562 // CHECK15-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
14563 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
14564 // CHECK15: omp.inner.for.end:
14565 // CHECK15-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
14566 // CHECK15: omp.dispatch.inc:
14567 // CHECK15-NEXT: br label [[OMP_DISPATCH_COND]]
14568 // CHECK15: omp.dispatch.end:
14569 // CHECK15-NEXT: ret void
14570 //
14571 //
14572 // CHECK15-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136
14573 // CHECK15-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
14574 // CHECK15-NEXT: entry:
14575 // CHECK15-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
14576 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14577 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
14578 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
14579 // CHECK15-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
14580 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14581 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14582 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
14583 // CHECK15-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
14584 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
14585 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
14586 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
14587 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
14588 // CHECK15-NEXT: ret void
14589 //
14590 //
14591 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..29
14592 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
14593 // CHECK15-NEXT: entry:
14594 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14595 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14596 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14597 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
14598 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
14599 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
14600 // CHECK15-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
14601 // CHECK15-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
14602 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14603 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14604 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
14605 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
14606 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14607 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14608 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14609 // CHECK15-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14610 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14611 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
14612 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
14613 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14614 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14615 // CHECK15-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14616 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
14617 // CHECK15-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
14618 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14619 // CHECK15-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
14620 // CHECK15-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
14621 // CHECK15: cond.true:
14622 // CHECK15-NEXT: br label [[COND_END:%.*]]
14623 // CHECK15: cond.false:
14624 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14625 // CHECK15-NEXT: br label [[COND_END]]
14626 // CHECK15: cond.end:
14627 // CHECK15-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
14628 // CHECK15-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
14629 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
14630 // CHECK15-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
14631 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
14632 // CHECK15: omp.inner.for.cond:
14633 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14634 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14635 // CHECK15-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
14636 // CHECK15-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14637 // CHECK15: omp.inner.for.body:
14638 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
14639 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
14640 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14641 // CHECK15-NEXT: store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
14642 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
14643 // CHECK15-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
14644 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
14645 // CHECK15: omp.inner.for.inc:
14646 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
14647 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
14648 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
14649 // CHECK15-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
14650 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]]
14651 // CHECK15: omp.inner.for.end:
14652 // CHECK15-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
14653 // CHECK15: omp.loop.exit:
14654 // CHECK15-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
14655 // CHECK15-NEXT: ret void
14656 //
14657 //
14658 // CHECK15-LABEL: define {{[^@]+}}@.omp_outlined..30
14659 // CHECK15-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
14660 // CHECK15-NEXT: entry:
14661 // CHECK15-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
14662 // CHECK15-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
14663 // CHECK15-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
14664 // CHECK15-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
14665 // CHECK15-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
14666 // CHECK15-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
14667 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
14668 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
14669 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
14670 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
14671 // CHECK15-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
14672 // CHECK15-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
14673 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
14674 // CHECK15-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
14675 // CHECK15-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
14676 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
14677 // CHECK15-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
14678 // CHECK15-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
14679 // CHECK15-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14680 // CHECK15-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
14681 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
14682 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
14683 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
14684 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
14685 // CHECK15-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
14686 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
14687 // CHECK15-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
14688 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
14689 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
14690 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14691 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
14692 // CHECK15-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
14693 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
14694 // CHECK15-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 1073741859, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
14695 // CHECK15-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
14696 // CHECK15: omp.dispatch.cond:
14697 // CHECK15-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
14698 // CHECK15-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
14699 // CHECK15-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
14700 // CHECK15: omp.dispatch.body:
14701 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14702 // CHECK15-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
14703 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
14704 // CHECK15: omp.inner.for.cond:
14705 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
14706 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
14707 // CHECK15-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
14708 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14709 // CHECK15: omp.inner.for.body:
14710 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
14711 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
14712 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14713 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !25
14714 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !25
14715 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
14716 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
14717 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
14718 // CHECK15: omp.body.continue:
14719 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
14720 // CHECK15: omp.inner.for.inc:
14721 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
14722 // CHECK15-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP14]], 1
14723 // CHECK15-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
14724 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
14725 // CHECK15: omp.inner.for.end:
14726 // CHECK15-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
14727 // CHECK15: omp.dispatch.inc:
14728 // CHECK15-NEXT: br label [[OMP_DISPATCH_COND]]
14729 // CHECK15: omp.dispatch.end:
14730 // CHECK15-NEXT: ret void
14731 //
14732 //
14733 // CHECK15-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
14734 // CHECK15-SAME: () #[[ATTR5:[0-9]+]] {
14735 // CHECK15-NEXT: entry:
14736 // CHECK15-NEXT: call void @__tgt_register_requires(i64 1)
14737 // CHECK15-NEXT: ret void
14738 //
14739 //
14740 // CHECK16-LABEL: define {{[^@]+}}@main
14741 // CHECK16-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
14742 // CHECK16-NEXT: entry:
14743 // CHECK16-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
14744 // CHECK16-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
14745 // CHECK16-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
14746 // CHECK16-NEXT: [[N:%.*]] = alloca i32, align 4
14747 // CHECK16-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
14748 // CHECK16-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
14749 // CHECK16-NEXT: [[M:%.*]] = alloca i32, align 4
14750 // CHECK16-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
14751 // CHECK16-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
14752 // CHECK16-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
14753 // CHECK16-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
14754 // CHECK16-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
14755 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
14756 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
14757 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
14758 // CHECK16-NEXT: [[N_CASTED3:%.*]] = alloca i32, align 4
14759 // CHECK16-NEXT: [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
14760 // CHECK16-NEXT: [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
14761 // CHECK16-NEXT: [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
14762 // CHECK16-NEXT: [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
14763 // CHECK16-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
14764 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
14765 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
14766 // CHECK16-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
14767 // CHECK16-NEXT: [[N_CASTED17:%.*]] = alloca i32, align 4
14768 // CHECK16-NEXT: [[DOTOFFLOAD_BASEPTRS18:%.*]] = alloca [4 x i8*], align 4
14769 // CHECK16-NEXT: [[DOTOFFLOAD_PTRS19:%.*]] = alloca [4 x i8*], align 4
14770 // CHECK16-NEXT: [[DOTOFFLOAD_MAPPERS20:%.*]] = alloca [4 x i8*], align 4
14771 // CHECK16-NEXT: [[DOTOFFLOAD_SIZES21:%.*]] = alloca [4 x i64], align 4
14772 // CHECK16-NEXT: [[_TMP22:%.*]] = alloca i32, align 4
14773 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_23:%.*]] = alloca i32, align 4
14774 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
14775 // CHECK16-NEXT: [[N_CASTED31:%.*]] = alloca i32, align 4
14776 // CHECK16-NEXT: [[DOTOFFLOAD_BASEPTRS32:%.*]] = alloca [3 x i8*], align 4
14777 // CHECK16-NEXT: [[DOTOFFLOAD_PTRS33:%.*]] = alloca [3 x i8*], align 4
14778 // CHECK16-NEXT: [[DOTOFFLOAD_MAPPERS34:%.*]] = alloca [3 x i8*], align 4
14779 // CHECK16-NEXT: [[DOTOFFLOAD_SIZES35:%.*]] = alloca [3 x i64], align 4
14780 // CHECK16-NEXT: [[_TMP36:%.*]] = alloca i32, align 4
14781 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_37:%.*]] = alloca i32, align 4
14782 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
14783 // CHECK16-NEXT: [[M_CASTED45:%.*]] = alloca i32, align 4
14784 // CHECK16-NEXT: [[N_CASTED46:%.*]] = alloca i32, align 4
14785 // CHECK16-NEXT: [[DOTOFFLOAD_BASEPTRS47:%.*]] = alloca [4 x i8*], align 4
14786 // CHECK16-NEXT: [[DOTOFFLOAD_PTRS48:%.*]] = alloca [4 x i8*], align 4
14787 // CHECK16-NEXT: [[DOTOFFLOAD_MAPPERS49:%.*]] = alloca [4 x i8*], align 4
14788 // CHECK16-NEXT: [[DOTOFFLOAD_SIZES50:%.*]] = alloca [4 x i64], align 4
14789 // CHECK16-NEXT: [[_TMP51:%.*]] = alloca i32, align 4
14790 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_52:%.*]] = alloca i32, align 4
14791 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_53:%.*]] = alloca i32, align 4
14792 // CHECK16-NEXT: store i32 0, i32* [[RETVAL]], align 4
14793 // CHECK16-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
14794 // CHECK16-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
14795 // CHECK16-NEXT: store i32 100, i32* [[N]], align 4
14796 // CHECK16-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
14797 // CHECK16-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
14798 // CHECK16-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
14799 // CHECK16-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
14800 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
14801 // CHECK16-NEXT: store i32 10, i32* [[M]], align 4
14802 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
14803 // CHECK16-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
14804 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
14805 // CHECK16-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
14806 // CHECK16-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
14807 // CHECK16-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
14808 // CHECK16-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
14809 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[TMP7]], align 4
14810 // CHECK16-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
14811 // CHECK16-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
14812 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[TMP9]], align 4
14813 // CHECK16-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
14814 // CHECK16-NEXT: store i64 4, i64* [[TMP10]], align 4
14815 // CHECK16-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
14816 // CHECK16-NEXT: store i8* null, i8** [[TMP11]], align 4
14817 // CHECK16-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
14818 // CHECK16-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
14819 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
14820 // CHECK16-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
14821 // CHECK16-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
14822 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
14823 // CHECK16-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
14824 // CHECK16-NEXT: store i64 4, i64* [[TMP16]], align 4
14825 // CHECK16-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
14826 // CHECK16-NEXT: store i8* null, i8** [[TMP17]], align 4
14827 // CHECK16-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
14828 // CHECK16-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
14829 // CHECK16-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 4
14830 // CHECK16-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
14831 // CHECK16-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
14832 // CHECK16-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 4
14833 // CHECK16-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
14834 // CHECK16-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 4
14835 // CHECK16-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
14836 // CHECK16-NEXT: store i8* null, i8** [[TMP23]], align 4
14837 // CHECK16-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
14838 // CHECK16-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
14839 // CHECK16-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
14840 // CHECK16-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
14841 // CHECK16-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
14842 // CHECK16-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
14843 // CHECK16-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
14844 // CHECK16-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
14845 // CHECK16-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
14846 // CHECK16-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
14847 // CHECK16-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
14848 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
14849 // CHECK16-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
14850 // CHECK16-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
14851 // CHECK16-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
14852 // CHECK16-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
14853 // CHECK16-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
14854 // CHECK16: omp_offload.failed:
14855 // CHECK16-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
14856 // CHECK16-NEXT: br label [[OMP_OFFLOAD_CONT]]
14857 // CHECK16: omp_offload.cont:
14858 // CHECK16-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
14859 // CHECK16-NEXT: store i32 [[TMP33]], i32* [[N_CASTED3]], align 4
14860 // CHECK16-NEXT: [[TMP34:%.*]] = load i32, i32* [[N_CASTED3]], align 4
14861 // CHECK16-NEXT: [[TMP35:%.*]] = mul nuw i32 [[TMP0]], 4
14862 // CHECK16-NEXT: [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
14863 // CHECK16-NEXT: [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
14864 // CHECK16-NEXT: [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
14865 // CHECK16-NEXT: store i32 [[TMP34]], i32* [[TMP38]], align 4
14866 // CHECK16-NEXT: [[TMP39:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
14867 // CHECK16-NEXT: [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
14868 // CHECK16-NEXT: store i32 [[TMP34]], i32* [[TMP40]], align 4
14869 // CHECK16-NEXT: [[TMP41:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
14870 // CHECK16-NEXT: store i64 4, i64* [[TMP41]], align 4
14871 // CHECK16-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
14872 // CHECK16-NEXT: store i8* null, i8** [[TMP42]], align 4
14873 // CHECK16-NEXT: [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
14874 // CHECK16-NEXT: [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i32*
14875 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[TMP44]], align 4
14876 // CHECK16-NEXT: [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
14877 // CHECK16-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
14878 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[TMP46]], align 4
14879 // CHECK16-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 1
14880 // CHECK16-NEXT: store i64 4, i64* [[TMP47]], align 4
14881 // CHECK16-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
14882 // CHECK16-NEXT: store i8* null, i8** [[TMP48]], align 4
14883 // CHECK16-NEXT: [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
14884 // CHECK16-NEXT: [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32**
14885 // CHECK16-NEXT: store i32* [[VLA]], i32** [[TMP50]], align 4
14886 // CHECK16-NEXT: [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
14887 // CHECK16-NEXT: [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32**
14888 // CHECK16-NEXT: store i32* [[VLA]], i32** [[TMP52]], align 4
14889 // CHECK16-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
14890 // CHECK16-NEXT: store i64 [[TMP36]], i64* [[TMP53]], align 4
14891 // CHECK16-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
14892 // CHECK16-NEXT: store i8* null, i8** [[TMP54]], align 4
14893 // CHECK16-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
14894 // CHECK16-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
14895 // CHECK16-NEXT: [[TMP57:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
14896 // CHECK16-NEXT: [[TMP58:%.*]] = load i32, i32* [[N]], align 4
14897 // CHECK16-NEXT: store i32 [[TMP58]], i32* [[DOTCAPTURE_EXPR_9]], align 4
14898 // CHECK16-NEXT: [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
14899 // CHECK16-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP59]], 0
14900 // CHECK16-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
14901 // CHECK16-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
14902 // CHECK16-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
14903 // CHECK16-NEXT: [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
14904 // CHECK16-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP60]], 1
14905 // CHECK16-NEXT: [[TMP61:%.*]] = zext i32 [[ADD14]] to i64
14906 // CHECK16-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP61]])
14907 // CHECK16-NEXT: [[TMP62:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153.region_id, i32 3, i8** [[TMP55]], i8** [[TMP56]], i64* [[TMP57]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
14908 // CHECK16-NEXT: [[TMP63:%.*]] = icmp ne i32 [[TMP62]], 0
14909 // CHECK16-NEXT: br i1 [[TMP63]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
14910 // CHECK16: omp_offload.failed15:
14911 // CHECK16-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153(i32 [[TMP34]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
14912 // CHECK16-NEXT: br label [[OMP_OFFLOAD_CONT16]]
14913 // CHECK16: omp_offload.cont16:
14914 // CHECK16-NEXT: [[TMP64:%.*]] = load i32, i32* [[M]], align 4
14915 // CHECK16-NEXT: store i32 [[TMP64]], i32* [[M_CASTED]], align 4
14916 // CHECK16-NEXT: [[TMP65:%.*]] = load i32, i32* [[M_CASTED]], align 4
14917 // CHECK16-NEXT: [[TMP66:%.*]] = load i32, i32* [[N]], align 4
14918 // CHECK16-NEXT: store i32 [[TMP66]], i32* [[N_CASTED17]], align 4
14919 // CHECK16-NEXT: [[TMP67:%.*]] = load i32, i32* [[N_CASTED17]], align 4
14920 // CHECK16-NEXT: [[TMP68:%.*]] = mul nuw i32 [[TMP0]], 4
14921 // CHECK16-NEXT: [[TMP69:%.*]] = sext i32 [[TMP68]] to i64
14922 // CHECK16-NEXT: [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
14923 // CHECK16-NEXT: [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i32*
14924 // CHECK16-NEXT: store i32 [[TMP65]], i32* [[TMP71]], align 4
14925 // CHECK16-NEXT: [[TMP72:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
14926 // CHECK16-NEXT: [[TMP73:%.*]] = bitcast i8** [[TMP72]] to i32*
14927 // CHECK16-NEXT: store i32 [[TMP65]], i32* [[TMP73]], align 4
14928 // CHECK16-NEXT: [[TMP74:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
14929 // CHECK16-NEXT: store i64 4, i64* [[TMP74]], align 4
14930 // CHECK16-NEXT: [[TMP75:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 0
14931 // CHECK16-NEXT: store i8* null, i8** [[TMP75]], align 4
14932 // CHECK16-NEXT: [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 1
14933 // CHECK16-NEXT: [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i32*
14934 // CHECK16-NEXT: store i32 [[TMP67]], i32* [[TMP77]], align 4
14935 // CHECK16-NEXT: [[TMP78:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 1
14936 // CHECK16-NEXT: [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
14937 // CHECK16-NEXT: store i32 [[TMP67]], i32* [[TMP79]], align 4
14938 // CHECK16-NEXT: [[TMP80:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 1
14939 // CHECK16-NEXT: store i64 4, i64* [[TMP80]], align 4
14940 // CHECK16-NEXT: [[TMP81:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 1
14941 // CHECK16-NEXT: store i8* null, i8** [[TMP81]], align 4
14942 // CHECK16-NEXT: [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 2
14943 // CHECK16-NEXT: [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i32*
14944 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[TMP83]], align 4
14945 // CHECK16-NEXT: [[TMP84:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 2
14946 // CHECK16-NEXT: [[TMP85:%.*]] = bitcast i8** [[TMP84]] to i32*
14947 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[TMP85]], align 4
14948 // CHECK16-NEXT: [[TMP86:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 2
14949 // CHECK16-NEXT: store i64 4, i64* [[TMP86]], align 4
14950 // CHECK16-NEXT: [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 2
14951 // CHECK16-NEXT: store i8* null, i8** [[TMP87]], align 4
14952 // CHECK16-NEXT: [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 3
14953 // CHECK16-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
14954 // CHECK16-NEXT: store i32* [[VLA]], i32** [[TMP89]], align 4
14955 // CHECK16-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 3
14956 // CHECK16-NEXT: [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i32**
14957 // CHECK16-NEXT: store i32* [[VLA]], i32** [[TMP91]], align 4
14958 // CHECK16-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 3
14959 // CHECK16-NEXT: store i64 [[TMP69]], i64* [[TMP92]], align 4
14960 // CHECK16-NEXT: [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 3
14961 // CHECK16-NEXT: store i8* null, i8** [[TMP93]], align 4
14962 // CHECK16-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
14963 // CHECK16-NEXT: [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
14964 // CHECK16-NEXT: [[TMP96:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
14965 // CHECK16-NEXT: [[TMP97:%.*]] = load i32, i32* [[N]], align 4
14966 // CHECK16-NEXT: store i32 [[TMP97]], i32* [[DOTCAPTURE_EXPR_23]], align 4
14967 // CHECK16-NEXT: [[TMP98:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_23]], align 4
14968 // CHECK16-NEXT: [[SUB25:%.*]] = sub nsw i32 [[TMP98]], 0
14969 // CHECK16-NEXT: [[DIV26:%.*]] = sdiv i32 [[SUB25]], 1
14970 // CHECK16-NEXT: [[SUB27:%.*]] = sub nsw i32 [[DIV26]], 1
14971 // CHECK16-NEXT: store i32 [[SUB27]], i32* [[DOTCAPTURE_EXPR_24]], align 4
14972 // CHECK16-NEXT: [[TMP99:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
14973 // CHECK16-NEXT: [[ADD28:%.*]] = add nsw i32 [[TMP99]], 1
14974 // CHECK16-NEXT: [[TMP100:%.*]] = zext i32 [[ADD28]] to i64
14975 // CHECK16-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP100]])
14976 // CHECK16-NEXT: [[TMP101:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158.region_id, i32 4, i8** [[TMP94]], i8** [[TMP95]], i64* [[TMP96]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
14977 // CHECK16-NEXT: [[TMP102:%.*]] = icmp ne i32 [[TMP101]], 0
14978 // CHECK16-NEXT: br i1 [[TMP102]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
14979 // CHECK16: omp_offload.failed29:
14980 // CHECK16-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158(i32 [[TMP65]], i32 [[TMP67]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
14981 // CHECK16-NEXT: br label [[OMP_OFFLOAD_CONT30]]
14982 // CHECK16: omp_offload.cont30:
14983 // CHECK16-NEXT: [[TMP103:%.*]] = load i32, i32* [[N]], align 4
14984 // CHECK16-NEXT: store i32 [[TMP103]], i32* [[N_CASTED31]], align 4
14985 // CHECK16-NEXT: [[TMP104:%.*]] = load i32, i32* [[N_CASTED31]], align 4
14986 // CHECK16-NEXT: [[TMP105:%.*]] = mul nuw i32 [[TMP0]], 4
14987 // CHECK16-NEXT: [[TMP106:%.*]] = sext i32 [[TMP105]] to i64
14988 // CHECK16-NEXT: [[TMP107:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
14989 // CHECK16-NEXT: [[TMP108:%.*]] = bitcast i8** [[TMP107]] to i32*
14990 // CHECK16-NEXT: store i32 [[TMP104]], i32* [[TMP108]], align 4
14991 // CHECK16-NEXT: [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
14992 // CHECK16-NEXT: [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i32*
14993 // CHECK16-NEXT: store i32 [[TMP104]], i32* [[TMP110]], align 4
14994 // CHECK16-NEXT: [[TMP111:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
14995 // CHECK16-NEXT: store i64 4, i64* [[TMP111]], align 4
14996 // CHECK16-NEXT: [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 0
14997 // CHECK16-NEXT: store i8* null, i8** [[TMP112]], align 4
14998 // CHECK16-NEXT: [[TMP113:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 1
14999 // CHECK16-NEXT: [[TMP114:%.*]] = bitcast i8** [[TMP113]] to i32*
15000 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[TMP114]], align 4
15001 // CHECK16-NEXT: [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 1
15002 // CHECK16-NEXT: [[TMP116:%.*]] = bitcast i8** [[TMP115]] to i32*
15003 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[TMP116]], align 4
15004 // CHECK16-NEXT: [[TMP117:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 1
15005 // CHECK16-NEXT: store i64 4, i64* [[TMP117]], align 4
15006 // CHECK16-NEXT: [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 1
15007 // CHECK16-NEXT: store i8* null, i8** [[TMP118]], align 4
15008 // CHECK16-NEXT: [[TMP119:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 2
15009 // CHECK16-NEXT: [[TMP120:%.*]] = bitcast i8** [[TMP119]] to i32**
15010 // CHECK16-NEXT: store i32* [[VLA]], i32** [[TMP120]], align 4
15011 // CHECK16-NEXT: [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 2
15012 // CHECK16-NEXT: [[TMP122:%.*]] = bitcast i8** [[TMP121]] to i32**
15013 // CHECK16-NEXT: store i32* [[VLA]], i32** [[TMP122]], align 4
15014 // CHECK16-NEXT: [[TMP123:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 2
15015 // CHECK16-NEXT: store i64 [[TMP106]], i64* [[TMP123]], align 4
15016 // CHECK16-NEXT: [[TMP124:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 2
15017 // CHECK16-NEXT: store i8* null, i8** [[TMP124]], align 4
15018 // CHECK16-NEXT: [[TMP125:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
15019 // CHECK16-NEXT: [[TMP126:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
15020 // CHECK16-NEXT: [[TMP127:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
15021 // CHECK16-NEXT: [[TMP128:%.*]] = load i32, i32* [[N]], align 4
15022 // CHECK16-NEXT: store i32 [[TMP128]], i32* [[DOTCAPTURE_EXPR_37]], align 4
15023 // CHECK16-NEXT: [[TMP129:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_37]], align 4
15024 // CHECK16-NEXT: [[SUB39:%.*]] = sub nsw i32 [[TMP129]], 0
15025 // CHECK16-NEXT: [[DIV40:%.*]] = sdiv i32 [[SUB39]], 1
15026 // CHECK16-NEXT: [[SUB41:%.*]] = sub nsw i32 [[DIV40]], 1
15027 // CHECK16-NEXT: store i32 [[SUB41]], i32* [[DOTCAPTURE_EXPR_38]], align 4
15028 // CHECK16-NEXT: [[TMP130:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_38]], align 4
15029 // CHECK16-NEXT: [[ADD42:%.*]] = add nsw i32 [[TMP130]], 1
15030 // CHECK16-NEXT: [[TMP131:%.*]] = zext i32 [[ADD42]] to i64
15031 // CHECK16-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP131]])
15032 // CHECK16-NEXT: [[TMP132:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163.region_id, i32 3, i8** [[TMP125]], i8** [[TMP126]], i64* [[TMP127]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
15033 // CHECK16-NEXT: [[TMP133:%.*]] = icmp ne i32 [[TMP132]], 0
15034 // CHECK16-NEXT: br i1 [[TMP133]], label [[OMP_OFFLOAD_FAILED43:%.*]], label [[OMP_OFFLOAD_CONT44:%.*]]
15035 // CHECK16: omp_offload.failed43:
15036 // CHECK16-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163(i32 [[TMP104]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
15037 // CHECK16-NEXT: br label [[OMP_OFFLOAD_CONT44]]
15038 // CHECK16: omp_offload.cont44:
15039 // CHECK16-NEXT: [[TMP134:%.*]] = load i32, i32* [[M]], align 4
15040 // CHECK16-NEXT: store i32 [[TMP134]], i32* [[M_CASTED45]], align 4
15041 // CHECK16-NEXT: [[TMP135:%.*]] = load i32, i32* [[M_CASTED45]], align 4
15042 // CHECK16-NEXT: [[TMP136:%.*]] = load i32, i32* [[N]], align 4
15043 // CHECK16-NEXT: store i32 [[TMP136]], i32* [[N_CASTED46]], align 4
15044 // CHECK16-NEXT: [[TMP137:%.*]] = load i32, i32* [[N_CASTED46]], align 4
15045 // CHECK16-NEXT: [[TMP138:%.*]] = mul nuw i32 [[TMP0]], 4
15046 // CHECK16-NEXT: [[TMP139:%.*]] = sext i32 [[TMP138]] to i64
15047 // CHECK16-NEXT: [[TMP140:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
15048 // CHECK16-NEXT: [[TMP141:%.*]] = bitcast i8** [[TMP140]] to i32*
15049 // CHECK16-NEXT: store i32 [[TMP135]], i32* [[TMP141]], align 4
15050 // CHECK16-NEXT: [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
15051 // CHECK16-NEXT: [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i32*
15052 // CHECK16-NEXT: store i32 [[TMP135]], i32* [[TMP143]], align 4
15053 // CHECK16-NEXT: [[TMP144:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
15054 // CHECK16-NEXT: store i64 4, i64* [[TMP144]], align 4
15055 // CHECK16-NEXT: [[TMP145:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 0
15056 // CHECK16-NEXT: store i8* null, i8** [[TMP145]], align 4
15057 // CHECK16-NEXT: [[TMP146:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 1
15058 // CHECK16-NEXT: [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i32*
15059 // CHECK16-NEXT: store i32 [[TMP137]], i32* [[TMP147]], align 4
15060 // CHECK16-NEXT: [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 1
15061 // CHECK16-NEXT: [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
15062 // CHECK16-NEXT: store i32 [[TMP137]], i32* [[TMP149]], align 4
15063 // CHECK16-NEXT: [[TMP150:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 1
15064 // CHECK16-NEXT: store i64 4, i64* [[TMP150]], align 4
15065 // CHECK16-NEXT: [[TMP151:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 1
15066 // CHECK16-NEXT: store i8* null, i8** [[TMP151]], align 4
15067 // CHECK16-NEXT: [[TMP152:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 2
15068 // CHECK16-NEXT: [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i32*
15069 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[TMP153]], align 4
15070 // CHECK16-NEXT: [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 2
15071 // CHECK16-NEXT: [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32*
15072 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[TMP155]], align 4
15073 // CHECK16-NEXT: [[TMP156:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 2
15074 // CHECK16-NEXT: store i64 4, i64* [[TMP156]], align 4
15075 // CHECK16-NEXT: [[TMP157:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 2
15076 // CHECK16-NEXT: store i8* null, i8** [[TMP157]], align 4
15077 // CHECK16-NEXT: [[TMP158:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 3
15078 // CHECK16-NEXT: [[TMP159:%.*]] = bitcast i8** [[TMP158]] to i32**
15079 // CHECK16-NEXT: store i32* [[VLA]], i32** [[TMP159]], align 4
15080 // CHECK16-NEXT: [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 3
15081 // CHECK16-NEXT: [[TMP161:%.*]] = bitcast i8** [[TMP160]] to i32**
15082 // CHECK16-NEXT: store i32* [[VLA]], i32** [[TMP161]], align 4
15083 // CHECK16-NEXT: [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 3
15084 // CHECK16-NEXT: store i64 [[TMP139]], i64* [[TMP162]], align 4
15085 // CHECK16-NEXT: [[TMP163:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 3
15086 // CHECK16-NEXT: store i8* null, i8** [[TMP163]], align 4
15087 // CHECK16-NEXT: [[TMP164:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
15088 // CHECK16-NEXT: [[TMP165:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
15089 // CHECK16-NEXT: [[TMP166:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
15090 // CHECK16-NEXT: [[TMP167:%.*]] = load i32, i32* [[N]], align 4
15091 // CHECK16-NEXT: store i32 [[TMP167]], i32* [[DOTCAPTURE_EXPR_52]], align 4
15092 // CHECK16-NEXT: [[TMP168:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_52]], align 4
15093 // CHECK16-NEXT: [[SUB54:%.*]] = sub nsw i32 [[TMP168]], 0
15094 // CHECK16-NEXT: [[DIV55:%.*]] = sdiv i32 [[SUB54]], 1
15095 // CHECK16-NEXT: [[SUB56:%.*]] = sub nsw i32 [[DIV55]], 1
15096 // CHECK16-NEXT: store i32 [[SUB56]], i32* [[DOTCAPTURE_EXPR_53]], align 4
15097 // CHECK16-NEXT: [[TMP169:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_53]], align 4
15098 // CHECK16-NEXT: [[ADD57:%.*]] = add nsw i32 [[TMP169]], 1
15099 // CHECK16-NEXT: [[TMP170:%.*]] = zext i32 [[ADD57]] to i64
15100 // CHECK16-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP170]])
15101 // CHECK16-NEXT: [[TMP171:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168.region_id, i32 4, i8** [[TMP164]], i8** [[TMP165]], i64* [[TMP166]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
15102 // CHECK16-NEXT: [[TMP172:%.*]] = icmp ne i32 [[TMP171]], 0
15103 // CHECK16-NEXT: br i1 [[TMP172]], label [[OMP_OFFLOAD_FAILED58:%.*]], label [[OMP_OFFLOAD_CONT59:%.*]]
15104 // CHECK16: omp_offload.failed58:
15105 // CHECK16-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168(i32 [[TMP135]], i32 [[TMP137]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
15106 // CHECK16-NEXT: br label [[OMP_OFFLOAD_CONT59]]
15107 // CHECK16: omp_offload.cont59:
15108 // CHECK16-NEXT: [[TMP173:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
15109 // CHECK16-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP173]])
15110 // CHECK16-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
15111 // CHECK16-NEXT: [[TMP174:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
15112 // CHECK16-NEXT: call void @llvm.stackrestore(i8* [[TMP174]])
15113 // CHECK16-NEXT: [[TMP175:%.*]] = load i32, i32* [[RETVAL]], align 4
15114 // CHECK16-NEXT: ret i32 [[TMP175]]
15115 //
15116 //
15117 // CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148
15118 // CHECK16-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
15119 // CHECK16-NEXT: entry:
15120 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
15121 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15122 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15123 // CHECK16-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
15124 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15125 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15126 // CHECK16-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15127 // CHECK16-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15128 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
15129 // CHECK16-NEXT: ret void
15130 //
15131 //
15132 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined.
15133 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
15134 // CHECK16-NEXT: entry:
15135 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15136 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15137 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
15138 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15139 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15140 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
15141 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
15142 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
15143 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
15144 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
15145 // CHECK16-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
15146 // CHECK16-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
15147 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15148 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15149 // CHECK16-NEXT: [[I3:%.*]] = alloca i32, align 4
15150 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15151 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15152 // CHECK16-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
15153 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15154 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15155 // CHECK16-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
15156 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15157 // CHECK16-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15158 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
15159 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
15160 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15161 // CHECK16-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
15162 // CHECK16-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
15163 // CHECK16-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
15164 // CHECK16-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
15165 // CHECK16-NEXT: store i32 0, i32* [[I]], align 4
15166 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15167 // CHECK16-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
15168 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
15169 // CHECK16: omp.precond.then:
15170 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
15171 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15172 // CHECK16-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
15173 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15174 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15175 // CHECK16-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15176 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
15177 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
15178 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15179 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15180 // CHECK16-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
15181 // CHECK16-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
15182 // CHECK16: cond.true:
15183 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15184 // CHECK16-NEXT: br label [[COND_END:%.*]]
15185 // CHECK16: cond.false:
15186 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15187 // CHECK16-NEXT: br label [[COND_END]]
15188 // CHECK16: cond.end:
15189 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
15190 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
15191 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
15192 // CHECK16-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
15193 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
15194 // CHECK16: omp.inner.for.cond:
15195 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15196 // CHECK16-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15197 // CHECK16-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
15198 // CHECK16-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15199 // CHECK16: omp.inner.for.body:
15200 // CHECK16-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
15201 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15202 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
15203 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
15204 // CHECK16: omp.inner.for.inc:
15205 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15206 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
15207 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
15208 // CHECK16-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
15209 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
15210 // CHECK16: omp.inner.for.end:
15211 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
15212 // CHECK16: omp.loop.exit:
15213 // CHECK16-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15214 // CHECK16-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
15215 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
15216 // CHECK16-NEXT: br label [[OMP_PRECOND_END]]
15217 // CHECK16: omp.precond.end:
15218 // CHECK16-NEXT: ret void
15219 //
15220 //
15221 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..1
15222 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
15223 // CHECK16-NEXT: entry:
15224 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15225 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15226 // CHECK16-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
15227 // CHECK16-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
15228 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
15229 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15230 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15231 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
15232 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
15233 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
15234 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
15235 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
15236 // CHECK16-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
15237 // CHECK16-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
15238 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15239 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15240 // CHECK16-NEXT: [[I3:%.*]] = alloca i32, align 4
15241 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15242 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15243 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
15244 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
15245 // CHECK16-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
15246 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15247 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15248 // CHECK16-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
15249 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15250 // CHECK16-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15251 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
15252 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
15253 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15254 // CHECK16-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
15255 // CHECK16-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
15256 // CHECK16-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
15257 // CHECK16-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
15258 // CHECK16-NEXT: store i32 0, i32* [[I]], align 4
15259 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15260 // CHECK16-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
15261 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
15262 // CHECK16: omp.precond.then:
15263 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
15264 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15265 // CHECK16-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
15266 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
15267 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
15268 // CHECK16-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
15269 // CHECK16-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
15270 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15271 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15272 // CHECK16-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15273 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
15274 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
15275 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15276 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15277 // CHECK16-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
15278 // CHECK16-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
15279 // CHECK16: cond.true:
15280 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15281 // CHECK16-NEXT: br label [[COND_END:%.*]]
15282 // CHECK16: cond.false:
15283 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15284 // CHECK16-NEXT: br label [[COND_END]]
15285 // CHECK16: cond.end:
15286 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
15287 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
15288 // CHECK16-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
15289 // CHECK16-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
15290 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
15291 // CHECK16: omp.inner.for.cond:
15292 // CHECK16-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15293 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15294 // CHECK16-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
15295 // CHECK16-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15296 // CHECK16: omp.inner.for.body:
15297 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15298 // CHECK16-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
15299 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
15300 // CHECK16-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
15301 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
15302 // CHECK16-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
15303 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
15304 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
15305 // CHECK16: omp.body.continue:
15306 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
15307 // CHECK16: omp.inner.for.inc:
15308 // CHECK16-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15309 // CHECK16-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
15310 // CHECK16-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
15311 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
15312 // CHECK16: omp.inner.for.end:
15313 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
15314 // CHECK16: omp.loop.exit:
15315 // CHECK16-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15316 // CHECK16-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
15317 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
15318 // CHECK16-NEXT: br label [[OMP_PRECOND_END]]
15319 // CHECK16: omp.precond.end:
15320 // CHECK16-NEXT: ret void
15321 //
15322 //
15323 // CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153
15324 // CHECK16-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
15325 // CHECK16-NEXT: entry:
15326 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
15327 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15328 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15329 // CHECK16-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
15330 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15331 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15332 // CHECK16-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15333 // CHECK16-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15334 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
15335 // CHECK16-NEXT: ret void
15336 //
15337 //
15338 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..2
15339 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
15340 // CHECK16-NEXT: entry:
15341 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15342 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15343 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
15344 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15345 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15346 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
15347 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
15348 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
15349 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
15350 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
15351 // CHECK16-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
15352 // CHECK16-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
15353 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15354 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15355 // CHECK16-NEXT: [[I3:%.*]] = alloca i32, align 4
15356 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15357 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15358 // CHECK16-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
15359 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15360 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15361 // CHECK16-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
15362 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15363 // CHECK16-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15364 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
15365 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
15366 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15367 // CHECK16-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
15368 // CHECK16-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
15369 // CHECK16-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
15370 // CHECK16-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
15371 // CHECK16-NEXT: store i32 0, i32* [[I]], align 4
15372 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15373 // CHECK16-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
15374 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
15375 // CHECK16: omp.precond.then:
15376 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
15377 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15378 // CHECK16-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
15379 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15380 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15381 // CHECK16-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15382 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
15383 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
15384 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15385 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15386 // CHECK16-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
15387 // CHECK16-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
15388 // CHECK16: cond.true:
15389 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15390 // CHECK16-NEXT: br label [[COND_END:%.*]]
15391 // CHECK16: cond.false:
15392 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15393 // CHECK16-NEXT: br label [[COND_END]]
15394 // CHECK16: cond.end:
15395 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
15396 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
15397 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
15398 // CHECK16-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
15399 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
15400 // CHECK16: omp.inner.for.cond:
15401 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15402 // CHECK16-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15403 // CHECK16-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
15404 // CHECK16-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15405 // CHECK16: omp.inner.for.body:
15406 // CHECK16-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
15407 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15408 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
15409 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
15410 // CHECK16: omp.inner.for.inc:
15411 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15412 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
15413 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
15414 // CHECK16-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
15415 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
15416 // CHECK16: omp.inner.for.end:
15417 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
15418 // CHECK16: omp.loop.exit:
15419 // CHECK16-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15420 // CHECK16-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
15421 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
15422 // CHECK16-NEXT: br label [[OMP_PRECOND_END]]
15423 // CHECK16: omp.precond.end:
15424 // CHECK16-NEXT: ret void
15425 //
15426 //
15427 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..3
15428 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
15429 // CHECK16-NEXT: entry:
15430 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15431 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15432 // CHECK16-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
15433 // CHECK16-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
15434 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
15435 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15436 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15437 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
15438 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
15439 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
15440 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
15441 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
15442 // CHECK16-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
15443 // CHECK16-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
15444 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15445 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15446 // CHECK16-NEXT: [[I3:%.*]] = alloca i32, align 4
15447 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15448 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15449 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
15450 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
15451 // CHECK16-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
15452 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15453 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15454 // CHECK16-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
15455 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15456 // CHECK16-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15457 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
15458 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
15459 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15460 // CHECK16-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
15461 // CHECK16-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
15462 // CHECK16-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
15463 // CHECK16-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
15464 // CHECK16-NEXT: store i32 0, i32* [[I]], align 4
15465 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15466 // CHECK16-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
15467 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
15468 // CHECK16: omp.precond.then:
15469 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
15470 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15471 // CHECK16-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
15472 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
15473 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
15474 // CHECK16-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
15475 // CHECK16-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
15476 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15477 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15478 // CHECK16-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15479 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
15480 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
15481 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15482 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15483 // CHECK16-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
15484 // CHECK16-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
15485 // CHECK16: cond.true:
15486 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15487 // CHECK16-NEXT: br label [[COND_END:%.*]]
15488 // CHECK16: cond.false:
15489 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15490 // CHECK16-NEXT: br label [[COND_END]]
15491 // CHECK16: cond.end:
15492 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
15493 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
15494 // CHECK16-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
15495 // CHECK16-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
15496 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
15497 // CHECK16: omp.inner.for.cond:
15498 // CHECK16-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15499 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15500 // CHECK16-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
15501 // CHECK16-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15502 // CHECK16: omp.inner.for.body:
15503 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15504 // CHECK16-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
15505 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
15506 // CHECK16-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
15507 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
15508 // CHECK16-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
15509 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
15510 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
15511 // CHECK16: omp.body.continue:
15512 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
15513 // CHECK16: omp.inner.for.inc:
15514 // CHECK16-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15515 // CHECK16-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
15516 // CHECK16-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
15517 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
15518 // CHECK16: omp.inner.for.end:
15519 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
15520 // CHECK16: omp.loop.exit:
15521 // CHECK16-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15522 // CHECK16-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
15523 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
15524 // CHECK16-NEXT: br label [[OMP_PRECOND_END]]
15525 // CHECK16: omp.precond.end:
15526 // CHECK16-NEXT: ret void
15527 //
15528 //
15529 // CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158
15530 // CHECK16-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
15531 // CHECK16-NEXT: entry:
15532 // CHECK16-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
15533 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
15534 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15535 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15536 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
15537 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
15538 // CHECK16-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
15539 // CHECK16-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
15540 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15541 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15542 // CHECK16-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15543 // CHECK16-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15544 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
15545 // CHECK16-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
15546 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15547 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
15548 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
15549 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
15550 // CHECK16-NEXT: ret void
15551 //
15552 //
15553 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..5
15554 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
15555 // CHECK16-NEXT: entry:
15556 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15557 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15558 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
15559 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15560 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15561 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
15562 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
15563 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
15564 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
15565 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
15566 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
15567 // CHECK16-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
15568 // CHECK16-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
15569 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15570 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15571 // CHECK16-NEXT: [[I4:%.*]] = alloca i32, align 4
15572 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
15573 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15574 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15575 // CHECK16-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
15576 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15577 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15578 // CHECK16-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
15579 // CHECK16-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
15580 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15581 // CHECK16-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15582 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
15583 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
15584 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15585 // CHECK16-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
15586 // CHECK16-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
15587 // CHECK16-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
15588 // CHECK16-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
15589 // CHECK16-NEXT: store i32 0, i32* [[I]], align 4
15590 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15591 // CHECK16-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
15592 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
15593 // CHECK16: omp.precond.then:
15594 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
15595 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15596 // CHECK16-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
15597 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15598 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15599 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
15600 // CHECK16-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15601 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
15602 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
15603 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15604 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15605 // CHECK16-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
15606 // CHECK16-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
15607 // CHECK16: cond.true:
15608 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15609 // CHECK16-NEXT: br label [[COND_END:%.*]]
15610 // CHECK16: cond.false:
15611 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15612 // CHECK16-NEXT: br label [[COND_END]]
15613 // CHECK16: cond.end:
15614 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
15615 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
15616 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
15617 // CHECK16-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
15618 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
15619 // CHECK16: omp.inner.for.cond:
15620 // CHECK16-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15621 // CHECK16-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15622 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
15623 // CHECK16-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
15624 // CHECK16-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15625 // CHECK16: omp.inner.for.body:
15626 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
15627 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15628 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
15629 // CHECK16-NEXT: store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
15630 // CHECK16-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
15631 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]])
15632 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
15633 // CHECK16: omp.inner.for.inc:
15634 // CHECK16-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15635 // CHECK16-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
15636 // CHECK16-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
15637 // CHECK16-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
15638 // CHECK16-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
15639 // CHECK16-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
15640 // CHECK16-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
15641 // CHECK16-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4
15642 // CHECK16-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15643 // CHECK16-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
15644 // CHECK16-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
15645 // CHECK16-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4
15646 // CHECK16-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15647 // CHECK16-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15648 // CHECK16-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
15649 // CHECK16-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
15650 // CHECK16: cond.true11:
15651 // CHECK16-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15652 // CHECK16-NEXT: br label [[COND_END13:%.*]]
15653 // CHECK16: cond.false12:
15654 // CHECK16-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15655 // CHECK16-NEXT: br label [[COND_END13]]
15656 // CHECK16: cond.end13:
15657 // CHECK16-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
15658 // CHECK16-NEXT: store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4
15659 // CHECK16-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
15660 // CHECK16-NEXT: store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
15661 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
15662 // CHECK16: omp.inner.for.end:
15663 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
15664 // CHECK16: omp.loop.exit:
15665 // CHECK16-NEXT: [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15666 // CHECK16-NEXT: [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
15667 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
15668 // CHECK16-NEXT: br label [[OMP_PRECOND_END]]
15669 // CHECK16: omp.precond.end:
15670 // CHECK16-NEXT: ret void
15671 //
15672 //
15673 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..6
15674 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
15675 // CHECK16-NEXT: entry:
15676 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15677 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15678 // CHECK16-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
15679 // CHECK16-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
15680 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
15681 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15682 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15683 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
15684 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
15685 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
15686 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
15687 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
15688 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
15689 // CHECK16-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
15690 // CHECK16-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
15691 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15692 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15693 // CHECK16-NEXT: [[I4:%.*]] = alloca i32, align 4
15694 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15695 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15696 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
15697 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
15698 // CHECK16-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
15699 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15700 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15701 // CHECK16-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
15702 // CHECK16-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
15703 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15704 // CHECK16-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15705 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
15706 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
15707 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15708 // CHECK16-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
15709 // CHECK16-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
15710 // CHECK16-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
15711 // CHECK16-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
15712 // CHECK16-NEXT: store i32 0, i32* [[I]], align 4
15713 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15714 // CHECK16-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
15715 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
15716 // CHECK16: omp.precond.then:
15717 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
15718 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15719 // CHECK16-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
15720 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
15721 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
15722 // CHECK16-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
15723 // CHECK16-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
15724 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15725 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15726 // CHECK16-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15727 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
15728 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
15729 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15730 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15731 // CHECK16-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
15732 // CHECK16-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
15733 // CHECK16: cond.true:
15734 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
15735 // CHECK16-NEXT: br label [[COND_END:%.*]]
15736 // CHECK16: cond.false:
15737 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15738 // CHECK16-NEXT: br label [[COND_END]]
15739 // CHECK16: cond.end:
15740 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
15741 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
15742 // CHECK16-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
15743 // CHECK16-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
15744 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
15745 // CHECK16: omp.inner.for.cond:
15746 // CHECK16-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15747 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15748 // CHECK16-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
15749 // CHECK16-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15750 // CHECK16: omp.inner.for.body:
15751 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15752 // CHECK16-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
15753 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
15754 // CHECK16-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
15755 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
15756 // CHECK16-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
15757 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
15758 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
15759 // CHECK16: omp.body.continue:
15760 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
15761 // CHECK16: omp.inner.for.inc:
15762 // CHECK16-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15763 // CHECK16-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
15764 // CHECK16-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
15765 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
15766 // CHECK16: omp.inner.for.end:
15767 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
15768 // CHECK16: omp.loop.exit:
15769 // CHECK16-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15770 // CHECK16-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
15771 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
15772 // CHECK16-NEXT: br label [[OMP_PRECOND_END]]
15773 // CHECK16: omp.precond.end:
15774 // CHECK16-NEXT: ret void
15775 //
15776 //
15777 // CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163
15778 // CHECK16-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
15779 // CHECK16-NEXT: entry:
15780 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
15781 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15782 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15783 // CHECK16-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
15784 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15785 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15786 // CHECK16-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15787 // CHECK16-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15788 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
15789 // CHECK16-NEXT: ret void
15790 //
15791 //
15792 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..8
15793 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
15794 // CHECK16-NEXT: entry:
15795 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15796 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15797 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
15798 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15799 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15800 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
15801 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
15802 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
15803 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
15804 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
15805 // CHECK16-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
15806 // CHECK16-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
15807 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15808 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15809 // CHECK16-NEXT: [[I3:%.*]] = alloca i32, align 4
15810 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15811 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15812 // CHECK16-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
15813 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15814 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15815 // CHECK16-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
15816 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15817 // CHECK16-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15818 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
15819 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
15820 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15821 // CHECK16-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
15822 // CHECK16-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
15823 // CHECK16-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
15824 // CHECK16-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
15825 // CHECK16-NEXT: store i32 0, i32* [[I]], align 4
15826 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15827 // CHECK16-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
15828 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
15829 // CHECK16: omp.precond.then:
15830 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
15831 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15832 // CHECK16-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
15833 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15834 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15835 // CHECK16-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15836 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
15837 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
15838 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15839 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15840 // CHECK16-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
15841 // CHECK16-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
15842 // CHECK16: cond.true:
15843 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15844 // CHECK16-NEXT: br label [[COND_END:%.*]]
15845 // CHECK16: cond.false:
15846 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15847 // CHECK16-NEXT: br label [[COND_END]]
15848 // CHECK16: cond.end:
15849 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
15850 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
15851 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
15852 // CHECK16-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
15853 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
15854 // CHECK16: omp.inner.for.cond:
15855 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15856 // CHECK16-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15857 // CHECK16-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
15858 // CHECK16-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15859 // CHECK16: omp.inner.for.body:
15860 // CHECK16-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
15861 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
15862 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
15863 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
15864 // CHECK16: omp.inner.for.inc:
15865 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
15866 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
15867 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
15868 // CHECK16-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
15869 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
15870 // CHECK16: omp.inner.for.end:
15871 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
15872 // CHECK16: omp.loop.exit:
15873 // CHECK16-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15874 // CHECK16-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
15875 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
15876 // CHECK16-NEXT: br label [[OMP_PRECOND_END]]
15877 // CHECK16: omp.precond.end:
15878 // CHECK16-NEXT: ret void
15879 //
15880 //
15881 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..9
15882 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
15883 // CHECK16-NEXT: entry:
15884 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
15885 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
15886 // CHECK16-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
15887 // CHECK16-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
15888 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
15889 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15890 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15891 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
15892 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
15893 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
15894 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
15895 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
15896 // CHECK16-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
15897 // CHECK16-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
15898 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
15899 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
15900 // CHECK16-NEXT: [[I3:%.*]] = alloca i32, align 4
15901 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
15902 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
15903 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
15904 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
15905 // CHECK16-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
15906 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15907 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15908 // CHECK16-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
15909 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15910 // CHECK16-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15911 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
15912 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
15913 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15914 // CHECK16-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
15915 // CHECK16-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
15916 // CHECK16-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
15917 // CHECK16-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
15918 // CHECK16-NEXT: store i32 0, i32* [[I]], align 4
15919 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15920 // CHECK16-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
15921 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
15922 // CHECK16: omp.precond.then:
15923 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
15924 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
15925 // CHECK16-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
15926 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
15927 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
15928 // CHECK16-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
15929 // CHECK16-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
15930 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
15931 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
15932 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
15933 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
15934 // CHECK16-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15935 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
15936 // CHECK16-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 1073741859, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
15937 // CHECK16-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
15938 // CHECK16: omp.dispatch.cond:
15939 // CHECK16-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
15940 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
15941 // CHECK16-NEXT: [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
15942 // CHECK16-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
15943 // CHECK16-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
15944 // CHECK16: omp.dispatch.body:
15945 // CHECK16-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
15946 // CHECK16-NEXT: store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
15947 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
15948 // CHECK16: omp.inner.for.cond:
15949 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
15950 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !16
15951 // CHECK16-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
15952 // CHECK16-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
15953 // CHECK16: omp.inner.for.body:
15954 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
15955 // CHECK16-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
15956 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
15957 // CHECK16-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !16
15958 // CHECK16-NEXT: [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !16
15959 // CHECK16-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
15960 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !16
15961 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
15962 // CHECK16: omp.body.continue:
15963 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
15964 // CHECK16: omp.inner.for.inc:
15965 // CHECK16-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
15966 // CHECK16-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP21]], 1
15967 // CHECK16-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
15968 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
15969 // CHECK16: omp.inner.for.end:
15970 // CHECK16-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
15971 // CHECK16: omp.dispatch.inc:
15972 // CHECK16-NEXT: br label [[OMP_DISPATCH_COND]]
15973 // CHECK16: omp.dispatch.end:
15974 // CHECK16-NEXT: br label [[OMP_PRECOND_END]]
15975 // CHECK16: omp.precond.end:
15976 // CHECK16-NEXT: ret void
15977 //
15978 //
15979 // CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168
15980 // CHECK16-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
15981 // CHECK16-NEXT: entry:
15982 // CHECK16-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
15983 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
15984 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
15985 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
15986 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
15987 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
15988 // CHECK16-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
15989 // CHECK16-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
15990 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
15991 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
15992 // CHECK16-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
15993 // CHECK16-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
15994 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
15995 // CHECK16-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
15996 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
15997 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
15998 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
15999 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
16000 // CHECK16-NEXT: ret void
16001 //
16002 //
16003 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..11
16004 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
16005 // CHECK16-NEXT: entry:
16006 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16007 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16008 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
16009 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
16010 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
16011 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
16012 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
16013 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
16014 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
16015 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
16016 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
16017 // CHECK16-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
16018 // CHECK16-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
16019 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16020 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16021 // CHECK16-NEXT: [[I4:%.*]] = alloca i32, align 4
16022 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
16023 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16024 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16025 // CHECK16-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
16026 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
16027 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
16028 // CHECK16-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16029 // CHECK16-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
16030 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
16031 // CHECK16-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
16032 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
16033 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
16034 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
16035 // CHECK16-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
16036 // CHECK16-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
16037 // CHECK16-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
16038 // CHECK16-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
16039 // CHECK16-NEXT: store i32 0, i32* [[I]], align 4
16040 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
16041 // CHECK16-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
16042 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
16043 // CHECK16: omp.precond.then:
16044 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
16045 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
16046 // CHECK16-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
16047 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16048 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16049 // CHECK16-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16050 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
16051 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
16052 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16053 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
16054 // CHECK16-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
16055 // CHECK16-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16056 // CHECK16: cond.true:
16057 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
16058 // CHECK16-NEXT: br label [[COND_END:%.*]]
16059 // CHECK16: cond.false:
16060 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16061 // CHECK16-NEXT: br label [[COND_END]]
16062 // CHECK16: cond.end:
16063 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
16064 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
16065 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
16066 // CHECK16-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
16067 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
16068 // CHECK16: omp.inner.for.cond:
16069 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16070 // CHECK16-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16071 // CHECK16-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
16072 // CHECK16-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16073 // CHECK16: omp.inner.for.body:
16074 // CHECK16-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
16075 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16076 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16077 // CHECK16-NEXT: store i32 [[TMP18]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
16078 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
16079 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP19]])
16080 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
16081 // CHECK16: omp.inner.for.inc:
16082 // CHECK16-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16083 // CHECK16-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
16084 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
16085 // CHECK16-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
16086 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
16087 // CHECK16: omp.inner.for.end:
16088 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
16089 // CHECK16: omp.loop.exit:
16090 // CHECK16-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16091 // CHECK16-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
16092 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
16093 // CHECK16-NEXT: br label [[OMP_PRECOND_END]]
16094 // CHECK16: omp.precond.end:
16095 // CHECK16-NEXT: ret void
16096 //
16097 //
16098 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..12
16099 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
16100 // CHECK16-NEXT: entry:
16101 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16102 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16103 // CHECK16-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
16104 // CHECK16-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
16105 // CHECK16-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
16106 // CHECK16-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
16107 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
16108 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
16109 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
16110 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
16111 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
16112 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
16113 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
16114 // CHECK16-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
16115 // CHECK16-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
16116 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16117 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16118 // CHECK16-NEXT: [[I4:%.*]] = alloca i32, align 4
16119 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16120 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16121 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
16122 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
16123 // CHECK16-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
16124 // CHECK16-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
16125 // CHECK16-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
16126 // CHECK16-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16127 // CHECK16-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
16128 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
16129 // CHECK16-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
16130 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
16131 // CHECK16-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
16132 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
16133 // CHECK16-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
16134 // CHECK16-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
16135 // CHECK16-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
16136 // CHECK16-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
16137 // CHECK16-NEXT: store i32 0, i32* [[I]], align 4
16138 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
16139 // CHECK16-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
16140 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
16141 // CHECK16: omp.precond.then:
16142 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
16143 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
16144 // CHECK16-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
16145 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
16146 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
16147 // CHECK16-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
16148 // CHECK16-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
16149 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16150 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16151 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16152 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16153 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16154 // CHECK16-NEXT: [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16155 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
16156 // CHECK16-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 1073741859, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
16157 // CHECK16-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
16158 // CHECK16: omp.dispatch.cond:
16159 // CHECK16-NEXT: [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16160 // CHECK16-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
16161 // CHECK16-NEXT: [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
16162 // CHECK16-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
16163 // CHECK16-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
16164 // CHECK16: omp.dispatch.body:
16165 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16166 // CHECK16-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
16167 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
16168 // CHECK16: omp.inner.for.cond:
16169 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
16170 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
16171 // CHECK16-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
16172 // CHECK16-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16173 // CHECK16: omp.inner.for.body:
16174 // CHECK16-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
16175 // CHECK16-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
16176 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
16177 // CHECK16-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !19
16178 // CHECK16-NEXT: [[TMP21:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !19
16179 // CHECK16-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP21]]
16180 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !19
16181 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
16182 // CHECK16: omp.body.continue:
16183 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
16184 // CHECK16: omp.inner.for.inc:
16185 // CHECK16-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
16186 // CHECK16-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
16187 // CHECK16-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
16188 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
16189 // CHECK16: omp.inner.for.end:
16190 // CHECK16-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
16191 // CHECK16: omp.dispatch.inc:
16192 // CHECK16-NEXT: br label [[OMP_DISPATCH_COND]]
16193 // CHECK16: omp.dispatch.end:
16194 // CHECK16-NEXT: br label [[OMP_PRECOND_END]]
16195 // CHECK16: omp.precond.end:
16196 // CHECK16-NEXT: ret void
16197 //
16198 //
16199 // CHECK16-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
16200 // CHECK16-SAME: (i32 [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
16201 // CHECK16-NEXT: entry:
16202 // CHECK16-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
16203 // CHECK16-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
16204 // CHECK16-NEXT: [[M:%.*]] = alloca i32, align 4
16205 // CHECK16-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
16206 // CHECK16-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
16207 // CHECK16-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
16208 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
16209 // CHECK16-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
16210 // CHECK16-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
16211 // CHECK16-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
16212 // CHECK16-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
16213 // CHECK16-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
16214 // CHECK16-NEXT: [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 4
16215 // CHECK16-NEXT: [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 4
16216 // CHECK16-NEXT: [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 4
16217 // CHECK16-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
16218 // CHECK16-NEXT: [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 4
16219 // CHECK16-NEXT: [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 4
16220 // CHECK16-NEXT: [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 4
16221 // CHECK16-NEXT: [[_TMP16:%.*]] = alloca i32, align 4
16222 // CHECK16-NEXT: [[M_CASTED19:%.*]] = alloca i32, align 4
16223 // CHECK16-NEXT: [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [2 x i8*], align 4
16224 // CHECK16-NEXT: [[DOTOFFLOAD_PTRS21:%.*]] = alloca [2 x i8*], align 4
16225 // CHECK16-NEXT: [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [2 x i8*], align 4
16226 // CHECK16-NEXT: [[_TMP23:%.*]] = alloca i32, align 4
16227 // CHECK16-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
16228 // CHECK16-NEXT: store i32 10, i32* [[M]], align 4
16229 // CHECK16-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
16230 // CHECK16-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
16231 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
16232 // CHECK16-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
16233 // CHECK16-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
16234 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
16235 // CHECK16-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
16236 // CHECK16-NEXT: store i8* null, i8** [[TMP4]], align 4
16237 // CHECK16-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
16238 // CHECK16-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
16239 // CHECK16-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
16240 // CHECK16-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
16241 // CHECK16-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
16242 // CHECK16-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
16243 // CHECK16: omp_offload.failed:
16244 // CHECK16-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116([10 x i32]* [[A]]) #[[ATTR3]]
16245 // CHECK16-NEXT: br label [[OMP_OFFLOAD_CONT]]
16246 // CHECK16: omp_offload.cont:
16247 // CHECK16-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
16248 // CHECK16-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
16249 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 4
16250 // CHECK16-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
16251 // CHECK16-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
16252 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 4
16253 // CHECK16-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
16254 // CHECK16-NEXT: store i8* null, i8** [[TMP13]], align 4
16255 // CHECK16-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
16256 // CHECK16-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
16257 // CHECK16-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
16258 // CHECK16-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
16259 // CHECK16-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
16260 // CHECK16-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
16261 // CHECK16: omp_offload.failed5:
16262 // CHECK16-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121([10 x i32]* [[A]]) #[[ATTR3]]
16263 // CHECK16-NEXT: br label [[OMP_OFFLOAD_CONT6]]
16264 // CHECK16: omp_offload.cont6:
16265 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[M]], align 4
16266 // CHECK16-NEXT: store i32 [[TMP18]], i32* [[M_CASTED]], align 4
16267 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[M_CASTED]], align 4
16268 // CHECK16-NEXT: [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
16269 // CHECK16-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
16270 // CHECK16-NEXT: store i32 [[TMP19]], i32* [[TMP21]], align 4
16271 // CHECK16-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
16272 // CHECK16-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
16273 // CHECK16-NEXT: store i32 [[TMP19]], i32* [[TMP23]], align 4
16274 // CHECK16-NEXT: [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 0
16275 // CHECK16-NEXT: store i8* null, i8** [[TMP24]], align 4
16276 // CHECK16-NEXT: [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
16277 // CHECK16-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
16278 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 4
16279 // CHECK16-NEXT: [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
16280 // CHECK16-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
16281 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 4
16282 // CHECK16-NEXT: [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 1
16283 // CHECK16-NEXT: store i8* null, i8** [[TMP29]], align 4
16284 // CHECK16-NEXT: [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
16285 // CHECK16-NEXT: [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
16286 // CHECK16-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
16287 // CHECK16-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
16288 // CHECK16-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
16289 // CHECK16-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
16290 // CHECK16: omp_offload.failed11:
16291 // CHECK16-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126(i32 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
16292 // CHECK16-NEXT: br label [[OMP_OFFLOAD_CONT12]]
16293 // CHECK16: omp_offload.cont12:
16294 // CHECK16-NEXT: [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
16295 // CHECK16-NEXT: [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
16296 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 4
16297 // CHECK16-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
16298 // CHECK16-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
16299 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 4
16300 // CHECK16-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
16301 // CHECK16-NEXT: store i8* null, i8** [[TMP38]], align 4
16302 // CHECK16-NEXT: [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
16303 // CHECK16-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
16304 // CHECK16-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
16305 // CHECK16-NEXT: [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
16306 // CHECK16-NEXT: [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
16307 // CHECK16-NEXT: br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
16308 // CHECK16: omp_offload.failed17:
16309 // CHECK16-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131([10 x i32]* [[A]]) #[[ATTR3]]
16310 // CHECK16-NEXT: br label [[OMP_OFFLOAD_CONT18]]
16311 // CHECK16: omp_offload.cont18:
16312 // CHECK16-NEXT: [[TMP43:%.*]] = load i32, i32* [[M]], align 4
16313 // CHECK16-NEXT: store i32 [[TMP43]], i32* [[M_CASTED19]], align 4
16314 // CHECK16-NEXT: [[TMP44:%.*]] = load i32, i32* [[M_CASTED19]], align 4
16315 // CHECK16-NEXT: [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
16316 // CHECK16-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
16317 // CHECK16-NEXT: store i32 [[TMP44]], i32* [[TMP46]], align 4
16318 // CHECK16-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
16319 // CHECK16-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i32*
16320 // CHECK16-NEXT: store i32 [[TMP44]], i32* [[TMP48]], align 4
16321 // CHECK16-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
16322 // CHECK16-NEXT: store i8* null, i8** [[TMP49]], align 4
16323 // CHECK16-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
16324 // CHECK16-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
16325 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 4
16326 // CHECK16-NEXT: [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
16327 // CHECK16-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
16328 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 4
16329 // CHECK16-NEXT: [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
16330 // CHECK16-NEXT: store i8* null, i8** [[TMP54]], align 4
16331 // CHECK16-NEXT: [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
16332 // CHECK16-NEXT: [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
16333 // CHECK16-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
16334 // CHECK16-NEXT: [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
16335 // CHECK16-NEXT: [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
16336 // CHECK16-NEXT: br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
16337 // CHECK16: omp_offload.failed24:
16338 // CHECK16-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136(i32 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
16339 // CHECK16-NEXT: br label [[OMP_OFFLOAD_CONT25]]
16340 // CHECK16: omp_offload.cont25:
16341 // CHECK16-NEXT: ret i32 0
16342 //
16343 //
16344 // CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116
16345 // CHECK16-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
16346 // CHECK16-NEXT: entry:
16347 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16348 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16349 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16350 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
16351 // CHECK16-NEXT: ret void
16352 //
16353 //
16354 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..14
16355 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
16356 // CHECK16-NEXT: entry:
16357 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16358 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16359 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16360 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
16361 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
16362 // CHECK16-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
16363 // CHECK16-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
16364 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16365 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16366 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
16367 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16368 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16369 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16370 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16371 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
16372 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
16373 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16374 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16375 // CHECK16-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16376 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
16377 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
16378 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16379 // CHECK16-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
16380 // CHECK16-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16381 // CHECK16: cond.true:
16382 // CHECK16-NEXT: br label [[COND_END:%.*]]
16383 // CHECK16: cond.false:
16384 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16385 // CHECK16-NEXT: br label [[COND_END]]
16386 // CHECK16: cond.end:
16387 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
16388 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
16389 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
16390 // CHECK16-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
16391 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
16392 // CHECK16: omp.inner.for.cond:
16393 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16394 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16395 // CHECK16-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
16396 // CHECK16-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16397 // CHECK16: omp.inner.for.body:
16398 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
16399 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16400 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
16401 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
16402 // CHECK16: omp.inner.for.inc:
16403 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16404 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
16405 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
16406 // CHECK16-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
16407 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
16408 // CHECK16: omp.inner.for.end:
16409 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
16410 // CHECK16: omp.loop.exit:
16411 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
16412 // CHECK16-NEXT: ret void
16413 //
16414 //
16415 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..15
16416 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
16417 // CHECK16-NEXT: entry:
16418 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16419 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16420 // CHECK16-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
16421 // CHECK16-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
16422 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16423 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
16424 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
16425 // CHECK16-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
16426 // CHECK16-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
16427 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16428 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16429 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
16430 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16431 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16432 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
16433 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
16434 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16435 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16436 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
16437 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
16438 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
16439 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
16440 // CHECK16-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
16441 // CHECK16-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
16442 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16443 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16444 // CHECK16-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16445 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
16446 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
16447 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16448 // CHECK16-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
16449 // CHECK16-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16450 // CHECK16: cond.true:
16451 // CHECK16-NEXT: br label [[COND_END:%.*]]
16452 // CHECK16: cond.false:
16453 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16454 // CHECK16-NEXT: br label [[COND_END]]
16455 // CHECK16: cond.end:
16456 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
16457 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
16458 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16459 // CHECK16-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
16460 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
16461 // CHECK16: omp.inner.for.cond:
16462 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16463 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16464 // CHECK16-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
16465 // CHECK16-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16466 // CHECK16: omp.inner.for.body:
16467 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16468 // CHECK16-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
16469 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
16470 // CHECK16-NEXT: store i32 [[ADD]], i32* [[I]], align 4
16471 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
16472 // CHECK16-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
16473 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
16474 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
16475 // CHECK16: omp.body.continue:
16476 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
16477 // CHECK16: omp.inner.for.inc:
16478 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16479 // CHECK16-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
16480 // CHECK16-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
16481 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
16482 // CHECK16: omp.inner.for.end:
16483 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
16484 // CHECK16: omp.loop.exit:
16485 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
16486 // CHECK16-NEXT: ret void
16487 //
16488 //
16489 // CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121
16490 // CHECK16-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
16491 // CHECK16-NEXT: entry:
16492 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16493 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16494 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16495 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
16496 // CHECK16-NEXT: ret void
16497 //
16498 //
16499 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..17
16500 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
16501 // CHECK16-NEXT: entry:
16502 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16503 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16504 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16505 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
16506 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
16507 // CHECK16-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
16508 // CHECK16-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
16509 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16510 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16511 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
16512 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16513 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16514 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16515 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16516 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
16517 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
16518 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16519 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16520 // CHECK16-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16521 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
16522 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
16523 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16524 // CHECK16-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
16525 // CHECK16-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16526 // CHECK16: cond.true:
16527 // CHECK16-NEXT: br label [[COND_END:%.*]]
16528 // CHECK16: cond.false:
16529 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16530 // CHECK16-NEXT: br label [[COND_END]]
16531 // CHECK16: cond.end:
16532 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
16533 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
16534 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
16535 // CHECK16-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
16536 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
16537 // CHECK16: omp.inner.for.cond:
16538 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16539 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16540 // CHECK16-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
16541 // CHECK16-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16542 // CHECK16: omp.inner.for.body:
16543 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
16544 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16545 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
16546 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
16547 // CHECK16: omp.inner.for.inc:
16548 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16549 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
16550 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
16551 // CHECK16-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
16552 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
16553 // CHECK16: omp.inner.for.end:
16554 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
16555 // CHECK16: omp.loop.exit:
16556 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
16557 // CHECK16-NEXT: ret void
16558 //
16559 //
16560 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..18
16561 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
16562 // CHECK16-NEXT: entry:
16563 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16564 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16565 // CHECK16-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
16566 // CHECK16-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
16567 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16568 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
16569 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
16570 // CHECK16-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
16571 // CHECK16-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
16572 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16573 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16574 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
16575 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16576 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16577 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
16578 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
16579 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16580 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16581 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
16582 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
16583 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
16584 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
16585 // CHECK16-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
16586 // CHECK16-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
16587 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16588 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16589 // CHECK16-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16590 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
16591 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
16592 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16593 // CHECK16-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
16594 // CHECK16-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16595 // CHECK16: cond.true:
16596 // CHECK16-NEXT: br label [[COND_END:%.*]]
16597 // CHECK16: cond.false:
16598 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16599 // CHECK16-NEXT: br label [[COND_END]]
16600 // CHECK16: cond.end:
16601 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
16602 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
16603 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16604 // CHECK16-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
16605 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
16606 // CHECK16: omp.inner.for.cond:
16607 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16608 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16609 // CHECK16-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
16610 // CHECK16-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16611 // CHECK16: omp.inner.for.body:
16612 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16613 // CHECK16-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
16614 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
16615 // CHECK16-NEXT: store i32 [[ADD]], i32* [[I]], align 4
16616 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
16617 // CHECK16-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
16618 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
16619 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
16620 // CHECK16: omp.body.continue:
16621 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
16622 // CHECK16: omp.inner.for.inc:
16623 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16624 // CHECK16-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
16625 // CHECK16-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
16626 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
16627 // CHECK16: omp.inner.for.end:
16628 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
16629 // CHECK16: omp.loop.exit:
16630 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
16631 // CHECK16-NEXT: ret void
16632 //
16633 //
16634 // CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126
16635 // CHECK16-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
16636 // CHECK16-NEXT: entry:
16637 // CHECK16-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
16638 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16639 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
16640 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
16641 // CHECK16-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
16642 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16643 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16644 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
16645 // CHECK16-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
16646 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
16647 // CHECK16-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
16648 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
16649 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
16650 // CHECK16-NEXT: ret void
16651 //
16652 //
16653 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..21
16654 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
16655 // CHECK16-NEXT: entry:
16656 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16657 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16658 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16659 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
16660 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
16661 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
16662 // CHECK16-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
16663 // CHECK16-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
16664 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16665 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16666 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
16667 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
16668 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16669 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16670 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16671 // CHECK16-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16672 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16673 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
16674 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
16675 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16676 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16677 // CHECK16-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16678 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
16679 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
16680 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16681 // CHECK16-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
16682 // CHECK16-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16683 // CHECK16: cond.true:
16684 // CHECK16-NEXT: br label [[COND_END:%.*]]
16685 // CHECK16: cond.false:
16686 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16687 // CHECK16-NEXT: br label [[COND_END]]
16688 // CHECK16: cond.end:
16689 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
16690 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
16691 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
16692 // CHECK16-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
16693 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
16694 // CHECK16: omp.inner.for.cond:
16695 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16696 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16697 // CHECK16-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
16698 // CHECK16-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16699 // CHECK16: omp.inner.for.body:
16700 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
16701 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16702 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16703 // CHECK16-NEXT: store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
16704 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
16705 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
16706 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
16707 // CHECK16: omp.inner.for.inc:
16708 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16709 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
16710 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
16711 // CHECK16-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
16712 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
16713 // CHECK16: omp.inner.for.end:
16714 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
16715 // CHECK16: omp.loop.exit:
16716 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
16717 // CHECK16-NEXT: ret void
16718 //
16719 //
16720 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..22
16721 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
16722 // CHECK16-NEXT: entry:
16723 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16724 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16725 // CHECK16-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
16726 // CHECK16-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
16727 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16728 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
16729 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
16730 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
16731 // CHECK16-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
16732 // CHECK16-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
16733 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16734 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16735 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
16736 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16737 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16738 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
16739 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
16740 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16741 // CHECK16-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16742 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16743 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
16744 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
16745 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
16746 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
16747 // CHECK16-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
16748 // CHECK16-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
16749 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16750 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16751 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16752 // CHECK16-NEXT: [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16753 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
16754 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
16755 // CHECK16-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
16756 // CHECK16: omp.dispatch.cond:
16757 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16758 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
16759 // CHECK16-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
16760 // CHECK16-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16761 // CHECK16: cond.true:
16762 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
16763 // CHECK16-NEXT: br label [[COND_END:%.*]]
16764 // CHECK16: cond.false:
16765 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16766 // CHECK16-NEXT: br label [[COND_END]]
16767 // CHECK16: cond.end:
16768 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
16769 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
16770 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16771 // CHECK16-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
16772 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16773 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16774 // CHECK16-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
16775 // CHECK16-NEXT: br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
16776 // CHECK16: omp.dispatch.body:
16777 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
16778 // CHECK16: omp.inner.for.cond:
16779 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16780 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16781 // CHECK16-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
16782 // CHECK16-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16783 // CHECK16: omp.inner.for.body:
16784 // CHECK16-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16785 // CHECK16-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
16786 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
16787 // CHECK16-NEXT: store i32 [[ADD]], i32* [[I]], align 4
16788 // CHECK16-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4
16789 // CHECK16-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP16]]
16790 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
16791 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
16792 // CHECK16: omp.body.continue:
16793 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
16794 // CHECK16: omp.inner.for.inc:
16795 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16796 // CHECK16-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
16797 // CHECK16-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
16798 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
16799 // CHECK16: omp.inner.for.end:
16800 // CHECK16-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
16801 // CHECK16: omp.dispatch.inc:
16802 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16803 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
16804 // CHECK16-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
16805 // CHECK16-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
16806 // CHECK16-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16807 // CHECK16-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
16808 // CHECK16-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
16809 // CHECK16-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
16810 // CHECK16-NEXT: br label [[OMP_DISPATCH_COND]]
16811 // CHECK16: omp.dispatch.end:
16812 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
16813 // CHECK16-NEXT: ret void
16814 //
16815 //
16816 // CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
16817 // CHECK16-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
16818 // CHECK16-NEXT: entry:
16819 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16820 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16821 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16822 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
16823 // CHECK16-NEXT: ret void
16824 //
16825 //
16826 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..25
16827 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
16828 // CHECK16-NEXT: entry:
16829 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16830 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16831 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16832 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
16833 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
16834 // CHECK16-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
16835 // CHECK16-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
16836 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16837 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16838 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
16839 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16840 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16841 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16842 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16843 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
16844 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
16845 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16846 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16847 // CHECK16-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16848 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
16849 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
16850 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16851 // CHECK16-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
16852 // CHECK16-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
16853 // CHECK16: cond.true:
16854 // CHECK16-NEXT: br label [[COND_END:%.*]]
16855 // CHECK16: cond.false:
16856 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16857 // CHECK16-NEXT: br label [[COND_END]]
16858 // CHECK16: cond.end:
16859 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
16860 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
16861 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
16862 // CHECK16-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
16863 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
16864 // CHECK16: omp.inner.for.cond:
16865 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16866 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16867 // CHECK16-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
16868 // CHECK16-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16869 // CHECK16: omp.inner.for.body:
16870 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
16871 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
16872 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
16873 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
16874 // CHECK16: omp.inner.for.inc:
16875 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
16876 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
16877 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
16878 // CHECK16-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
16879 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
16880 // CHECK16: omp.inner.for.end:
16881 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
16882 // CHECK16: omp.loop.exit:
16883 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
16884 // CHECK16-NEXT: ret void
16885 //
16886 //
16887 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..26
16888 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
16889 // CHECK16-NEXT: entry:
16890 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16891 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16892 // CHECK16-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
16893 // CHECK16-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
16894 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16895 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
16896 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
16897 // CHECK16-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
16898 // CHECK16-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
16899 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16900 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16901 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
16902 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16903 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16904 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
16905 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
16906 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16907 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16908 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
16909 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
16910 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
16911 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
16912 // CHECK16-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
16913 // CHECK16-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
16914 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
16915 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
16916 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16917 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
16918 // CHECK16-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
16919 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
16920 // CHECK16-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 1073741859, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
16921 // CHECK16-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
16922 // CHECK16: omp.dispatch.cond:
16923 // CHECK16-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
16924 // CHECK16-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
16925 // CHECK16-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
16926 // CHECK16: omp.dispatch.body:
16927 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
16928 // CHECK16-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
16929 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
16930 // CHECK16: omp.inner.for.cond:
16931 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
16932 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
16933 // CHECK16-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
16934 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
16935 // CHECK16: omp.inner.for.body:
16936 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
16937 // CHECK16-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
16938 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
16939 // CHECK16-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
16940 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !22
16941 // CHECK16-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP12]]
16942 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !22
16943 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
16944 // CHECK16: omp.body.continue:
16945 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
16946 // CHECK16: omp.inner.for.inc:
16947 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
16948 // CHECK16-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
16949 // CHECK16-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
16950 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
16951 // CHECK16: omp.inner.for.end:
16952 // CHECK16-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
16953 // CHECK16: omp.dispatch.inc:
16954 // CHECK16-NEXT: br label [[OMP_DISPATCH_COND]]
16955 // CHECK16: omp.dispatch.end:
16956 // CHECK16-NEXT: ret void
16957 //
16958 //
16959 // CHECK16-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136
16960 // CHECK16-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
16961 // CHECK16-NEXT: entry:
16962 // CHECK16-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
16963 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16964 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
16965 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
16966 // CHECK16-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
16967 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16968 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16969 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
16970 // CHECK16-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
16971 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
16972 // CHECK16-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
16973 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
16974 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
16975 // CHECK16-NEXT: ret void
16976 //
16977 //
16978 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..29
16979 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
16980 // CHECK16-NEXT: entry:
16981 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
16982 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
16983 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
16984 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
16985 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
16986 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
16987 // CHECK16-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
16988 // CHECK16-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
16989 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
16990 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
16991 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
16992 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
16993 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
16994 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
16995 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
16996 // CHECK16-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
16997 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
16998 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
16999 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
17000 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17001 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17002 // CHECK16-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
17003 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
17004 // CHECK16-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17005 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
17006 // CHECK16-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
17007 // CHECK16-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17008 // CHECK16: cond.true:
17009 // CHECK16-NEXT: br label [[COND_END:%.*]]
17010 // CHECK16: cond.false:
17011 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
17012 // CHECK16-NEXT: br label [[COND_END]]
17013 // CHECK16: cond.end:
17014 // CHECK16-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
17015 // CHECK16-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
17016 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
17017 // CHECK16-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
17018 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
17019 // CHECK16: omp.inner.for.cond:
17020 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17021 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
17022 // CHECK16-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
17023 // CHECK16-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17024 // CHECK16: omp.inner.for.body:
17025 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
17026 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
17027 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
17028 // CHECK16-NEXT: store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
17029 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
17030 // CHECK16-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
17031 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
17032 // CHECK16: omp.inner.for.inc:
17033 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17034 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
17035 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
17036 // CHECK16-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
17037 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]]
17038 // CHECK16: omp.inner.for.end:
17039 // CHECK16-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
17040 // CHECK16: omp.loop.exit:
17041 // CHECK16-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
17042 // CHECK16-NEXT: ret void
17043 //
17044 //
17045 // CHECK16-LABEL: define {{[^@]+}}@.omp_outlined..30
17046 // CHECK16-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
17047 // CHECK16-NEXT: entry:
17048 // CHECK16-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
17049 // CHECK16-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
17050 // CHECK16-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
17051 // CHECK16-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
17052 // CHECK16-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
17053 // CHECK16-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
17054 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
17055 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
17056 // CHECK16-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
17057 // CHECK16-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
17058 // CHECK16-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17059 // CHECK16-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17060 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
17061 // CHECK16-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
17062 // CHECK16-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
17063 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
17064 // CHECK16-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
17065 // CHECK16-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
17066 // CHECK16-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
17067 // CHECK16-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
17068 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
17069 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
17070 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
17071 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
17072 // CHECK16-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
17073 // CHECK16-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
17074 // CHECK16-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17075 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17076 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
17077 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17078 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17079 // CHECK16-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
17080 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
17081 // CHECK16-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 1073741859, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
17082 // CHECK16-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
17083 // CHECK16: omp.dispatch.cond:
17084 // CHECK16-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
17085 // CHECK16-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
17086 // CHECK16-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
17087 // CHECK16: omp.dispatch.body:
17088 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17089 // CHECK16-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
17090 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
17091 // CHECK16: omp.inner.for.cond:
17092 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
17093 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
17094 // CHECK16-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
17095 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17096 // CHECK16: omp.inner.for.body:
17097 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
17098 // CHECK16-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
17099 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
17100 // CHECK16-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !25
17101 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !25
17102 // CHECK16-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
17103 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
17104 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
17105 // CHECK16: omp.body.continue:
17106 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
17107 // CHECK16: omp.inner.for.inc:
17108 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
17109 // CHECK16-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP14]], 1
17110 // CHECK16-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
17111 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
17112 // CHECK16: omp.inner.for.end:
17113 // CHECK16-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
17114 // CHECK16: omp.dispatch.inc:
17115 // CHECK16-NEXT: br label [[OMP_DISPATCH_COND]]
17116 // CHECK16: omp.dispatch.end:
17117 // CHECK16-NEXT: ret void
17118 //
17119 //
17120 // CHECK16-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
17121 // CHECK16-SAME: () #[[ATTR5:[0-9]+]] {
17122 // CHECK16-NEXT: entry:
17123 // CHECK16-NEXT: call void @__tgt_register_requires(i64 1)
17124 // CHECK16-NEXT: ret void
17125 //
17126 //
17127 // CHECK17-LABEL: define {{[^@]+}}@main
17128 // CHECK17-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
17129 // CHECK17-NEXT: entry:
17130 // CHECK17-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
17131 // CHECK17-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
17132 // CHECK17-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
17133 // CHECK17-NEXT: [[N:%.*]] = alloca i32, align 4
17134 // CHECK17-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
17135 // CHECK17-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
17136 // CHECK17-NEXT: [[M:%.*]] = alloca i32, align 4
17137 // CHECK17-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
17138 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
17139 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
17140 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
17141 // CHECK17-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
17142 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
17143 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
17144 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
17145 // CHECK17-NEXT: [[N_CASTED3:%.*]] = alloca i64, align 8
17146 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
17147 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
17148 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
17149 // CHECK17-NEXT: [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
17150 // CHECK17-NEXT: [[_TMP9:%.*]] = alloca i32, align 4
17151 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
17152 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
17153 // CHECK17-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
17154 // CHECK17-NEXT: [[N_CASTED19:%.*]] = alloca i64, align 8
17155 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [4 x i8*], align 8
17156 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS22:%.*]] = alloca [4 x i8*], align 8
17157 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [4 x i8*], align 8
17158 // CHECK17-NEXT: [[DOTOFFLOAD_SIZES24:%.*]] = alloca [4 x i64], align 8
17159 // CHECK17-NEXT: [[_TMP25:%.*]] = alloca i32, align 4
17160 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
17161 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
17162 // CHECK17-NEXT: [[N_CASTED34:%.*]] = alloca i64, align 8
17163 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS36:%.*]] = alloca [3 x i8*], align 8
17164 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS37:%.*]] = alloca [3 x i8*], align 8
17165 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS38:%.*]] = alloca [3 x i8*], align 8
17166 // CHECK17-NEXT: [[DOTOFFLOAD_SIZES39:%.*]] = alloca [3 x i64], align 8
17167 // CHECK17-NEXT: [[_TMP40:%.*]] = alloca i32, align 4
17168 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
17169 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
17170 // CHECK17-NEXT: [[M_CASTED49:%.*]] = alloca i64, align 8
17171 // CHECK17-NEXT: [[N_CASTED51:%.*]] = alloca i64, align 8
17172 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS53:%.*]] = alloca [4 x i8*], align 8
17173 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS54:%.*]] = alloca [4 x i8*], align 8
17174 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS55:%.*]] = alloca [4 x i8*], align 8
17175 // CHECK17-NEXT: [[DOTOFFLOAD_SIZES56:%.*]] = alloca [4 x i64], align 8
17176 // CHECK17-NEXT: [[_TMP57:%.*]] = alloca i32, align 4
17177 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_58:%.*]] = alloca i32, align 4
17178 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_59:%.*]] = alloca i32, align 4
17179 // CHECK17-NEXT: store i32 0, i32* [[RETVAL]], align 4
17180 // CHECK17-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
17181 // CHECK17-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
17182 // CHECK17-NEXT: store i32 100, i32* [[N]], align 4
17183 // CHECK17-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
17184 // CHECK17-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
17185 // CHECK17-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
17186 // CHECK17-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
17187 // CHECK17-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
17188 // CHECK17-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
17189 // CHECK17-NEXT: store i32 10, i32* [[M]], align 4
17190 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
17191 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
17192 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
17193 // CHECK17-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
17194 // CHECK17-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
17195 // CHECK17-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
17196 // CHECK17-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
17197 // CHECK17-NEXT: store i64 [[TMP4]], i64* [[TMP7]], align 8
17198 // CHECK17-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
17199 // CHECK17-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
17200 // CHECK17-NEXT: store i64 [[TMP4]], i64* [[TMP9]], align 8
17201 // CHECK17-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
17202 // CHECK17-NEXT: store i64 4, i64* [[TMP10]], align 8
17203 // CHECK17-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
17204 // CHECK17-NEXT: store i8* null, i8** [[TMP11]], align 8
17205 // CHECK17-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
17206 // CHECK17-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
17207 // CHECK17-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
17208 // CHECK17-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
17209 // CHECK17-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
17210 // CHECK17-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
17211 // CHECK17-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
17212 // CHECK17-NEXT: store i64 8, i64* [[TMP16]], align 8
17213 // CHECK17-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
17214 // CHECK17-NEXT: store i8* null, i8** [[TMP17]], align 8
17215 // CHECK17-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
17216 // CHECK17-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
17217 // CHECK17-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 8
17218 // CHECK17-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
17219 // CHECK17-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
17220 // CHECK17-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 8
17221 // CHECK17-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
17222 // CHECK17-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 8
17223 // CHECK17-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
17224 // CHECK17-NEXT: store i8* null, i8** [[TMP23]], align 8
17225 // CHECK17-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
17226 // CHECK17-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
17227 // CHECK17-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
17228 // CHECK17-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
17229 // CHECK17-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
17230 // CHECK17-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17231 // CHECK17-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
17232 // CHECK17-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
17233 // CHECK17-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
17234 // CHECK17-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
17235 // CHECK17-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17236 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
17237 // CHECK17-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
17238 // CHECK17-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
17239 // CHECK17-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
17240 // CHECK17-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
17241 // CHECK17-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
17242 // CHECK17: omp_offload.failed:
17243 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
17244 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT]]
17245 // CHECK17: omp_offload.cont:
17246 // CHECK17-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
17247 // CHECK17-NEXT: [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
17248 // CHECK17-NEXT: store i32 [[TMP33]], i32* [[CONV4]], align 4
17249 // CHECK17-NEXT: [[TMP34:%.*]] = load i64, i64* [[N_CASTED3]], align 8
17250 // CHECK17-NEXT: [[TMP35:%.*]] = mul nuw i64 [[TMP1]], 4
17251 // CHECK17-NEXT: [[TMP36:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
17252 // CHECK17-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
17253 // CHECK17-NEXT: store i64 [[TMP34]], i64* [[TMP37]], align 8
17254 // CHECK17-NEXT: [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
17255 // CHECK17-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
17256 // CHECK17-NEXT: store i64 [[TMP34]], i64* [[TMP39]], align 8
17257 // CHECK17-NEXT: [[TMP40:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
17258 // CHECK17-NEXT: store i64 4, i64* [[TMP40]], align 8
17259 // CHECK17-NEXT: [[TMP41:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
17260 // CHECK17-NEXT: store i8* null, i8** [[TMP41]], align 8
17261 // CHECK17-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
17262 // CHECK17-NEXT: [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
17263 // CHECK17-NEXT: store i64 [[TMP1]], i64* [[TMP43]], align 8
17264 // CHECK17-NEXT: [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
17265 // CHECK17-NEXT: [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i64*
17266 // CHECK17-NEXT: store i64 [[TMP1]], i64* [[TMP45]], align 8
17267 // CHECK17-NEXT: [[TMP46:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 1
17268 // CHECK17-NEXT: store i64 8, i64* [[TMP46]], align 8
17269 // CHECK17-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
17270 // CHECK17-NEXT: store i8* null, i8** [[TMP47]], align 8
17271 // CHECK17-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
17272 // CHECK17-NEXT: [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32**
17273 // CHECK17-NEXT: store i32* [[VLA]], i32** [[TMP49]], align 8
17274 // CHECK17-NEXT: [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
17275 // CHECK17-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i32**
17276 // CHECK17-NEXT: store i32* [[VLA]], i32** [[TMP51]], align 8
17277 // CHECK17-NEXT: [[TMP52:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
17278 // CHECK17-NEXT: store i64 [[TMP35]], i64* [[TMP52]], align 8
17279 // CHECK17-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
17280 // CHECK17-NEXT: store i8* null, i8** [[TMP53]], align 8
17281 // CHECK17-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
17282 // CHECK17-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
17283 // CHECK17-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
17284 // CHECK17-NEXT: [[TMP57:%.*]] = load i32, i32* [[N]], align 4
17285 // CHECK17-NEXT: store i32 [[TMP57]], i32* [[DOTCAPTURE_EXPR_10]], align 4
17286 // CHECK17-NEXT: [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
17287 // CHECK17-NEXT: [[SUB12:%.*]] = sub nsw i32 [[TMP58]], 0
17288 // CHECK17-NEXT: [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
17289 // CHECK17-NEXT: [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
17290 // CHECK17-NEXT: store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
17291 // CHECK17-NEXT: [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
17292 // CHECK17-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP59]], 1
17293 // CHECK17-NEXT: [[TMP60:%.*]] = zext i32 [[ADD15]] to i64
17294 // CHECK17-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP60]])
17295 // CHECK17-NEXT: [[TMP61:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153.region_id, i32 3, i8** [[TMP54]], i8** [[TMP55]], i64* [[TMP56]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
17296 // CHECK17-NEXT: [[TMP62:%.*]] = icmp ne i32 [[TMP61]], 0
17297 // CHECK17-NEXT: br i1 [[TMP62]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
17298 // CHECK17: omp_offload.failed16:
17299 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153(i64 [[TMP34]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
17300 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT17]]
17301 // CHECK17: omp_offload.cont17:
17302 // CHECK17-NEXT: [[TMP63:%.*]] = load i32, i32* [[M]], align 4
17303 // CHECK17-NEXT: [[CONV18:%.*]] = bitcast i64* [[M_CASTED]] to i32*
17304 // CHECK17-NEXT: store i32 [[TMP63]], i32* [[CONV18]], align 4
17305 // CHECK17-NEXT: [[TMP64:%.*]] = load i64, i64* [[M_CASTED]], align 8
17306 // CHECK17-NEXT: [[TMP65:%.*]] = load i32, i32* [[N]], align 4
17307 // CHECK17-NEXT: [[CONV20:%.*]] = bitcast i64* [[N_CASTED19]] to i32*
17308 // CHECK17-NEXT: store i32 [[TMP65]], i32* [[CONV20]], align 4
17309 // CHECK17-NEXT: [[TMP66:%.*]] = load i64, i64* [[N_CASTED19]], align 8
17310 // CHECK17-NEXT: [[TMP67:%.*]] = mul nuw i64 [[TMP1]], 4
17311 // CHECK17-NEXT: [[TMP68:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
17312 // CHECK17-NEXT: [[TMP69:%.*]] = bitcast i8** [[TMP68]] to i64*
17313 // CHECK17-NEXT: store i64 [[TMP64]], i64* [[TMP69]], align 8
17314 // CHECK17-NEXT: [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
17315 // CHECK17-NEXT: [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i64*
17316 // CHECK17-NEXT: store i64 [[TMP64]], i64* [[TMP71]], align 8
17317 // CHECK17-NEXT: [[TMP72:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
17318 // CHECK17-NEXT: store i64 4, i64* [[TMP72]], align 8
17319 // CHECK17-NEXT: [[TMP73:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
17320 // CHECK17-NEXT: store i8* null, i8** [[TMP73]], align 8
17321 // CHECK17-NEXT: [[TMP74:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
17322 // CHECK17-NEXT: [[TMP75:%.*]] = bitcast i8** [[TMP74]] to i64*
17323 // CHECK17-NEXT: store i64 [[TMP66]], i64* [[TMP75]], align 8
17324 // CHECK17-NEXT: [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
17325 // CHECK17-NEXT: [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i64*
17326 // CHECK17-NEXT: store i64 [[TMP66]], i64* [[TMP77]], align 8
17327 // CHECK17-NEXT: [[TMP78:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 1
17328 // CHECK17-NEXT: store i64 4, i64* [[TMP78]], align 8
17329 // CHECK17-NEXT: [[TMP79:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
17330 // CHECK17-NEXT: store i8* null, i8** [[TMP79]], align 8
17331 // CHECK17-NEXT: [[TMP80:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 2
17332 // CHECK17-NEXT: [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
17333 // CHECK17-NEXT: store i64 [[TMP1]], i64* [[TMP81]], align 8
17334 // CHECK17-NEXT: [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 2
17335 // CHECK17-NEXT: [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
17336 // CHECK17-NEXT: store i64 [[TMP1]], i64* [[TMP83]], align 8
17337 // CHECK17-NEXT: [[TMP84:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 2
17338 // CHECK17-NEXT: store i64 8, i64* [[TMP84]], align 8
17339 // CHECK17-NEXT: [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 2
17340 // CHECK17-NEXT: store i8* null, i8** [[TMP85]], align 8
17341 // CHECK17-NEXT: [[TMP86:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 3
17342 // CHECK17-NEXT: [[TMP87:%.*]] = bitcast i8** [[TMP86]] to i32**
17343 // CHECK17-NEXT: store i32* [[VLA]], i32** [[TMP87]], align 8
17344 // CHECK17-NEXT: [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 3
17345 // CHECK17-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
17346 // CHECK17-NEXT: store i32* [[VLA]], i32** [[TMP89]], align 8
17347 // CHECK17-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 3
17348 // CHECK17-NEXT: store i64 [[TMP67]], i64* [[TMP90]], align 8
17349 // CHECK17-NEXT: [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 3
17350 // CHECK17-NEXT: store i8* null, i8** [[TMP91]], align 8
17351 // CHECK17-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
17352 // CHECK17-NEXT: [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
17353 // CHECK17-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
17354 // CHECK17-NEXT: [[TMP95:%.*]] = load i32, i32* [[N]], align 4
17355 // CHECK17-NEXT: store i32 [[TMP95]], i32* [[DOTCAPTURE_EXPR_26]], align 4
17356 // CHECK17-NEXT: [[TMP96:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_26]], align 4
17357 // CHECK17-NEXT: [[SUB28:%.*]] = sub nsw i32 [[TMP96]], 0
17358 // CHECK17-NEXT: [[DIV29:%.*]] = sdiv i32 [[SUB28]], 1
17359 // CHECK17-NEXT: [[SUB30:%.*]] = sub nsw i32 [[DIV29]], 1
17360 // CHECK17-NEXT: store i32 [[SUB30]], i32* [[DOTCAPTURE_EXPR_27]], align 4
17361 // CHECK17-NEXT: [[TMP97:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
17362 // CHECK17-NEXT: [[ADD31:%.*]] = add nsw i32 [[TMP97]], 1
17363 // CHECK17-NEXT: [[TMP98:%.*]] = zext i32 [[ADD31]] to i64
17364 // CHECK17-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP98]])
17365 // CHECK17-NEXT: [[TMP99:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158.region_id, i32 4, i8** [[TMP92]], i8** [[TMP93]], i64* [[TMP94]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
17366 // CHECK17-NEXT: [[TMP100:%.*]] = icmp ne i32 [[TMP99]], 0
17367 // CHECK17-NEXT: br i1 [[TMP100]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
17368 // CHECK17: omp_offload.failed32:
17369 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158(i64 [[TMP64]], i64 [[TMP66]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
17370 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT33]]
17371 // CHECK17: omp_offload.cont33:
17372 // CHECK17-NEXT: [[TMP101:%.*]] = load i32, i32* [[N]], align 4
17373 // CHECK17-NEXT: [[CONV35:%.*]] = bitcast i64* [[N_CASTED34]] to i32*
17374 // CHECK17-NEXT: store i32 [[TMP101]], i32* [[CONV35]], align 4
17375 // CHECK17-NEXT: [[TMP102:%.*]] = load i64, i64* [[N_CASTED34]], align 8
17376 // CHECK17-NEXT: [[TMP103:%.*]] = mul nuw i64 [[TMP1]], 4
17377 // CHECK17-NEXT: [[TMP104:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
17378 // CHECK17-NEXT: [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i64*
17379 // CHECK17-NEXT: store i64 [[TMP102]], i64* [[TMP105]], align 8
17380 // CHECK17-NEXT: [[TMP106:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
17381 // CHECK17-NEXT: [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i64*
17382 // CHECK17-NEXT: store i64 [[TMP102]], i64* [[TMP107]], align 8
17383 // CHECK17-NEXT: [[TMP108:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
17384 // CHECK17-NEXT: store i64 4, i64* [[TMP108]], align 8
17385 // CHECK17-NEXT: [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 0
17386 // CHECK17-NEXT: store i8* null, i8** [[TMP109]], align 8
17387 // CHECK17-NEXT: [[TMP110:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 1
17388 // CHECK17-NEXT: [[TMP111:%.*]] = bitcast i8** [[TMP110]] to i64*
17389 // CHECK17-NEXT: store i64 [[TMP1]], i64* [[TMP111]], align 8
17390 // CHECK17-NEXT: [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 1
17391 // CHECK17-NEXT: [[TMP113:%.*]] = bitcast i8** [[TMP112]] to i64*
17392 // CHECK17-NEXT: store i64 [[TMP1]], i64* [[TMP113]], align 8
17393 // CHECK17-NEXT: [[TMP114:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 1
17394 // CHECK17-NEXT: store i64 8, i64* [[TMP114]], align 8
17395 // CHECK17-NEXT: [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 1
17396 // CHECK17-NEXT: store i8* null, i8** [[TMP115]], align 8
17397 // CHECK17-NEXT: [[TMP116:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 2
17398 // CHECK17-NEXT: [[TMP117:%.*]] = bitcast i8** [[TMP116]] to i32**
17399 // CHECK17-NEXT: store i32* [[VLA]], i32** [[TMP117]], align 8
17400 // CHECK17-NEXT: [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 2
17401 // CHECK17-NEXT: [[TMP119:%.*]] = bitcast i8** [[TMP118]] to i32**
17402 // CHECK17-NEXT: store i32* [[VLA]], i32** [[TMP119]], align 8
17403 // CHECK17-NEXT: [[TMP120:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 2
17404 // CHECK17-NEXT: store i64 [[TMP103]], i64* [[TMP120]], align 8
17405 // CHECK17-NEXT: [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 2
17406 // CHECK17-NEXT: store i8* null, i8** [[TMP121]], align 8
17407 // CHECK17-NEXT: [[TMP122:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
17408 // CHECK17-NEXT: [[TMP123:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
17409 // CHECK17-NEXT: [[TMP124:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
17410 // CHECK17-NEXT: [[TMP125:%.*]] = load i32, i32* [[N]], align 4
17411 // CHECK17-NEXT: store i32 [[TMP125]], i32* [[DOTCAPTURE_EXPR_41]], align 4
17412 // CHECK17-NEXT: [[TMP126:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
17413 // CHECK17-NEXT: [[SUB43:%.*]] = sub nsw i32 [[TMP126]], 0
17414 // CHECK17-NEXT: [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
17415 // CHECK17-NEXT: [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
17416 // CHECK17-NEXT: store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
17417 // CHECK17-NEXT: [[TMP127:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
17418 // CHECK17-NEXT: [[ADD46:%.*]] = add nsw i32 [[TMP127]], 1
17419 // CHECK17-NEXT: [[TMP128:%.*]] = zext i32 [[ADD46]] to i64
17420 // CHECK17-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP128]])
17421 // CHECK17-NEXT: [[TMP129:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163.region_id, i32 3, i8** [[TMP122]], i8** [[TMP123]], i64* [[TMP124]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
17422 // CHECK17-NEXT: [[TMP130:%.*]] = icmp ne i32 [[TMP129]], 0
17423 // CHECK17-NEXT: br i1 [[TMP130]], label [[OMP_OFFLOAD_FAILED47:%.*]], label [[OMP_OFFLOAD_CONT48:%.*]]
17424 // CHECK17: omp_offload.failed47:
17425 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163(i64 [[TMP102]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
17426 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT48]]
17427 // CHECK17: omp_offload.cont48:
17428 // CHECK17-NEXT: [[TMP131:%.*]] = load i32, i32* [[M]], align 4
17429 // CHECK17-NEXT: [[CONV50:%.*]] = bitcast i64* [[M_CASTED49]] to i32*
17430 // CHECK17-NEXT: store i32 [[TMP131]], i32* [[CONV50]], align 4
17431 // CHECK17-NEXT: [[TMP132:%.*]] = load i64, i64* [[M_CASTED49]], align 8
17432 // CHECK17-NEXT: [[TMP133:%.*]] = load i32, i32* [[N]], align 4
17433 // CHECK17-NEXT: [[CONV52:%.*]] = bitcast i64* [[N_CASTED51]] to i32*
17434 // CHECK17-NEXT: store i32 [[TMP133]], i32* [[CONV52]], align 4
17435 // CHECK17-NEXT: [[TMP134:%.*]] = load i64, i64* [[N_CASTED51]], align 8
17436 // CHECK17-NEXT: [[TMP135:%.*]] = mul nuw i64 [[TMP1]], 4
17437 // CHECK17-NEXT: [[TMP136:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
17438 // CHECK17-NEXT: [[TMP137:%.*]] = bitcast i8** [[TMP136]] to i64*
17439 // CHECK17-NEXT: store i64 [[TMP132]], i64* [[TMP137]], align 8
17440 // CHECK17-NEXT: [[TMP138:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
17441 // CHECK17-NEXT: [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i64*
17442 // CHECK17-NEXT: store i64 [[TMP132]], i64* [[TMP139]], align 8
17443 // CHECK17-NEXT: [[TMP140:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
17444 // CHECK17-NEXT: store i64 4, i64* [[TMP140]], align 8
17445 // CHECK17-NEXT: [[TMP141:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 0
17446 // CHECK17-NEXT: store i8* null, i8** [[TMP141]], align 8
17447 // CHECK17-NEXT: [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 1
17448 // CHECK17-NEXT: [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i64*
17449 // CHECK17-NEXT: store i64 [[TMP134]], i64* [[TMP143]], align 8
17450 // CHECK17-NEXT: [[TMP144:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 1
17451 // CHECK17-NEXT: [[TMP145:%.*]] = bitcast i8** [[TMP144]] to i64*
17452 // CHECK17-NEXT: store i64 [[TMP134]], i64* [[TMP145]], align 8
17453 // CHECK17-NEXT: [[TMP146:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 1
17454 // CHECK17-NEXT: store i64 4, i64* [[TMP146]], align 8
17455 // CHECK17-NEXT: [[TMP147:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 1
17456 // CHECK17-NEXT: store i8* null, i8** [[TMP147]], align 8
17457 // CHECK17-NEXT: [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 2
17458 // CHECK17-NEXT: [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
17459 // CHECK17-NEXT: store i64 [[TMP1]], i64* [[TMP149]], align 8
17460 // CHECK17-NEXT: [[TMP150:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 2
17461 // CHECK17-NEXT: [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
17462 // CHECK17-NEXT: store i64 [[TMP1]], i64* [[TMP151]], align 8
17463 // CHECK17-NEXT: [[TMP152:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 2
17464 // CHECK17-NEXT: store i64 8, i64* [[TMP152]], align 8
17465 // CHECK17-NEXT: [[TMP153:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 2
17466 // CHECK17-NEXT: store i8* null, i8** [[TMP153]], align 8
17467 // CHECK17-NEXT: [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 3
17468 // CHECK17-NEXT: [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32**
17469 // CHECK17-NEXT: store i32* [[VLA]], i32** [[TMP155]], align 8
17470 // CHECK17-NEXT: [[TMP156:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 3
17471 // CHECK17-NEXT: [[TMP157:%.*]] = bitcast i8** [[TMP156]] to i32**
17472 // CHECK17-NEXT: store i32* [[VLA]], i32** [[TMP157]], align 8
17473 // CHECK17-NEXT: [[TMP158:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 3
17474 // CHECK17-NEXT: store i64 [[TMP135]], i64* [[TMP158]], align 8
17475 // CHECK17-NEXT: [[TMP159:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 3
17476 // CHECK17-NEXT: store i8* null, i8** [[TMP159]], align 8
17477 // CHECK17-NEXT: [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
17478 // CHECK17-NEXT: [[TMP161:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
17479 // CHECK17-NEXT: [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
17480 // CHECK17-NEXT: [[TMP163:%.*]] = load i32, i32* [[N]], align 4
17481 // CHECK17-NEXT: store i32 [[TMP163]], i32* [[DOTCAPTURE_EXPR_58]], align 4
17482 // CHECK17-NEXT: [[TMP164:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_58]], align 4
17483 // CHECK17-NEXT: [[SUB60:%.*]] = sub nsw i32 [[TMP164]], 0
17484 // CHECK17-NEXT: [[DIV61:%.*]] = sdiv i32 [[SUB60]], 1
17485 // CHECK17-NEXT: [[SUB62:%.*]] = sub nsw i32 [[DIV61]], 1
17486 // CHECK17-NEXT: store i32 [[SUB62]], i32* [[DOTCAPTURE_EXPR_59]], align 4
17487 // CHECK17-NEXT: [[TMP165:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_59]], align 4
17488 // CHECK17-NEXT: [[ADD63:%.*]] = add nsw i32 [[TMP165]], 1
17489 // CHECK17-NEXT: [[TMP166:%.*]] = zext i32 [[ADD63]] to i64
17490 // CHECK17-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP166]])
17491 // CHECK17-NEXT: [[TMP167:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168.region_id, i32 4, i8** [[TMP160]], i8** [[TMP161]], i64* [[TMP162]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
17492 // CHECK17-NEXT: [[TMP168:%.*]] = icmp ne i32 [[TMP167]], 0
17493 // CHECK17-NEXT: br i1 [[TMP168]], label [[OMP_OFFLOAD_FAILED64:%.*]], label [[OMP_OFFLOAD_CONT65:%.*]]
17494 // CHECK17: omp_offload.failed64:
17495 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168(i64 [[TMP132]], i64 [[TMP134]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
17496 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT65]]
17497 // CHECK17: omp_offload.cont65:
17498 // CHECK17-NEXT: [[TMP169:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
17499 // CHECK17-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP169]])
17500 // CHECK17-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
17501 // CHECK17-NEXT: [[TMP170:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
17502 // CHECK17-NEXT: call void @llvm.stackrestore(i8* [[TMP170]])
17503 // CHECK17-NEXT: [[TMP171:%.*]] = load i32, i32* [[RETVAL]], align 4
17504 // CHECK17-NEXT: ret i32 [[TMP171]]
17505 //
17506 //
17507 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148
17508 // CHECK17-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
17509 // CHECK17-NEXT: entry:
17510 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
17511 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
17512 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
17513 // CHECK17-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
17514 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
17515 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
17516 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
17517 // CHECK17-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
17518 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
17519 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
17520 // CHECK17-NEXT: ret void
17521 //
17522 //
17523 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
17524 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
17525 // CHECK17-NEXT: entry:
17526 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17527 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17528 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
17529 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
17530 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
17531 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
17532 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
17533 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
17534 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
17535 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
17536 // CHECK17-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
17537 // CHECK17-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
17538 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17539 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17540 // CHECK17-NEXT: [[I3:%.*]] = alloca i32, align 4
17541 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17542 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17543 // CHECK17-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
17544 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
17545 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
17546 // CHECK17-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
17547 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
17548 // CHECK17-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
17549 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
17550 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
17551 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17552 // CHECK17-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
17553 // CHECK17-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
17554 // CHECK17-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
17555 // CHECK17-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
17556 // CHECK17-NEXT: store i32 0, i32* [[I]], align 4
17557 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17558 // CHECK17-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
17559 // CHECK17-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
17560 // CHECK17: omp.precond.then:
17561 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
17562 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17563 // CHECK17-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
17564 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17565 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17566 // CHECK17-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17567 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
17568 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17569 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
17570 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17571 // CHECK17-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
17572 // CHECK17-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17573 // CHECK17: cond.true:
17574 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17575 // CHECK17-NEXT: br label [[COND_END:%.*]]
17576 // CHECK17: cond.false:
17577 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
17578 // CHECK17-NEXT: br label [[COND_END]]
17579 // CHECK17: cond.end:
17580 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
17581 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
17582 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
17583 // CHECK17-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
17584 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
17585 // CHECK17: omp.inner.for.cond:
17586 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17587 // CHECK17-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
17588 // CHECK17-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
17589 // CHECK17-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17590 // CHECK17: omp.inner.for.body:
17591 // CHECK17-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
17592 // CHECK17-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
17593 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
17594 // CHECK17-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
17595 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
17596 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
17597 // CHECK17: omp.inner.for.inc:
17598 // CHECK17-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17599 // CHECK17-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
17600 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
17601 // CHECK17-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
17602 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
17603 // CHECK17: omp.inner.for.end:
17604 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
17605 // CHECK17: omp.loop.exit:
17606 // CHECK17-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17607 // CHECK17-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
17608 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
17609 // CHECK17-NEXT: br label [[OMP_PRECOND_END]]
17610 // CHECK17: omp.precond.end:
17611 // CHECK17-NEXT: ret void
17612 //
17613 //
17614 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..1
17615 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
17616 // CHECK17-NEXT: entry:
17617 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17618 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17619 // CHECK17-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
17620 // CHECK17-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
17621 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
17622 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
17623 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
17624 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
17625 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
17626 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
17627 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
17628 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
17629 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
17630 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
17631 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17632 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17633 // CHECK17-NEXT: [[I4:%.*]] = alloca i32, align 4
17634 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17635 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17636 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
17637 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
17638 // CHECK17-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
17639 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
17640 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
17641 // CHECK17-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
17642 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
17643 // CHECK17-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
17644 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
17645 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
17646 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17647 // CHECK17-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
17648 // CHECK17-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
17649 // CHECK17-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
17650 // CHECK17-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
17651 // CHECK17-NEXT: store i32 0, i32* [[I]], align 4
17652 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17653 // CHECK17-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
17654 // CHECK17-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
17655 // CHECK17: omp.precond.then:
17656 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
17657 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17658 // CHECK17-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
17659 // CHECK17-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
17660 // CHECK17-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
17661 // CHECK17-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
17662 // CHECK17-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
17663 // CHECK17-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
17664 // CHECK17-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
17665 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17666 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17667 // CHECK17-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17668 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
17669 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17670 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17671 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17672 // CHECK17-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
17673 // CHECK17-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17674 // CHECK17: cond.true:
17675 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17676 // CHECK17-NEXT: br label [[COND_END:%.*]]
17677 // CHECK17: cond.false:
17678 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17679 // CHECK17-NEXT: br label [[COND_END]]
17680 // CHECK17: cond.end:
17681 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
17682 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
17683 // CHECK17-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17684 // CHECK17-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
17685 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
17686 // CHECK17: omp.inner.for.cond:
17687 // CHECK17-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17688 // CHECK17-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17689 // CHECK17-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
17690 // CHECK17-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17691 // CHECK17: omp.inner.for.body:
17692 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17693 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
17694 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
17695 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
17696 // CHECK17-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
17697 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
17698 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
17699 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
17700 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
17701 // CHECK17: omp.body.continue:
17702 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
17703 // CHECK17: omp.inner.for.inc:
17704 // CHECK17-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17705 // CHECK17-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
17706 // CHECK17-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
17707 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
17708 // CHECK17: omp.inner.for.end:
17709 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
17710 // CHECK17: omp.loop.exit:
17711 // CHECK17-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17712 // CHECK17-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
17713 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
17714 // CHECK17-NEXT: br label [[OMP_PRECOND_END]]
17715 // CHECK17: omp.precond.end:
17716 // CHECK17-NEXT: ret void
17717 //
17718 //
17719 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153
17720 // CHECK17-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
17721 // CHECK17-NEXT: entry:
17722 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
17723 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
17724 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
17725 // CHECK17-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
17726 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
17727 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
17728 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
17729 // CHECK17-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
17730 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
17731 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
17732 // CHECK17-NEXT: ret void
17733 //
17734 //
17735 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..2
17736 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
17737 // CHECK17-NEXT: entry:
17738 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17739 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17740 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
17741 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
17742 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
17743 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
17744 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
17745 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
17746 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
17747 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
17748 // CHECK17-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
17749 // CHECK17-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
17750 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17751 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17752 // CHECK17-NEXT: [[I3:%.*]] = alloca i32, align 4
17753 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17754 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17755 // CHECK17-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
17756 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
17757 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
17758 // CHECK17-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
17759 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
17760 // CHECK17-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
17761 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
17762 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
17763 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17764 // CHECK17-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
17765 // CHECK17-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
17766 // CHECK17-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
17767 // CHECK17-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
17768 // CHECK17-NEXT: store i32 0, i32* [[I]], align 4
17769 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17770 // CHECK17-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
17771 // CHECK17-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
17772 // CHECK17: omp.precond.then:
17773 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
17774 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17775 // CHECK17-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
17776 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17777 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17778 // CHECK17-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17779 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
17780 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17781 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
17782 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17783 // CHECK17-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
17784 // CHECK17-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17785 // CHECK17: cond.true:
17786 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17787 // CHECK17-NEXT: br label [[COND_END:%.*]]
17788 // CHECK17: cond.false:
17789 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
17790 // CHECK17-NEXT: br label [[COND_END]]
17791 // CHECK17: cond.end:
17792 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
17793 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
17794 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
17795 // CHECK17-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
17796 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
17797 // CHECK17: omp.inner.for.cond:
17798 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17799 // CHECK17-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
17800 // CHECK17-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
17801 // CHECK17-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17802 // CHECK17: omp.inner.for.body:
17803 // CHECK17-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
17804 // CHECK17-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
17805 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
17806 // CHECK17-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
17807 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
17808 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
17809 // CHECK17: omp.inner.for.inc:
17810 // CHECK17-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17811 // CHECK17-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
17812 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
17813 // CHECK17-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
17814 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
17815 // CHECK17: omp.inner.for.end:
17816 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
17817 // CHECK17: omp.loop.exit:
17818 // CHECK17-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17819 // CHECK17-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
17820 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
17821 // CHECK17-NEXT: br label [[OMP_PRECOND_END]]
17822 // CHECK17: omp.precond.end:
17823 // CHECK17-NEXT: ret void
17824 //
17825 //
17826 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..3
17827 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
17828 // CHECK17-NEXT: entry:
17829 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17830 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17831 // CHECK17-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
17832 // CHECK17-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
17833 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
17834 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
17835 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
17836 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
17837 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
17838 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
17839 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
17840 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
17841 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
17842 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
17843 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17844 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17845 // CHECK17-NEXT: [[I4:%.*]] = alloca i32, align 4
17846 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17847 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17848 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
17849 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
17850 // CHECK17-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
17851 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
17852 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
17853 // CHECK17-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
17854 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
17855 // CHECK17-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
17856 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
17857 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
17858 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17859 // CHECK17-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
17860 // CHECK17-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
17861 // CHECK17-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
17862 // CHECK17-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
17863 // CHECK17-NEXT: store i32 0, i32* [[I]], align 4
17864 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17865 // CHECK17-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
17866 // CHECK17-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
17867 // CHECK17: omp.precond.then:
17868 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
17869 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17870 // CHECK17-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
17871 // CHECK17-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
17872 // CHECK17-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
17873 // CHECK17-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
17874 // CHECK17-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
17875 // CHECK17-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
17876 // CHECK17-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
17877 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
17878 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
17879 // CHECK17-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17880 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
17881 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
17882 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17883 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17884 // CHECK17-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
17885 // CHECK17-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
17886 // CHECK17: cond.true:
17887 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17888 // CHECK17-NEXT: br label [[COND_END:%.*]]
17889 // CHECK17: cond.false:
17890 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17891 // CHECK17-NEXT: br label [[COND_END]]
17892 // CHECK17: cond.end:
17893 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
17894 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
17895 // CHECK17-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
17896 // CHECK17-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
17897 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
17898 // CHECK17: omp.inner.for.cond:
17899 // CHECK17-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17900 // CHECK17-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
17901 // CHECK17-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
17902 // CHECK17-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
17903 // CHECK17: omp.inner.for.body:
17904 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17905 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
17906 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
17907 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
17908 // CHECK17-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
17909 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
17910 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
17911 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
17912 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
17913 // CHECK17: omp.body.continue:
17914 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
17915 // CHECK17: omp.inner.for.inc:
17916 // CHECK17-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
17917 // CHECK17-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
17918 // CHECK17-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
17919 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
17920 // CHECK17: omp.inner.for.end:
17921 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
17922 // CHECK17: omp.loop.exit:
17923 // CHECK17-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
17924 // CHECK17-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
17925 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
17926 // CHECK17-NEXT: br label [[OMP_PRECOND_END]]
17927 // CHECK17: omp.precond.end:
17928 // CHECK17-NEXT: ret void
17929 //
17930 //
17931 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158
17932 // CHECK17-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
17933 // CHECK17-NEXT: entry:
17934 // CHECK17-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
17935 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
17936 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
17937 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
17938 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
17939 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
17940 // CHECK17-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
17941 // CHECK17-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
17942 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
17943 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
17944 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
17945 // CHECK17-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
17946 // CHECK17-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
17947 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
17948 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
17949 // CHECK17-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
17950 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
17951 // CHECK17-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
17952 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
17953 // CHECK17-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
17954 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
17955 // CHECK17-NEXT: ret void
17956 //
17957 //
17958 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..5
17959 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
17960 // CHECK17-NEXT: entry:
17961 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
17962 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
17963 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
17964 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
17965 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
17966 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
17967 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
17968 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
17969 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
17970 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
17971 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
17972 // CHECK17-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
17973 // CHECK17-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
17974 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
17975 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
17976 // CHECK17-NEXT: [[I4:%.*]] = alloca i32, align 4
17977 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
17978 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
17979 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
17980 // CHECK17-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
17981 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
17982 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
17983 // CHECK17-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
17984 // CHECK17-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
17985 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
17986 // CHECK17-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
17987 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
17988 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
17989 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
17990 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17991 // CHECK17-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
17992 // CHECK17-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
17993 // CHECK17-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
17994 // CHECK17-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
17995 // CHECK17-NEXT: store i32 0, i32* [[I]], align 4
17996 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
17997 // CHECK17-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
17998 // CHECK17-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
17999 // CHECK17: omp.precond.then:
18000 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
18001 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18002 // CHECK17-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
18003 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18004 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18005 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[CONV]], align 8
18006 // CHECK17-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18007 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
18008 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
18009 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18010 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18011 // CHECK17-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
18012 // CHECK17-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18013 // CHECK17: cond.true:
18014 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18015 // CHECK17-NEXT: br label [[COND_END:%.*]]
18016 // CHECK17: cond.false:
18017 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18018 // CHECK17-NEXT: br label [[COND_END]]
18019 // CHECK17: cond.end:
18020 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
18021 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
18022 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
18023 // CHECK17-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
18024 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
18025 // CHECK17: omp.inner.for.cond:
18026 // CHECK17-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18027 // CHECK17-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18028 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
18029 // CHECK17-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
18030 // CHECK17-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18031 // CHECK17: omp.inner.for.body:
18032 // CHECK17-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
18033 // CHECK17-NEXT: [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
18034 // CHECK17-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18035 // CHECK17-NEXT: [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
18036 // CHECK17-NEXT: [[TMP21:%.*]] = load i32, i32* [[CONV]], align 8
18037 // CHECK17-NEXT: [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
18038 // CHECK17-NEXT: store i32 [[TMP21]], i32* [[CONV7]], align 4
18039 // CHECK17-NEXT: [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
18040 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]])
18041 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
18042 // CHECK17: omp.inner.for.inc:
18043 // CHECK17-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18044 // CHECK17-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
18045 // CHECK17-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
18046 // CHECK17-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
18047 // CHECK17-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
18048 // CHECK17-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
18049 // CHECK17-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
18050 // CHECK17-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4
18051 // CHECK17-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18052 // CHECK17-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
18053 // CHECK17-NEXT: [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
18054 // CHECK17-NEXT: store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4
18055 // CHECK17-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18056 // CHECK17-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18057 // CHECK17-NEXT: [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
18058 // CHECK17-NEXT: br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
18059 // CHECK17: cond.true12:
18060 // CHECK17-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18061 // CHECK17-NEXT: br label [[COND_END14:%.*]]
18062 // CHECK17: cond.false13:
18063 // CHECK17-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18064 // CHECK17-NEXT: br label [[COND_END14]]
18065 // CHECK17: cond.end14:
18066 // CHECK17-NEXT: [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
18067 // CHECK17-NEXT: store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4
18068 // CHECK17-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
18069 // CHECK17-NEXT: store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4
18070 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
18071 // CHECK17: omp.inner.for.end:
18072 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
18073 // CHECK17: omp.loop.exit:
18074 // CHECK17-NEXT: [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18075 // CHECK17-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
18076 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
18077 // CHECK17-NEXT: br label [[OMP_PRECOND_END]]
18078 // CHECK17: omp.precond.end:
18079 // CHECK17-NEXT: ret void
18080 //
18081 //
18082 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..6
18083 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
18084 // CHECK17-NEXT: entry:
18085 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18086 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18087 // CHECK17-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
18088 // CHECK17-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
18089 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
18090 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
18091 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
18092 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
18093 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
18094 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
18095 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
18096 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
18097 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
18098 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
18099 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
18100 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18101 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18102 // CHECK17-NEXT: [[I6:%.*]] = alloca i32, align 4
18103 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18104 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18105 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
18106 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
18107 // CHECK17-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
18108 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
18109 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
18110 // CHECK17-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
18111 // CHECK17-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
18112 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
18113 // CHECK17-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
18114 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
18115 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
18116 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
18117 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
18118 // CHECK17-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
18119 // CHECK17-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
18120 // CHECK17-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
18121 // CHECK17-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
18122 // CHECK17-NEXT: store i32 0, i32* [[I]], align 4
18123 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
18124 // CHECK17-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
18125 // CHECK17-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
18126 // CHECK17: omp.precond.then:
18127 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
18128 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18129 // CHECK17-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
18130 // CHECK17-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
18131 // CHECK17-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
18132 // CHECK17-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
18133 // CHECK17-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
18134 // CHECK17-NEXT: store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
18135 // CHECK17-NEXT: store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
18136 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18137 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18138 // CHECK17-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18139 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
18140 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18141 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18142 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18143 // CHECK17-NEXT: [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
18144 // CHECK17-NEXT: br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18145 // CHECK17: cond.true:
18146 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18147 // CHECK17-NEXT: br label [[COND_END:%.*]]
18148 // CHECK17: cond.false:
18149 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18150 // CHECK17-NEXT: br label [[COND_END]]
18151 // CHECK17: cond.end:
18152 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
18153 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
18154 // CHECK17-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18155 // CHECK17-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
18156 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
18157 // CHECK17: omp.inner.for.cond:
18158 // CHECK17-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18159 // CHECK17-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18160 // CHECK17-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
18161 // CHECK17-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18162 // CHECK17: omp.inner.for.body:
18163 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18164 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
18165 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
18166 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I6]], align 4
18167 // CHECK17-NEXT: [[TMP19:%.*]] = load i32, i32* [[I6]], align 4
18168 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
18169 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
18170 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
18171 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
18172 // CHECK17: omp.body.continue:
18173 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
18174 // CHECK17: omp.inner.for.inc:
18175 // CHECK17-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18176 // CHECK17-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
18177 // CHECK17-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
18178 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
18179 // CHECK17: omp.inner.for.end:
18180 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
18181 // CHECK17: omp.loop.exit:
18182 // CHECK17-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18183 // CHECK17-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
18184 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
18185 // CHECK17-NEXT: br label [[OMP_PRECOND_END]]
18186 // CHECK17: omp.precond.end:
18187 // CHECK17-NEXT: ret void
18188 //
18189 //
18190 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163
18191 // CHECK17-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
18192 // CHECK17-NEXT: entry:
18193 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
18194 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
18195 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
18196 // CHECK17-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
18197 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
18198 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
18199 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
18200 // CHECK17-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
18201 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
18202 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
18203 // CHECK17-NEXT: ret void
18204 //
18205 //
18206 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..8
18207 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
18208 // CHECK17-NEXT: entry:
18209 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18210 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18211 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
18212 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
18213 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
18214 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
18215 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
18216 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
18217 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
18218 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
18219 // CHECK17-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
18220 // CHECK17-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
18221 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18222 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18223 // CHECK17-NEXT: [[I3:%.*]] = alloca i32, align 4
18224 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18225 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18226 // CHECK17-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
18227 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
18228 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
18229 // CHECK17-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
18230 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
18231 // CHECK17-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
18232 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
18233 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
18234 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
18235 // CHECK17-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
18236 // CHECK17-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
18237 // CHECK17-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
18238 // CHECK17-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
18239 // CHECK17-NEXT: store i32 0, i32* [[I]], align 4
18240 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
18241 // CHECK17-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
18242 // CHECK17-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
18243 // CHECK17: omp.precond.then:
18244 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
18245 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
18246 // CHECK17-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
18247 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18248 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18249 // CHECK17-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18250 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
18251 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18252 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18253 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
18254 // CHECK17-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
18255 // CHECK17-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18256 // CHECK17: cond.true:
18257 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
18258 // CHECK17-NEXT: br label [[COND_END:%.*]]
18259 // CHECK17: cond.false:
18260 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18261 // CHECK17-NEXT: br label [[COND_END]]
18262 // CHECK17: cond.end:
18263 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
18264 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
18265 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
18266 // CHECK17-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
18267 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
18268 // CHECK17: omp.inner.for.cond:
18269 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18270 // CHECK17-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18271 // CHECK17-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
18272 // CHECK17-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18273 // CHECK17: omp.inner.for.body:
18274 // CHECK17-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
18275 // CHECK17-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
18276 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18277 // CHECK17-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
18278 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
18279 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
18280 // CHECK17: omp.inner.for.inc:
18281 // CHECK17-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18282 // CHECK17-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
18283 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
18284 // CHECK17-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
18285 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
18286 // CHECK17: omp.inner.for.end:
18287 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
18288 // CHECK17: omp.loop.exit:
18289 // CHECK17-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18290 // CHECK17-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
18291 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
18292 // CHECK17-NEXT: br label [[OMP_PRECOND_END]]
18293 // CHECK17: omp.precond.end:
18294 // CHECK17-NEXT: ret void
18295 //
18296 //
18297 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..9
18298 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
18299 // CHECK17-NEXT: entry:
18300 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18301 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18302 // CHECK17-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
18303 // CHECK17-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
18304 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
18305 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
18306 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
18307 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
18308 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
18309 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
18310 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
18311 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
18312 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
18313 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
18314 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18315 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18316 // CHECK17-NEXT: [[I4:%.*]] = alloca i32, align 4
18317 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18318 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18319 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
18320 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
18321 // CHECK17-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
18322 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
18323 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
18324 // CHECK17-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
18325 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
18326 // CHECK17-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
18327 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
18328 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
18329 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
18330 // CHECK17-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
18331 // CHECK17-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
18332 // CHECK17-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
18333 // CHECK17-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
18334 // CHECK17-NEXT: store i32 0, i32* [[I]], align 4
18335 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
18336 // CHECK17-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
18337 // CHECK17-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
18338 // CHECK17: omp.precond.then:
18339 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
18340 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
18341 // CHECK17-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
18342 // CHECK17-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
18343 // CHECK17-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
18344 // CHECK17-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
18345 // CHECK17-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
18346 // CHECK17-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
18347 // CHECK17-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
18348 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18349 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18350 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18351 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18352 // CHECK17-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18353 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
18354 // CHECK17-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 35, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
18355 // CHECK17-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
18356 // CHECK17: omp.dispatch.cond:
18357 // CHECK17-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18358 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
18359 // CHECK17-NEXT: [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
18360 // CHECK17-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
18361 // CHECK17-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
18362 // CHECK17: omp.dispatch.body:
18363 // CHECK17-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18364 // CHECK17-NEXT: store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
18365 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
18366 // CHECK17: omp.inner.for.cond:
18367 // CHECK17-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
18368 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !15
18369 // CHECK17-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
18370 // CHECK17-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18371 // CHECK17: omp.inner.for.body:
18372 // CHECK17-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
18373 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
18374 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
18375 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !15
18376 // CHECK17-NEXT: [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !15
18377 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
18378 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
18379 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !15
18380 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
18381 // CHECK17: omp.body.continue:
18382 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
18383 // CHECK17: omp.inner.for.inc:
18384 // CHECK17-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
18385 // CHECK17-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP21]], 1
18386 // CHECK17-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
18387 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
18388 // CHECK17: omp.inner.for.end:
18389 // CHECK17-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
18390 // CHECK17: omp.dispatch.inc:
18391 // CHECK17-NEXT: br label [[OMP_DISPATCH_COND]]
18392 // CHECK17: omp.dispatch.end:
18393 // CHECK17-NEXT: br label [[OMP_PRECOND_END]]
18394 // CHECK17: omp.precond.end:
18395 // CHECK17-NEXT: ret void
18396 //
18397 //
18398 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168
18399 // CHECK17-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
18400 // CHECK17-NEXT: entry:
18401 // CHECK17-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
18402 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
18403 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
18404 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
18405 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
18406 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
18407 // CHECK17-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
18408 // CHECK17-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
18409 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
18410 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
18411 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
18412 // CHECK17-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
18413 // CHECK17-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
18414 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
18415 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
18416 // CHECK17-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
18417 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
18418 // CHECK17-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
18419 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
18420 // CHECK17-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
18421 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
18422 // CHECK17-NEXT: ret void
18423 //
18424 //
18425 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..11
18426 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
18427 // CHECK17-NEXT: entry:
18428 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18429 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18430 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
18431 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
18432 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
18433 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
18434 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
18435 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
18436 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
18437 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
18438 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
18439 // CHECK17-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
18440 // CHECK17-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
18441 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18442 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18443 // CHECK17-NEXT: [[I4:%.*]] = alloca i32, align 4
18444 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
18445 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18446 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18447 // CHECK17-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
18448 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
18449 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
18450 // CHECK17-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
18451 // CHECK17-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
18452 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
18453 // CHECK17-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
18454 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
18455 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
18456 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
18457 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
18458 // CHECK17-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
18459 // CHECK17-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
18460 // CHECK17-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
18461 // CHECK17-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
18462 // CHECK17-NEXT: store i32 0, i32* [[I]], align 4
18463 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
18464 // CHECK17-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
18465 // CHECK17-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
18466 // CHECK17: omp.precond.then:
18467 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
18468 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18469 // CHECK17-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
18470 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18471 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18472 // CHECK17-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18473 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
18474 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18475 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18476 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18477 // CHECK17-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
18478 // CHECK17-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18479 // CHECK17: cond.true:
18480 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18481 // CHECK17-NEXT: br label [[COND_END:%.*]]
18482 // CHECK17: cond.false:
18483 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18484 // CHECK17-NEXT: br label [[COND_END]]
18485 // CHECK17: cond.end:
18486 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
18487 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
18488 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
18489 // CHECK17-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
18490 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
18491 // CHECK17: omp.inner.for.cond:
18492 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18493 // CHECK17-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18494 // CHECK17-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
18495 // CHECK17-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18496 // CHECK17: omp.inner.for.body:
18497 // CHECK17-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
18498 // CHECK17-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
18499 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18500 // CHECK17-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
18501 // CHECK17-NEXT: [[TMP20:%.*]] = load i32, i32* [[CONV]], align 8
18502 // CHECK17-NEXT: [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
18503 // CHECK17-NEXT: store i32 [[TMP20]], i32* [[CONV7]], align 4
18504 // CHECK17-NEXT: [[TMP21:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
18505 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP21]])
18506 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
18507 // CHECK17: omp.inner.for.inc:
18508 // CHECK17-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18509 // CHECK17-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
18510 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
18511 // CHECK17-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
18512 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
18513 // CHECK17: omp.inner.for.end:
18514 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
18515 // CHECK17: omp.loop.exit:
18516 // CHECK17-NEXT: [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18517 // CHECK17-NEXT: [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
18518 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
18519 // CHECK17-NEXT: br label [[OMP_PRECOND_END]]
18520 // CHECK17: omp.precond.end:
18521 // CHECK17-NEXT: ret void
18522 //
18523 //
18524 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..12
18525 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
18526 // CHECK17-NEXT: entry:
18527 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18528 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18529 // CHECK17-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
18530 // CHECK17-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
18531 // CHECK17-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
18532 // CHECK17-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
18533 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
18534 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
18535 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
18536 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
18537 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
18538 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
18539 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
18540 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
18541 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
18542 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18543 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18544 // CHECK17-NEXT: [[I6:%.*]] = alloca i32, align 4
18545 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18546 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18547 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
18548 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
18549 // CHECK17-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
18550 // CHECK17-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
18551 // CHECK17-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
18552 // CHECK17-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
18553 // CHECK17-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
18554 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
18555 // CHECK17-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
18556 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
18557 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
18558 // CHECK17-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
18559 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
18560 // CHECK17-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
18561 // CHECK17-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
18562 // CHECK17-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
18563 // CHECK17-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
18564 // CHECK17-NEXT: store i32 0, i32* [[I]], align 4
18565 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
18566 // CHECK17-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
18567 // CHECK17-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
18568 // CHECK17: omp.precond.then:
18569 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
18570 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
18571 // CHECK17-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
18572 // CHECK17-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
18573 // CHECK17-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
18574 // CHECK17-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
18575 // CHECK17-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
18576 // CHECK17-NEXT: store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
18577 // CHECK17-NEXT: store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
18578 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18579 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18580 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
18581 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18582 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18583 // CHECK17-NEXT: [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18584 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
18585 // CHECK17-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 35, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
18586 // CHECK17-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
18587 // CHECK17: omp.dispatch.cond:
18588 // CHECK17-NEXT: [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18589 // CHECK17-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
18590 // CHECK17-NEXT: [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
18591 // CHECK17-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
18592 // CHECK17-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
18593 // CHECK17: omp.dispatch.body:
18594 // CHECK17-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18595 // CHECK17-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
18596 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
18597 // CHECK17: omp.inner.for.cond:
18598 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
18599 // CHECK17-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
18600 // CHECK17-NEXT: [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
18601 // CHECK17-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18602 // CHECK17: omp.inner.for.body:
18603 // CHECK17-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
18604 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
18605 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
18606 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !18
18607 // CHECK17-NEXT: [[TMP21:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !18
18608 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
18609 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
18610 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
18611 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
18612 // CHECK17: omp.body.continue:
18613 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
18614 // CHECK17: omp.inner.for.inc:
18615 // CHECK17-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
18616 // CHECK17-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
18617 // CHECK17-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
18618 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
18619 // CHECK17: omp.inner.for.end:
18620 // CHECK17-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
18621 // CHECK17: omp.dispatch.inc:
18622 // CHECK17-NEXT: br label [[OMP_DISPATCH_COND]]
18623 // CHECK17: omp.dispatch.end:
18624 // CHECK17-NEXT: br label [[OMP_PRECOND_END]]
18625 // CHECK17: omp.precond.end:
18626 // CHECK17-NEXT: ret void
18627 //
18628 //
18629 // CHECK17-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
18630 // CHECK17-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
18631 // CHECK17-NEXT: entry:
18632 // CHECK17-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
18633 // CHECK17-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
18634 // CHECK17-NEXT: [[M:%.*]] = alloca i32, align 4
18635 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
18636 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
18637 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
18638 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
18639 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
18640 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
18641 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
18642 // CHECK17-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
18643 // CHECK17-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
18644 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
18645 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
18646 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
18647 // CHECK17-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
18648 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 8
18649 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 8
18650 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 8
18651 // CHECK17-NEXT: [[_TMP16:%.*]] = alloca i32, align 4
18652 // CHECK17-NEXT: [[M_CASTED19:%.*]] = alloca i64, align 8
18653 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [2 x i8*], align 8
18654 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS22:%.*]] = alloca [2 x i8*], align 8
18655 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [2 x i8*], align 8
18656 // CHECK17-NEXT: [[_TMP24:%.*]] = alloca i32, align 4
18657 // CHECK17-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
18658 // CHECK17-NEXT: store i32 10, i32* [[M]], align 4
18659 // CHECK17-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
18660 // CHECK17-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
18661 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
18662 // CHECK17-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
18663 // CHECK17-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
18664 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
18665 // CHECK17-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
18666 // CHECK17-NEXT: store i8* null, i8** [[TMP4]], align 8
18667 // CHECK17-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
18668 // CHECK17-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
18669 // CHECK17-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
18670 // CHECK17-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
18671 // CHECK17-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
18672 // CHECK17-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
18673 // CHECK17: omp_offload.failed:
18674 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116([10 x i32]* [[A]]) #[[ATTR3]]
18675 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT]]
18676 // CHECK17: omp_offload.cont:
18677 // CHECK17-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
18678 // CHECK17-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
18679 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 8
18680 // CHECK17-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
18681 // CHECK17-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
18682 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 8
18683 // CHECK17-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
18684 // CHECK17-NEXT: store i8* null, i8** [[TMP13]], align 8
18685 // CHECK17-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
18686 // CHECK17-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
18687 // CHECK17-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
18688 // CHECK17-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
18689 // CHECK17-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
18690 // CHECK17-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
18691 // CHECK17: omp_offload.failed5:
18692 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121([10 x i32]* [[A]]) #[[ATTR3]]
18693 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT6]]
18694 // CHECK17: omp_offload.cont6:
18695 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[M]], align 4
18696 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
18697 // CHECK17-NEXT: store i32 [[TMP18]], i32* [[CONV]], align 4
18698 // CHECK17-NEXT: [[TMP19:%.*]] = load i64, i64* [[M_CASTED]], align 8
18699 // CHECK17-NEXT: [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
18700 // CHECK17-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
18701 // CHECK17-NEXT: store i64 [[TMP19]], i64* [[TMP21]], align 8
18702 // CHECK17-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
18703 // CHECK17-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
18704 // CHECK17-NEXT: store i64 [[TMP19]], i64* [[TMP23]], align 8
18705 // CHECK17-NEXT: [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
18706 // CHECK17-NEXT: store i8* null, i8** [[TMP24]], align 8
18707 // CHECK17-NEXT: [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
18708 // CHECK17-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
18709 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 8
18710 // CHECK17-NEXT: [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
18711 // CHECK17-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
18712 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 8
18713 // CHECK17-NEXT: [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
18714 // CHECK17-NEXT: store i8* null, i8** [[TMP29]], align 8
18715 // CHECK17-NEXT: [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
18716 // CHECK17-NEXT: [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
18717 // CHECK17-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
18718 // CHECK17-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
18719 // CHECK17-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
18720 // CHECK17-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
18721 // CHECK17: omp_offload.failed11:
18722 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126(i64 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
18723 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT12]]
18724 // CHECK17: omp_offload.cont12:
18725 // CHECK17-NEXT: [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
18726 // CHECK17-NEXT: [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
18727 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 8
18728 // CHECK17-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
18729 // CHECK17-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
18730 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 8
18731 // CHECK17-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i64 0, i64 0
18732 // CHECK17-NEXT: store i8* null, i8** [[TMP38]], align 8
18733 // CHECK17-NEXT: [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
18734 // CHECK17-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
18735 // CHECK17-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
18736 // CHECK17-NEXT: [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
18737 // CHECK17-NEXT: [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
18738 // CHECK17-NEXT: br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
18739 // CHECK17: omp_offload.failed17:
18740 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131([10 x i32]* [[A]]) #[[ATTR3]]
18741 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT18]]
18742 // CHECK17: omp_offload.cont18:
18743 // CHECK17-NEXT: [[TMP43:%.*]] = load i32, i32* [[M]], align 4
18744 // CHECK17-NEXT: [[CONV20:%.*]] = bitcast i64* [[M_CASTED19]] to i32*
18745 // CHECK17-NEXT: store i32 [[TMP43]], i32* [[CONV20]], align 4
18746 // CHECK17-NEXT: [[TMP44:%.*]] = load i64, i64* [[M_CASTED19]], align 8
18747 // CHECK17-NEXT: [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
18748 // CHECK17-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
18749 // CHECK17-NEXT: store i64 [[TMP44]], i64* [[TMP46]], align 8
18750 // CHECK17-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
18751 // CHECK17-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
18752 // CHECK17-NEXT: store i64 [[TMP44]], i64* [[TMP48]], align 8
18753 // CHECK17-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
18754 // CHECK17-NEXT: store i8* null, i8** [[TMP49]], align 8
18755 // CHECK17-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
18756 // CHECK17-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
18757 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 8
18758 // CHECK17-NEXT: [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
18759 // CHECK17-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
18760 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 8
18761 // CHECK17-NEXT: [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
18762 // CHECK17-NEXT: store i8* null, i8** [[TMP54]], align 8
18763 // CHECK17-NEXT: [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
18764 // CHECK17-NEXT: [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
18765 // CHECK17-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
18766 // CHECK17-NEXT: [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
18767 // CHECK17-NEXT: [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
18768 // CHECK17-NEXT: br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED25:%.*]], label [[OMP_OFFLOAD_CONT26:%.*]]
18769 // CHECK17: omp_offload.failed25:
18770 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136(i64 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
18771 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT26]]
18772 // CHECK17: omp_offload.cont26:
18773 // CHECK17-NEXT: ret i32 0
18774 //
18775 //
18776 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116
18777 // CHECK17-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
18778 // CHECK17-NEXT: entry:
18779 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
18780 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
18781 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
18782 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
18783 // CHECK17-NEXT: ret void
18784 //
18785 //
18786 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..14
18787 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
18788 // CHECK17-NEXT: entry:
18789 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18790 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18791 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
18792 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
18793 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
18794 // CHECK17-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
18795 // CHECK17-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
18796 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18797 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18798 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
18799 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18800 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18801 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
18802 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
18803 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
18804 // CHECK17-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
18805 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18806 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18807 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18808 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
18809 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18810 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18811 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
18812 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18813 // CHECK17: cond.true:
18814 // CHECK17-NEXT: br label [[COND_END:%.*]]
18815 // CHECK17: cond.false:
18816 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18817 // CHECK17-NEXT: br label [[COND_END]]
18818 // CHECK17: cond.end:
18819 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
18820 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
18821 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
18822 // CHECK17-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
18823 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
18824 // CHECK17: omp.inner.for.cond:
18825 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18826 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18827 // CHECK17-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
18828 // CHECK17-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18829 // CHECK17: omp.inner.for.body:
18830 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
18831 // CHECK17-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
18832 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18833 // CHECK17-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
18834 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
18835 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
18836 // CHECK17: omp.inner.for.inc:
18837 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18838 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
18839 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
18840 // CHECK17-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
18841 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
18842 // CHECK17: omp.inner.for.end:
18843 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
18844 // CHECK17: omp.loop.exit:
18845 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
18846 // CHECK17-NEXT: ret void
18847 //
18848 //
18849 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..15
18850 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
18851 // CHECK17-NEXT: entry:
18852 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18853 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18854 // CHECK17-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
18855 // CHECK17-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
18856 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
18857 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
18858 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
18859 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
18860 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
18861 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18862 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18863 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
18864 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18865 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18866 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
18867 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
18868 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
18869 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
18870 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
18871 // CHECK17-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
18872 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
18873 // CHECK17-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
18874 // CHECK17-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
18875 // CHECK17-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
18876 // CHECK17-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
18877 // CHECK17-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
18878 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18879 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18880 // CHECK17-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18881 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
18882 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18883 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18884 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
18885 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18886 // CHECK17: cond.true:
18887 // CHECK17-NEXT: br label [[COND_END:%.*]]
18888 // CHECK17: cond.false:
18889 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18890 // CHECK17-NEXT: br label [[COND_END]]
18891 // CHECK17: cond.end:
18892 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
18893 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
18894 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
18895 // CHECK17-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
18896 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
18897 // CHECK17: omp.inner.for.cond:
18898 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18899 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
18900 // CHECK17-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
18901 // CHECK17-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18902 // CHECK17: omp.inner.for.body:
18903 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18904 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
18905 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
18906 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I]], align 4
18907 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
18908 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
18909 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
18910 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
18911 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
18912 // CHECK17: omp.body.continue:
18913 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
18914 // CHECK17: omp.inner.for.inc:
18915 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18916 // CHECK17-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
18917 // CHECK17-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
18918 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
18919 // CHECK17: omp.inner.for.end:
18920 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
18921 // CHECK17: omp.loop.exit:
18922 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
18923 // CHECK17-NEXT: ret void
18924 //
18925 //
18926 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121
18927 // CHECK17-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
18928 // CHECK17-NEXT: entry:
18929 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
18930 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
18931 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
18932 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
18933 // CHECK17-NEXT: ret void
18934 //
18935 //
18936 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..17
18937 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
18938 // CHECK17-NEXT: entry:
18939 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
18940 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
18941 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
18942 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
18943 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
18944 // CHECK17-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
18945 // CHECK17-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
18946 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
18947 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
18948 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
18949 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
18950 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
18951 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
18952 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
18953 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
18954 // CHECK17-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
18955 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
18956 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
18957 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
18958 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
18959 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
18960 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18961 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
18962 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
18963 // CHECK17: cond.true:
18964 // CHECK17-NEXT: br label [[COND_END:%.*]]
18965 // CHECK17: cond.false:
18966 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18967 // CHECK17-NEXT: br label [[COND_END]]
18968 // CHECK17: cond.end:
18969 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
18970 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
18971 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
18972 // CHECK17-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
18973 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
18974 // CHECK17: omp.inner.for.cond:
18975 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18976 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18977 // CHECK17-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
18978 // CHECK17-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
18979 // CHECK17: omp.inner.for.body:
18980 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
18981 // CHECK17-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
18982 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
18983 // CHECK17-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
18984 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
18985 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
18986 // CHECK17: omp.inner.for.inc:
18987 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
18988 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
18989 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
18990 // CHECK17-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
18991 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
18992 // CHECK17: omp.inner.for.end:
18993 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
18994 // CHECK17: omp.loop.exit:
18995 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
18996 // CHECK17-NEXT: ret void
18997 //
18998 //
18999 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..18
19000 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
19001 // CHECK17-NEXT: entry:
19002 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
19003 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
19004 // CHECK17-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
19005 // CHECK17-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
19006 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
19007 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
19008 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
19009 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
19010 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
19011 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19012 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19013 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
19014 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
19015 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
19016 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
19017 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
19018 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
19019 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
19020 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
19021 // CHECK17-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
19022 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
19023 // CHECK17-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
19024 // CHECK17-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
19025 // CHECK17-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
19026 // CHECK17-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
19027 // CHECK17-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
19028 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19029 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19030 // CHECK17-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
19031 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
19032 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
19033 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19034 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
19035 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19036 // CHECK17: cond.true:
19037 // CHECK17-NEXT: br label [[COND_END:%.*]]
19038 // CHECK17: cond.false:
19039 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19040 // CHECK17-NEXT: br label [[COND_END]]
19041 // CHECK17: cond.end:
19042 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
19043 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
19044 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19045 // CHECK17-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
19046 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
19047 // CHECK17: omp.inner.for.cond:
19048 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19049 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19050 // CHECK17-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
19051 // CHECK17-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19052 // CHECK17: omp.inner.for.body:
19053 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19054 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
19055 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
19056 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I]], align 4
19057 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
19058 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
19059 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
19060 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
19061 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
19062 // CHECK17: omp.body.continue:
19063 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
19064 // CHECK17: omp.inner.for.inc:
19065 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19066 // CHECK17-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
19067 // CHECK17-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
19068 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
19069 // CHECK17: omp.inner.for.end:
19070 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
19071 // CHECK17: omp.loop.exit:
19072 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
19073 // CHECK17-NEXT: ret void
19074 //
19075 //
19076 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126
19077 // CHECK17-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
19078 // CHECK17-NEXT: entry:
19079 // CHECK17-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
19080 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
19081 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
19082 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
19083 // CHECK17-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
19084 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
19085 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
19086 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
19087 // CHECK17-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
19088 // CHECK17-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
19089 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
19090 // CHECK17-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
19091 // CHECK17-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
19092 // CHECK17-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
19093 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
19094 // CHECK17-NEXT: ret void
19095 //
19096 //
19097 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..21
19098 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
19099 // CHECK17-NEXT: entry:
19100 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
19101 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
19102 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
19103 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
19104 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
19105 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
19106 // CHECK17-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
19107 // CHECK17-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
19108 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19109 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19110 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
19111 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
19112 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
19113 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
19114 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
19115 // CHECK17-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
19116 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
19117 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
19118 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
19119 // CHECK17-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
19120 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19121 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19122 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
19123 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
19124 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
19125 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
19126 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
19127 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19128 // CHECK17: cond.true:
19129 // CHECK17-NEXT: br label [[COND_END:%.*]]
19130 // CHECK17: cond.false:
19131 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
19132 // CHECK17-NEXT: br label [[COND_END]]
19133 // CHECK17: cond.end:
19134 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
19135 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
19136 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
19137 // CHECK17-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
19138 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
19139 // CHECK17: omp.inner.for.cond:
19140 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19141 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
19142 // CHECK17-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
19143 // CHECK17-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19144 // CHECK17: omp.inner.for.body:
19145 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
19146 // CHECK17-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
19147 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
19148 // CHECK17-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
19149 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8
19150 // CHECK17-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
19151 // CHECK17-NEXT: store i32 [[TMP12]], i32* [[CONV2]], align 4
19152 // CHECK17-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
19153 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
19154 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
19155 // CHECK17: omp.inner.for.inc:
19156 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19157 // CHECK17-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
19158 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
19159 // CHECK17-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
19160 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
19161 // CHECK17: omp.inner.for.end:
19162 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
19163 // CHECK17: omp.loop.exit:
19164 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
19165 // CHECK17-NEXT: ret void
19166 //
19167 //
19168 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..22
19169 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
19170 // CHECK17-NEXT: entry:
19171 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
19172 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
19173 // CHECK17-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
19174 // CHECK17-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
19175 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
19176 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
19177 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
19178 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
19179 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
19180 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
19181 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19182 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19183 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
19184 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
19185 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
19186 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
19187 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
19188 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
19189 // CHECK17-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
19190 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
19191 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
19192 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
19193 // CHECK17-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
19194 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
19195 // CHECK17-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
19196 // CHECK17-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
19197 // CHECK17-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
19198 // CHECK17-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
19199 // CHECK17-NEXT: store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
19200 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19201 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19202 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
19203 // CHECK17-NEXT: [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
19204 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
19205 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
19206 // CHECK17-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
19207 // CHECK17: omp.dispatch.cond:
19208 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19209 // CHECK17-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
19210 // CHECK17-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
19211 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
19212 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19213 // CHECK17: cond.true:
19214 // CHECK17-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
19215 // CHECK17-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
19216 // CHECK17-NEXT: br label [[COND_END:%.*]]
19217 // CHECK17: cond.false:
19218 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19219 // CHECK17-NEXT: br label [[COND_END]]
19220 // CHECK17: cond.end:
19221 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
19222 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
19223 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19224 // CHECK17-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
19225 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19226 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19227 // CHECK17-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
19228 // CHECK17-NEXT: br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
19229 // CHECK17: omp.dispatch.body:
19230 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
19231 // CHECK17: omp.inner.for.cond:
19232 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19233 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19234 // CHECK17-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
19235 // CHECK17-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19236 // CHECK17: omp.inner.for.body:
19237 // CHECK17-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19238 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
19239 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
19240 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I]], align 4
19241 // CHECK17-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4
19242 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
19243 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
19244 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
19245 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
19246 // CHECK17: omp.body.continue:
19247 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
19248 // CHECK17: omp.inner.for.inc:
19249 // CHECK17-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19250 // CHECK17-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
19251 // CHECK17-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
19252 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
19253 // CHECK17: omp.inner.for.end:
19254 // CHECK17-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
19255 // CHECK17: omp.dispatch.inc:
19256 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19257 // CHECK17-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
19258 // CHECK17-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
19259 // CHECK17-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
19260 // CHECK17-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19261 // CHECK17-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
19262 // CHECK17-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
19263 // CHECK17-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
19264 // CHECK17-NEXT: br label [[OMP_DISPATCH_COND]]
19265 // CHECK17: omp.dispatch.end:
19266 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
19267 // CHECK17-NEXT: ret void
19268 //
19269 //
19270 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
19271 // CHECK17-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
19272 // CHECK17-NEXT: entry:
19273 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
19274 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
19275 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
19276 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
19277 // CHECK17-NEXT: ret void
19278 //
19279 //
19280 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..25
19281 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
19282 // CHECK17-NEXT: entry:
19283 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
19284 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
19285 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
19286 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
19287 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
19288 // CHECK17-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
19289 // CHECK17-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
19290 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19291 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19292 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
19293 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
19294 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
19295 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
19296 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
19297 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
19298 // CHECK17-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
19299 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19300 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19301 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
19302 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
19303 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
19304 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
19305 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
19306 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19307 // CHECK17: cond.true:
19308 // CHECK17-NEXT: br label [[COND_END:%.*]]
19309 // CHECK17: cond.false:
19310 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
19311 // CHECK17-NEXT: br label [[COND_END]]
19312 // CHECK17: cond.end:
19313 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
19314 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
19315 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
19316 // CHECK17-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
19317 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
19318 // CHECK17: omp.inner.for.cond:
19319 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19320 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
19321 // CHECK17-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
19322 // CHECK17-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19323 // CHECK17: omp.inner.for.body:
19324 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
19325 // CHECK17-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
19326 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
19327 // CHECK17-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
19328 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
19329 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
19330 // CHECK17: omp.inner.for.inc:
19331 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19332 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
19333 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
19334 // CHECK17-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
19335 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
19336 // CHECK17: omp.inner.for.end:
19337 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
19338 // CHECK17: omp.loop.exit:
19339 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
19340 // CHECK17-NEXT: ret void
19341 //
19342 //
19343 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..26
19344 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
19345 // CHECK17-NEXT: entry:
19346 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
19347 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
19348 // CHECK17-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
19349 // CHECK17-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
19350 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
19351 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
19352 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
19353 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
19354 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
19355 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19356 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19357 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
19358 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
19359 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
19360 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
19361 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
19362 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
19363 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
19364 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
19365 // CHECK17-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
19366 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
19367 // CHECK17-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
19368 // CHECK17-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
19369 // CHECK17-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
19370 // CHECK17-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
19371 // CHECK17-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
19372 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19373 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19374 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19375 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19376 // CHECK17-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
19377 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
19378 // CHECK17-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
19379 // CHECK17-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
19380 // CHECK17: omp.dispatch.cond:
19381 // CHECK17-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
19382 // CHECK17-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
19383 // CHECK17-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
19384 // CHECK17: omp.dispatch.body:
19385 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19386 // CHECK17-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
19387 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
19388 // CHECK17: omp.inner.for.cond:
19389 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
19390 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
19391 // CHECK17-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
19392 // CHECK17-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19393 // CHECK17: omp.inner.for.body:
19394 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
19395 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
19396 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
19397 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !21
19398 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !21
19399 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
19400 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
19401 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
19402 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
19403 // CHECK17: omp.body.continue:
19404 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
19405 // CHECK17: omp.inner.for.inc:
19406 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
19407 // CHECK17-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
19408 // CHECK17-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
19409 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
19410 // CHECK17: omp.inner.for.end:
19411 // CHECK17-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
19412 // CHECK17: omp.dispatch.inc:
19413 // CHECK17-NEXT: br label [[OMP_DISPATCH_COND]]
19414 // CHECK17: omp.dispatch.end:
19415 // CHECK17-NEXT: ret void
19416 //
19417 //
19418 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136
19419 // CHECK17-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
19420 // CHECK17-NEXT: entry:
19421 // CHECK17-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
19422 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
19423 // CHECK17-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
19424 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
19425 // CHECK17-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
19426 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
19427 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
19428 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
19429 // CHECK17-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
19430 // CHECK17-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
19431 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
19432 // CHECK17-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
19433 // CHECK17-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
19434 // CHECK17-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
19435 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
19436 // CHECK17-NEXT: ret void
19437 //
19438 //
19439 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..29
19440 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
19441 // CHECK17-NEXT: entry:
19442 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
19443 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
19444 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
19445 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
19446 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
19447 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
19448 // CHECK17-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
19449 // CHECK17-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
19450 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19451 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19452 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
19453 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
19454 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
19455 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
19456 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
19457 // CHECK17-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
19458 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
19459 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
19460 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
19461 // CHECK17-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
19462 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19463 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19464 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
19465 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
19466 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
19467 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
19468 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
19469 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
19470 // CHECK17: cond.true:
19471 // CHECK17-NEXT: br label [[COND_END:%.*]]
19472 // CHECK17: cond.false:
19473 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
19474 // CHECK17-NEXT: br label [[COND_END]]
19475 // CHECK17: cond.end:
19476 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
19477 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
19478 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
19479 // CHECK17-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
19480 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
19481 // CHECK17: omp.inner.for.cond:
19482 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19483 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
19484 // CHECK17-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
19485 // CHECK17-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19486 // CHECK17: omp.inner.for.body:
19487 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
19488 // CHECK17-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
19489 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
19490 // CHECK17-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
19491 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8
19492 // CHECK17-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
19493 // CHECK17-NEXT: store i32 [[TMP12]], i32* [[CONV2]], align 4
19494 // CHECK17-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
19495 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
19496 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
19497 // CHECK17: omp.inner.for.inc:
19498 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
19499 // CHECK17-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
19500 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
19501 // CHECK17-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
19502 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
19503 // CHECK17: omp.inner.for.end:
19504 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
19505 // CHECK17: omp.loop.exit:
19506 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
19507 // CHECK17-NEXT: ret void
19508 //
19509 //
19510 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined..30
19511 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
19512 // CHECK17-NEXT: entry:
19513 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
19514 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
19515 // CHECK17-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
19516 // CHECK17-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
19517 // CHECK17-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
19518 // CHECK17-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
19519 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
19520 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
19521 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
19522 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
19523 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
19524 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
19525 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
19526 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
19527 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
19528 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
19529 // CHECK17-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
19530 // CHECK17-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
19531 // CHECK17-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
19532 // CHECK17-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
19533 // CHECK17-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
19534 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
19535 // CHECK17-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
19536 // CHECK17-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
19537 // CHECK17-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
19538 // CHECK17-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
19539 // CHECK17-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
19540 // CHECK17-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
19541 // CHECK17-NEXT: store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
19542 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
19543 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
19544 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
19545 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19546 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
19547 // CHECK17-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
19548 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
19549 // CHECK17-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 35, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
19550 // CHECK17-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
19551 // CHECK17: omp.dispatch.cond:
19552 // CHECK17-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
19553 // CHECK17-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
19554 // CHECK17-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
19555 // CHECK17: omp.dispatch.body:
19556 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
19557 // CHECK17-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
19558 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
19559 // CHECK17: omp.inner.for.cond:
19560 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
19561 // CHECK17-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !24
19562 // CHECK17-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
19563 // CHECK17-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
19564 // CHECK17: omp.inner.for.body:
19565 // CHECK17-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
19566 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
19567 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
19568 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !24
19569 // CHECK17-NEXT: [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !24
19570 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
19571 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
19572 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !24
19573 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
19574 // CHECK17: omp.body.continue:
19575 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
19576 // CHECK17: omp.inner.for.inc:
19577 // CHECK17-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
19578 // CHECK17-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
19579 // CHECK17-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
19580 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
19581 // CHECK17: omp.inner.for.end:
19582 // CHECK17-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
19583 // CHECK17: omp.dispatch.inc:
19584 // CHECK17-NEXT: br label [[OMP_DISPATCH_COND]]
19585 // CHECK17: omp.dispatch.end:
19586 // CHECK17-NEXT: ret void
19587 //
19588 //
19589 // CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
19590 // CHECK17-SAME: () #[[ATTR5:[0-9]+]] {
19591 // CHECK17-NEXT: entry:
19592 // CHECK17-NEXT: call void @__tgt_register_requires(i64 1)
19593 // CHECK17-NEXT: ret void
19594 //
19595 //
19596 // CHECK18-LABEL: define {{[^@]+}}@main
19597 // CHECK18-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
19598 // CHECK18-NEXT: entry:
19599 // CHECK18-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
19600 // CHECK18-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
19601 // CHECK18-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
19602 // CHECK18-NEXT: [[N:%.*]] = alloca i32, align 4
19603 // CHECK18-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
19604 // CHECK18-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
19605 // CHECK18-NEXT: [[M:%.*]] = alloca i32, align 4
19606 // CHECK18-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
19607 // CHECK18-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
19608 // CHECK18-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
19609 // CHECK18-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
19610 // CHECK18-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
19611 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
19612 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
19613 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
19614 // CHECK18-NEXT: [[N_CASTED3:%.*]] = alloca i64, align 8
19615 // CHECK18-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
19616 // CHECK18-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
19617 // CHECK18-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
19618 // CHECK18-NEXT: [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
19619 // CHECK18-NEXT: [[_TMP9:%.*]] = alloca i32, align 4
19620 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
19621 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
19622 // CHECK18-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
19623 // CHECK18-NEXT: [[N_CASTED19:%.*]] = alloca i64, align 8
19624 // CHECK18-NEXT: [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [4 x i8*], align 8
19625 // CHECK18-NEXT: [[DOTOFFLOAD_PTRS22:%.*]] = alloca [4 x i8*], align 8
19626 // CHECK18-NEXT: [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [4 x i8*], align 8
19627 // CHECK18-NEXT: [[DOTOFFLOAD_SIZES24:%.*]] = alloca [4 x i64], align 8
19628 // CHECK18-NEXT: [[_TMP25:%.*]] = alloca i32, align 4
19629 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
19630 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
19631 // CHECK18-NEXT: [[N_CASTED34:%.*]] = alloca i64, align 8
19632 // CHECK18-NEXT: [[DOTOFFLOAD_BASEPTRS36:%.*]] = alloca [3 x i8*], align 8
19633 // CHECK18-NEXT: [[DOTOFFLOAD_PTRS37:%.*]] = alloca [3 x i8*], align 8
19634 // CHECK18-NEXT: [[DOTOFFLOAD_MAPPERS38:%.*]] = alloca [3 x i8*], align 8
19635 // CHECK18-NEXT: [[DOTOFFLOAD_SIZES39:%.*]] = alloca [3 x i64], align 8
19636 // CHECK18-NEXT: [[_TMP40:%.*]] = alloca i32, align 4
19637 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
19638 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
19639 // CHECK18-NEXT: [[M_CASTED49:%.*]] = alloca i64, align 8
19640 // CHECK18-NEXT: [[N_CASTED51:%.*]] = alloca i64, align 8
19641 // CHECK18-NEXT: [[DOTOFFLOAD_BASEPTRS53:%.*]] = alloca [4 x i8*], align 8
19642 // CHECK18-NEXT: [[DOTOFFLOAD_PTRS54:%.*]] = alloca [4 x i8*], align 8
19643 // CHECK18-NEXT: [[DOTOFFLOAD_MAPPERS55:%.*]] = alloca [4 x i8*], align 8
19644 // CHECK18-NEXT: [[DOTOFFLOAD_SIZES56:%.*]] = alloca [4 x i64], align 8
19645 // CHECK18-NEXT: [[_TMP57:%.*]] = alloca i32, align 4
19646 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_58:%.*]] = alloca i32, align 4
19647 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_59:%.*]] = alloca i32, align 4
19648 // CHECK18-NEXT: store i32 0, i32* [[RETVAL]], align 4
19649 // CHECK18-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
19650 // CHECK18-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
19651 // CHECK18-NEXT: store i32 100, i32* [[N]], align 4
19652 // CHECK18-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
19653 // CHECK18-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
19654 // CHECK18-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
19655 // CHECK18-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
19656 // CHECK18-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
19657 // CHECK18-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
19658 // CHECK18-NEXT: store i32 10, i32* [[M]], align 4
19659 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
19660 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
19661 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
19662 // CHECK18-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
19663 // CHECK18-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
19664 // CHECK18-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
19665 // CHECK18-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
19666 // CHECK18-NEXT: store i64 [[TMP4]], i64* [[TMP7]], align 8
19667 // CHECK18-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
19668 // CHECK18-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
19669 // CHECK18-NEXT: store i64 [[TMP4]], i64* [[TMP9]], align 8
19670 // CHECK18-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
19671 // CHECK18-NEXT: store i64 4, i64* [[TMP10]], align 8
19672 // CHECK18-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
19673 // CHECK18-NEXT: store i8* null, i8** [[TMP11]], align 8
19674 // CHECK18-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
19675 // CHECK18-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
19676 // CHECK18-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
19677 // CHECK18-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
19678 // CHECK18-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
19679 // CHECK18-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
19680 // CHECK18-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
19681 // CHECK18-NEXT: store i64 8, i64* [[TMP16]], align 8
19682 // CHECK18-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
19683 // CHECK18-NEXT: store i8* null, i8** [[TMP17]], align 8
19684 // CHECK18-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
19685 // CHECK18-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
19686 // CHECK18-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 8
19687 // CHECK18-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
19688 // CHECK18-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
19689 // CHECK18-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 8
19690 // CHECK18-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
19691 // CHECK18-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 8
19692 // CHECK18-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
19693 // CHECK18-NEXT: store i8* null, i8** [[TMP23]], align 8
19694 // CHECK18-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
19695 // CHECK18-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
19696 // CHECK18-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
19697 // CHECK18-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
19698 // CHECK18-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
19699 // CHECK18-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
19700 // CHECK18-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
19701 // CHECK18-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
19702 // CHECK18-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
19703 // CHECK18-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
19704 // CHECK18-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
19705 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
19706 // CHECK18-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
19707 // CHECK18-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
19708 // CHECK18-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
19709 // CHECK18-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
19710 // CHECK18-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
19711 // CHECK18: omp_offload.failed:
19712 // CHECK18-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
19713 // CHECK18-NEXT: br label [[OMP_OFFLOAD_CONT]]
19714 // CHECK18: omp_offload.cont:
19715 // CHECK18-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
19716 // CHECK18-NEXT: [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
19717 // CHECK18-NEXT: store i32 [[TMP33]], i32* [[CONV4]], align 4
19718 // CHECK18-NEXT: [[TMP34:%.*]] = load i64, i64* [[N_CASTED3]], align 8
19719 // CHECK18-NEXT: [[TMP35:%.*]] = mul nuw i64 [[TMP1]], 4
19720 // CHECK18-NEXT: [[TMP36:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
19721 // CHECK18-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
19722 // CHECK18-NEXT: store i64 [[TMP34]], i64* [[TMP37]], align 8
19723 // CHECK18-NEXT: [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
19724 // CHECK18-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
19725 // CHECK18-NEXT: store i64 [[TMP34]], i64* [[TMP39]], align 8
19726 // CHECK18-NEXT: [[TMP40:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
19727 // CHECK18-NEXT: store i64 4, i64* [[TMP40]], align 8
19728 // CHECK18-NEXT: [[TMP41:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
19729 // CHECK18-NEXT: store i8* null, i8** [[TMP41]], align 8
19730 // CHECK18-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
19731 // CHECK18-NEXT: [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
19732 // CHECK18-NEXT: store i64 [[TMP1]], i64* [[TMP43]], align 8
19733 // CHECK18-NEXT: [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
19734 // CHECK18-NEXT: [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i64*
19735 // CHECK18-NEXT: store i64 [[TMP1]], i64* [[TMP45]], align 8
19736 // CHECK18-NEXT: [[TMP46:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 1
19737 // CHECK18-NEXT: store i64 8, i64* [[TMP46]], align 8
19738 // CHECK18-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
19739 // CHECK18-NEXT: store i8* null, i8** [[TMP47]], align 8
19740 // CHECK18-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
19741 // CHECK18-NEXT: [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32**
19742 // CHECK18-NEXT: store i32* [[VLA]], i32** [[TMP49]], align 8
19743 // CHECK18-NEXT: [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
19744 // CHECK18-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i32**
19745 // CHECK18-NEXT: store i32* [[VLA]], i32** [[TMP51]], align 8
19746 // CHECK18-NEXT: [[TMP52:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
19747 // CHECK18-NEXT: store i64 [[TMP35]], i64* [[TMP52]], align 8
19748 // CHECK18-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
19749 // CHECK18-NEXT: store i8* null, i8** [[TMP53]], align 8
19750 // CHECK18-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
19751 // CHECK18-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
19752 // CHECK18-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
19753 // CHECK18-NEXT: [[TMP57:%.*]] = load i32, i32* [[N]], align 4
19754 // CHECK18-NEXT: store i32 [[TMP57]], i32* [[DOTCAPTURE_EXPR_10]], align 4
19755 // CHECK18-NEXT: [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
19756 // CHECK18-NEXT: [[SUB12:%.*]] = sub nsw i32 [[TMP58]], 0
19757 // CHECK18-NEXT: [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
19758 // CHECK18-NEXT: [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
19759 // CHECK18-NEXT: store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
19760 // CHECK18-NEXT: [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
19761 // CHECK18-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP59]], 1
19762 // CHECK18-NEXT: [[TMP60:%.*]] = zext i32 [[ADD15]] to i64
19763 // CHECK18-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP60]])
19764 // CHECK18-NEXT: [[TMP61:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153.region_id, i32 3, i8** [[TMP54]], i8** [[TMP55]], i64* [[TMP56]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
19765 // CHECK18-NEXT: [[TMP62:%.*]] = icmp ne i32 [[TMP61]], 0
19766 // CHECK18-NEXT: br i1 [[TMP62]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
19767 // CHECK18: omp_offload.failed16:
19768 // CHECK18-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153(i64 [[TMP34]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
19769 // CHECK18-NEXT: br label [[OMP_OFFLOAD_CONT17]]
19770 // CHECK18: omp_offload.cont17:
19771 // CHECK18-NEXT: [[TMP63:%.*]] = load i32, i32* [[M]], align 4
19772 // CHECK18-NEXT: [[CONV18:%.*]] = bitcast i64* [[M_CASTED]] to i32*
19773 // CHECK18-NEXT: store i32 [[TMP63]], i32* [[CONV18]], align 4
19774 // CHECK18-NEXT: [[TMP64:%.*]] = load i64, i64* [[M_CASTED]], align 8
19775 // CHECK18-NEXT: [[TMP65:%.*]] = load i32, i32* [[N]], align 4
19776 // CHECK18-NEXT: [[CONV20:%.*]] = bitcast i64* [[N_CASTED19]] to i32*
19777 // CHECK18-NEXT: store i32 [[TMP65]], i32* [[CONV20]], align 4
19778 // CHECK18-NEXT: [[TMP66:%.*]] = load i64, i64* [[N_CASTED19]], align 8
19779 // CHECK18-NEXT: [[TMP67:%.*]] = mul nuw i64 [[TMP1]], 4
19780 // CHECK18-NEXT: [[TMP68:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
19781 // CHECK18-NEXT: [[TMP69:%.*]] = bitcast i8** [[TMP68]] to i64*
19782 // CHECK18-NEXT: store i64 [[TMP64]], i64* [[TMP69]], align 8
19783 // CHECK18-NEXT: [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
19784 // CHECK18-NEXT: [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i64*
19785 // CHECK18-NEXT: store i64 [[TMP64]], i64* [[TMP71]], align 8
19786 // CHECK18-NEXT: [[TMP72:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
19787 // CHECK18-NEXT: store i64 4, i64* [[TMP72]], align 8
19788 // CHECK18-NEXT: [[TMP73:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
19789 // CHECK18-NEXT: store i8* null, i8** [[TMP73]], align 8
19790 // CHECK18-NEXT: [[TMP74:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
19791 // CHECK18-NEXT: [[TMP75:%.*]] = bitcast i8** [[TMP74]] to i64*
19792 // CHECK18-NEXT: store i64 [[TMP66]], i64* [[TMP75]], align 8
19793 // CHECK18-NEXT: [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
19794 // CHECK18-NEXT: [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i64*
19795 // CHECK18-NEXT: store i64 [[TMP66]], i64* [[TMP77]], align 8
19796 // CHECK18-NEXT: [[TMP78:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 1
19797 // CHECK18-NEXT: store i64 4, i64* [[TMP78]], align 8
19798 // CHECK18-NEXT: [[TMP79:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
19799 // CHECK18-NEXT: store i8* null, i8** [[TMP79]], align 8
19800 // CHECK18-NEXT: [[TMP80:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 2
19801 // CHECK18-NEXT: [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
19802 // CHECK18-NEXT: store i64 [[TMP1]], i64* [[TMP81]], align 8
19803 // CHECK18-NEXT: [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 2
19804 // CHECK18-NEXT: [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
19805 // CHECK18-NEXT: store i64 [[TMP1]], i64* [[TMP83]], align 8
19806 // CHECK18-NEXT: [[TMP84:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 2
19807 // CHECK18-NEXT: store i64 8, i64* [[TMP84]], align 8
19808 // CHECK18-NEXT: [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 2
19809 // CHECK18-NEXT: store i8* null, i8** [[TMP85]], align 8
19810 // CHECK18-NEXT: [[TMP86:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 3
19811 // CHECK18-NEXT: [[TMP87:%.*]] = bitcast i8** [[TMP86]] to i32**
19812 // CHECK18-NEXT: store i32* [[VLA]], i32** [[TMP87]], align 8
19813 // CHECK18-NEXT: [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 3
19814 // CHECK18-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
19815 // CHECK18-NEXT: store i32* [[VLA]], i32** [[TMP89]], align 8
19816 // CHECK18-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 3
19817 // CHECK18-NEXT: store i64 [[TMP67]], i64* [[TMP90]], align 8
19818 // CHECK18-NEXT: [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 3
19819 // CHECK18-NEXT: store i8* null, i8** [[TMP91]], align 8
19820 // CHECK18-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
19821 // CHECK18-NEXT: [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
19822 // CHECK18-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
19823 // CHECK18-NEXT: [[TMP95:%.*]] = load i32, i32* [[N]], align 4
19824 // CHECK18-NEXT: store i32 [[TMP95]], i32* [[DOTCAPTURE_EXPR_26]], align 4
19825 // CHECK18-NEXT: [[TMP96:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_26]], align 4
19826 // CHECK18-NEXT: [[SUB28:%.*]] = sub nsw i32 [[TMP96]], 0
19827 // CHECK18-NEXT: [[DIV29:%.*]] = sdiv i32 [[SUB28]], 1
19828 // CHECK18-NEXT: [[SUB30:%.*]] = sub nsw i32 [[DIV29]], 1
19829 // CHECK18-NEXT: store i32 [[SUB30]], i32* [[DOTCAPTURE_EXPR_27]], align 4
19830 // CHECK18-NEXT: [[TMP97:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
19831 // CHECK18-NEXT: [[ADD31:%.*]] = add nsw i32 [[TMP97]], 1
19832 // CHECK18-NEXT: [[TMP98:%.*]] = zext i32 [[ADD31]] to i64
19833 // CHECK18-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP98]])
19834 // CHECK18-NEXT: [[TMP99:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158.region_id, i32 4, i8** [[TMP92]], i8** [[TMP93]], i64* [[TMP94]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
19835 // CHECK18-NEXT: [[TMP100:%.*]] = icmp ne i32 [[TMP99]], 0
19836 // CHECK18-NEXT: br i1 [[TMP100]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
19837 // CHECK18: omp_offload.failed32:
19838 // CHECK18-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158(i64 [[TMP64]], i64 [[TMP66]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
19839 // CHECK18-NEXT: br label [[OMP_OFFLOAD_CONT33]]
19840 // CHECK18: omp_offload.cont33:
19841 // CHECK18-NEXT: [[TMP101:%.*]] = load i32, i32* [[N]], align 4
19842 // CHECK18-NEXT: [[CONV35:%.*]] = bitcast i64* [[N_CASTED34]] to i32*
19843 // CHECK18-NEXT: store i32 [[TMP101]], i32* [[CONV35]], align 4
19844 // CHECK18-NEXT: [[TMP102:%.*]] = load i64, i64* [[N_CASTED34]], align 8
19845 // CHECK18-NEXT: [[TMP103:%.*]] = mul nuw i64 [[TMP1]], 4
19846 // CHECK18-NEXT: [[TMP104:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
19847 // CHECK18-NEXT: [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i64*
19848 // CHECK18-NEXT: store i64 [[TMP102]], i64* [[TMP105]], align 8
19849 // CHECK18-NEXT: [[TMP106:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
19850 // CHECK18-NEXT: [[TMP107:%.*]] = bitcast i8** [[TMP106]] to i64*
19851 // CHECK18-NEXT: store i64 [[TMP102]], i64* [[TMP107]], align 8
19852 // CHECK18-NEXT: [[TMP108:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
19853 // CHECK18-NEXT: store i64 4, i64* [[TMP108]], align 8
19854 // CHECK18-NEXT: [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 0
19855 // CHECK18-NEXT: store i8* null, i8** [[TMP109]], align 8
19856 // CHECK18-NEXT: [[TMP110:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 1
19857 // CHECK18-NEXT: [[TMP111:%.*]] = bitcast i8** [[TMP110]] to i64*
19858 // CHECK18-NEXT: store i64 [[TMP1]], i64* [[TMP111]], align 8
19859 // CHECK18-NEXT: [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 1
19860 // CHECK18-NEXT: [[TMP113:%.*]] = bitcast i8** [[TMP112]] to i64*
19861 // CHECK18-NEXT: store i64 [[TMP1]], i64* [[TMP113]], align 8
19862 // CHECK18-NEXT: [[TMP114:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 1
19863 // CHECK18-NEXT: store i64 8, i64* [[TMP114]], align 8
19864 // CHECK18-NEXT: [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 1
19865 // CHECK18-NEXT: store i8* null, i8** [[TMP115]], align 8
19866 // CHECK18-NEXT: [[TMP116:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 2
19867 // CHECK18-NEXT: [[TMP117:%.*]] = bitcast i8** [[TMP116]] to i32**
19868 // CHECK18-NEXT: store i32* [[VLA]], i32** [[TMP117]], align 8
19869 // CHECK18-NEXT: [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 2
19870 // CHECK18-NEXT: [[TMP119:%.*]] = bitcast i8** [[TMP118]] to i32**
19871 // CHECK18-NEXT: store i32* [[VLA]], i32** [[TMP119]], align 8
19872 // CHECK18-NEXT: [[TMP120:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 2
19873 // CHECK18-NEXT: store i64 [[TMP103]], i64* [[TMP120]], align 8
19874 // CHECK18-NEXT: [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS38]], i64 0, i64 2
19875 // CHECK18-NEXT: store i8* null, i8** [[TMP121]], align 8
19876 // CHECK18-NEXT: [[TMP122:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS36]], i32 0, i32 0
19877 // CHECK18-NEXT: [[TMP123:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS37]], i32 0, i32 0
19878 // CHECK18-NEXT: [[TMP124:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES39]], i32 0, i32 0
19879 // CHECK18-NEXT: [[TMP125:%.*]] = load i32, i32* [[N]], align 4
19880 // CHECK18-NEXT: store i32 [[TMP125]], i32* [[DOTCAPTURE_EXPR_41]], align 4
19881 // CHECK18-NEXT: [[TMP126:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
19882 // CHECK18-NEXT: [[SUB43:%.*]] = sub nsw i32 [[TMP126]], 0
19883 // CHECK18-NEXT: [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
19884 // CHECK18-NEXT: [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
19885 // CHECK18-NEXT: store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
19886 // CHECK18-NEXT: [[TMP127:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
19887 // CHECK18-NEXT: [[ADD46:%.*]] = add nsw i32 [[TMP127]], 1
19888 // CHECK18-NEXT: [[TMP128:%.*]] = zext i32 [[ADD46]] to i64
19889 // CHECK18-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP128]])
19890 // CHECK18-NEXT: [[TMP129:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163.region_id, i32 3, i8** [[TMP122]], i8** [[TMP123]], i64* [[TMP124]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
19891 // CHECK18-NEXT: [[TMP130:%.*]] = icmp ne i32 [[TMP129]], 0
19892 // CHECK18-NEXT: br i1 [[TMP130]], label [[OMP_OFFLOAD_FAILED47:%.*]], label [[OMP_OFFLOAD_CONT48:%.*]]
19893 // CHECK18: omp_offload.failed47:
19894 // CHECK18-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163(i64 [[TMP102]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
19895 // CHECK18-NEXT: br label [[OMP_OFFLOAD_CONT48]]
19896 // CHECK18: omp_offload.cont48:
19897 // CHECK18-NEXT: [[TMP131:%.*]] = load i32, i32* [[M]], align 4
19898 // CHECK18-NEXT: [[CONV50:%.*]] = bitcast i64* [[M_CASTED49]] to i32*
19899 // CHECK18-NEXT: store i32 [[TMP131]], i32* [[CONV50]], align 4
19900 // CHECK18-NEXT: [[TMP132:%.*]] = load i64, i64* [[M_CASTED49]], align 8
19901 // CHECK18-NEXT: [[TMP133:%.*]] = load i32, i32* [[N]], align 4
19902 // CHECK18-NEXT: [[CONV52:%.*]] = bitcast i64* [[N_CASTED51]] to i32*
19903 // CHECK18-NEXT: store i32 [[TMP133]], i32* [[CONV52]], align 4
19904 // CHECK18-NEXT: [[TMP134:%.*]] = load i64, i64* [[N_CASTED51]], align 8
19905 // CHECK18-NEXT: [[TMP135:%.*]] = mul nuw i64 [[TMP1]], 4
19906 // CHECK18-NEXT: [[TMP136:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
19907 // CHECK18-NEXT: [[TMP137:%.*]] = bitcast i8** [[TMP136]] to i64*
19908 // CHECK18-NEXT: store i64 [[TMP132]], i64* [[TMP137]], align 8
19909 // CHECK18-NEXT: [[TMP138:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
19910 // CHECK18-NEXT: [[TMP139:%.*]] = bitcast i8** [[TMP138]] to i64*
19911 // CHECK18-NEXT: store i64 [[TMP132]], i64* [[TMP139]], align 8
19912 // CHECK18-NEXT: [[TMP140:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
19913 // CHECK18-NEXT: store i64 4, i64* [[TMP140]], align 8
19914 // CHECK18-NEXT: [[TMP141:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 0
19915 // CHECK18-NEXT: store i8* null, i8** [[TMP141]], align 8
19916 // CHECK18-NEXT: [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 1
19917 // CHECK18-NEXT: [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i64*
19918 // CHECK18-NEXT: store i64 [[TMP134]], i64* [[TMP143]], align 8
19919 // CHECK18-NEXT: [[TMP144:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 1
19920 // CHECK18-NEXT: [[TMP145:%.*]] = bitcast i8** [[TMP144]] to i64*
19921 // CHECK18-NEXT: store i64 [[TMP134]], i64* [[TMP145]], align 8
19922 // CHECK18-NEXT: [[TMP146:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 1
19923 // CHECK18-NEXT: store i64 4, i64* [[TMP146]], align 8
19924 // CHECK18-NEXT: [[TMP147:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 1
19925 // CHECK18-NEXT: store i8* null, i8** [[TMP147]], align 8
19926 // CHECK18-NEXT: [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 2
19927 // CHECK18-NEXT: [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i64*
19928 // CHECK18-NEXT: store i64 [[TMP1]], i64* [[TMP149]], align 8
19929 // CHECK18-NEXT: [[TMP150:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 2
19930 // CHECK18-NEXT: [[TMP151:%.*]] = bitcast i8** [[TMP150]] to i64*
19931 // CHECK18-NEXT: store i64 [[TMP1]], i64* [[TMP151]], align 8
19932 // CHECK18-NEXT: [[TMP152:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 2
19933 // CHECK18-NEXT: store i64 8, i64* [[TMP152]], align 8
19934 // CHECK18-NEXT: [[TMP153:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 2
19935 // CHECK18-NEXT: store i8* null, i8** [[TMP153]], align 8
19936 // CHECK18-NEXT: [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 3
19937 // CHECK18-NEXT: [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32**
19938 // CHECK18-NEXT: store i32* [[VLA]], i32** [[TMP155]], align 8
19939 // CHECK18-NEXT: [[TMP156:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 3
19940 // CHECK18-NEXT: [[TMP157:%.*]] = bitcast i8** [[TMP156]] to i32**
19941 // CHECK18-NEXT: store i32* [[VLA]], i32** [[TMP157]], align 8
19942 // CHECK18-NEXT: [[TMP158:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 3
19943 // CHECK18-NEXT: store i64 [[TMP135]], i64* [[TMP158]], align 8
19944 // CHECK18-NEXT: [[TMP159:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS55]], i64 0, i64 3
19945 // CHECK18-NEXT: store i8* null, i8** [[TMP159]], align 8
19946 // CHECK18-NEXT: [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS53]], i32 0, i32 0
19947 // CHECK18-NEXT: [[TMP161:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS54]], i32 0, i32 0
19948 // CHECK18-NEXT: [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES56]], i32 0, i32 0
19949 // CHECK18-NEXT: [[TMP163:%.*]] = load i32, i32* [[N]], align 4
19950 // CHECK18-NEXT: store i32 [[TMP163]], i32* [[DOTCAPTURE_EXPR_58]], align 4
19951 // CHECK18-NEXT: [[TMP164:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_58]], align 4
19952 // CHECK18-NEXT: [[SUB60:%.*]] = sub nsw i32 [[TMP164]], 0
19953 // CHECK18-NEXT: [[DIV61:%.*]] = sdiv i32 [[SUB60]], 1
19954 // CHECK18-NEXT: [[SUB62:%.*]] = sub nsw i32 [[DIV61]], 1
19955 // CHECK18-NEXT: store i32 [[SUB62]], i32* [[DOTCAPTURE_EXPR_59]], align 4
19956 // CHECK18-NEXT: [[TMP165:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_59]], align 4
19957 // CHECK18-NEXT: [[ADD63:%.*]] = add nsw i32 [[TMP165]], 1
19958 // CHECK18-NEXT: [[TMP166:%.*]] = zext i32 [[ADD63]] to i64
19959 // CHECK18-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP166]])
19960 // CHECK18-NEXT: [[TMP167:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168.region_id, i32 4, i8** [[TMP160]], i8** [[TMP161]], i64* [[TMP162]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
19961 // CHECK18-NEXT: [[TMP168:%.*]] = icmp ne i32 [[TMP167]], 0
19962 // CHECK18-NEXT: br i1 [[TMP168]], label [[OMP_OFFLOAD_FAILED64:%.*]], label [[OMP_OFFLOAD_CONT65:%.*]]
19963 // CHECK18: omp_offload.failed64:
19964 // CHECK18-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168(i64 [[TMP132]], i64 [[TMP134]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
19965 // CHECK18-NEXT: br label [[OMP_OFFLOAD_CONT65]]
19966 // CHECK18: omp_offload.cont65:
19967 // CHECK18-NEXT: [[TMP169:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
19968 // CHECK18-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP169]])
19969 // CHECK18-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
19970 // CHECK18-NEXT: [[TMP170:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
19971 // CHECK18-NEXT: call void @llvm.stackrestore(i8* [[TMP170]])
19972 // CHECK18-NEXT: [[TMP171:%.*]] = load i32, i32* [[RETVAL]], align 4
19973 // CHECK18-NEXT: ret i32 [[TMP171]]
19974 //
19975 //
19976 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148
19977 // CHECK18-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
19978 // CHECK18-NEXT: entry:
19979 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
19980 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
19981 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
19982 // CHECK18-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
19983 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
19984 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
19985 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
19986 // CHECK18-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
19987 // CHECK18-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
19988 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
19989 // CHECK18-NEXT: ret void
19990 //
19991 //
19992 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined.
19993 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
19994 // CHECK18-NEXT: entry:
19995 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
19996 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
19997 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
19998 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
19999 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20000 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
20001 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
20002 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
20003 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
20004 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
20005 // CHECK18-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
20006 // CHECK18-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
20007 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
20008 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
20009 // CHECK18-NEXT: [[I3:%.*]] = alloca i32, align 4
20010 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
20011 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
20012 // CHECK18-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
20013 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20014 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20015 // CHECK18-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
20016 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20017 // CHECK18-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20018 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
20019 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
20020 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20021 // CHECK18-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
20022 // CHECK18-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
20023 // CHECK18-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
20024 // CHECK18-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
20025 // CHECK18-NEXT: store i32 0, i32* [[I]], align 4
20026 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20027 // CHECK18-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
20028 // CHECK18-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
20029 // CHECK18: omp.precond.then:
20030 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
20031 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20032 // CHECK18-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
20033 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
20034 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
20035 // CHECK18-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20036 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
20037 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
20038 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20039 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20040 // CHECK18-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
20041 // CHECK18-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
20042 // CHECK18: cond.true:
20043 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20044 // CHECK18-NEXT: br label [[COND_END:%.*]]
20045 // CHECK18: cond.false:
20046 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20047 // CHECK18-NEXT: br label [[COND_END]]
20048 // CHECK18: cond.end:
20049 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
20050 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
20051 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
20052 // CHECK18-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
20053 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
20054 // CHECK18: omp.inner.for.cond:
20055 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20056 // CHECK18-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20057 // CHECK18-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
20058 // CHECK18-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
20059 // CHECK18: omp.inner.for.body:
20060 // CHECK18-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
20061 // CHECK18-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
20062 // CHECK18-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20063 // CHECK18-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
20064 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
20065 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
20066 // CHECK18: omp.inner.for.inc:
20067 // CHECK18-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20068 // CHECK18-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
20069 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
20070 // CHECK18-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
20071 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
20072 // CHECK18: omp.inner.for.end:
20073 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
20074 // CHECK18: omp.loop.exit:
20075 // CHECK18-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20076 // CHECK18-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
20077 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
20078 // CHECK18-NEXT: br label [[OMP_PRECOND_END]]
20079 // CHECK18: omp.precond.end:
20080 // CHECK18-NEXT: ret void
20081 //
20082 //
20083 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..1
20084 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
20085 // CHECK18-NEXT: entry:
20086 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
20087 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
20088 // CHECK18-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
20089 // CHECK18-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
20090 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
20091 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
20092 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20093 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
20094 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
20095 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
20096 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
20097 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
20098 // CHECK18-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
20099 // CHECK18-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
20100 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
20101 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
20102 // CHECK18-NEXT: [[I4:%.*]] = alloca i32, align 4
20103 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
20104 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
20105 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
20106 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
20107 // CHECK18-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
20108 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20109 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20110 // CHECK18-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
20111 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20112 // CHECK18-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20113 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
20114 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
20115 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20116 // CHECK18-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
20117 // CHECK18-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
20118 // CHECK18-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
20119 // CHECK18-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
20120 // CHECK18-NEXT: store i32 0, i32* [[I]], align 4
20121 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20122 // CHECK18-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
20123 // CHECK18-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
20124 // CHECK18: omp.precond.then:
20125 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
20126 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20127 // CHECK18-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
20128 // CHECK18-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
20129 // CHECK18-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
20130 // CHECK18-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
20131 // CHECK18-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
20132 // CHECK18-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
20133 // CHECK18-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
20134 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
20135 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
20136 // CHECK18-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20137 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
20138 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
20139 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20140 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20141 // CHECK18-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
20142 // CHECK18-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
20143 // CHECK18: cond.true:
20144 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20145 // CHECK18-NEXT: br label [[COND_END:%.*]]
20146 // CHECK18: cond.false:
20147 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20148 // CHECK18-NEXT: br label [[COND_END]]
20149 // CHECK18: cond.end:
20150 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
20151 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
20152 // CHECK18-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
20153 // CHECK18-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
20154 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
20155 // CHECK18: omp.inner.for.cond:
20156 // CHECK18-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20157 // CHECK18-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20158 // CHECK18-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
20159 // CHECK18-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
20160 // CHECK18: omp.inner.for.body:
20161 // CHECK18-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20162 // CHECK18-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
20163 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
20164 // CHECK18-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
20165 // CHECK18-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
20166 // CHECK18-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
20167 // CHECK18-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
20168 // CHECK18-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
20169 // CHECK18-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
20170 // CHECK18: omp.body.continue:
20171 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
20172 // CHECK18: omp.inner.for.inc:
20173 // CHECK18-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20174 // CHECK18-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
20175 // CHECK18-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
20176 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
20177 // CHECK18: omp.inner.for.end:
20178 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
20179 // CHECK18: omp.loop.exit:
20180 // CHECK18-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20181 // CHECK18-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
20182 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
20183 // CHECK18-NEXT: br label [[OMP_PRECOND_END]]
20184 // CHECK18: omp.precond.end:
20185 // CHECK18-NEXT: ret void
20186 //
20187 //
20188 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153
20189 // CHECK18-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
20190 // CHECK18-NEXT: entry:
20191 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
20192 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
20193 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20194 // CHECK18-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
20195 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20196 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20197 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
20198 // CHECK18-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20199 // CHECK18-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20200 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
20201 // CHECK18-NEXT: ret void
20202 //
20203 //
20204 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..2
20205 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
20206 // CHECK18-NEXT: entry:
20207 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
20208 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
20209 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
20210 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
20211 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20212 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
20213 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
20214 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
20215 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
20216 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
20217 // CHECK18-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
20218 // CHECK18-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
20219 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
20220 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
20221 // CHECK18-NEXT: [[I3:%.*]] = alloca i32, align 4
20222 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
20223 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
20224 // CHECK18-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
20225 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20226 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20227 // CHECK18-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
20228 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20229 // CHECK18-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20230 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
20231 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
20232 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20233 // CHECK18-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
20234 // CHECK18-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
20235 // CHECK18-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
20236 // CHECK18-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
20237 // CHECK18-NEXT: store i32 0, i32* [[I]], align 4
20238 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20239 // CHECK18-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
20240 // CHECK18-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
20241 // CHECK18: omp.precond.then:
20242 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
20243 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20244 // CHECK18-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
20245 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
20246 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
20247 // CHECK18-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20248 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
20249 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
20250 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20251 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20252 // CHECK18-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
20253 // CHECK18-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
20254 // CHECK18: cond.true:
20255 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20256 // CHECK18-NEXT: br label [[COND_END:%.*]]
20257 // CHECK18: cond.false:
20258 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20259 // CHECK18-NEXT: br label [[COND_END]]
20260 // CHECK18: cond.end:
20261 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
20262 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
20263 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
20264 // CHECK18-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
20265 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
20266 // CHECK18: omp.inner.for.cond:
20267 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20268 // CHECK18-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20269 // CHECK18-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
20270 // CHECK18-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
20271 // CHECK18: omp.inner.for.body:
20272 // CHECK18-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
20273 // CHECK18-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
20274 // CHECK18-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20275 // CHECK18-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
20276 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
20277 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
20278 // CHECK18: omp.inner.for.inc:
20279 // CHECK18-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20280 // CHECK18-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
20281 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
20282 // CHECK18-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
20283 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
20284 // CHECK18: omp.inner.for.end:
20285 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
20286 // CHECK18: omp.loop.exit:
20287 // CHECK18-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20288 // CHECK18-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
20289 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
20290 // CHECK18-NEXT: br label [[OMP_PRECOND_END]]
20291 // CHECK18: omp.precond.end:
20292 // CHECK18-NEXT: ret void
20293 //
20294 //
20295 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..3
20296 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
20297 // CHECK18-NEXT: entry:
20298 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
20299 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
20300 // CHECK18-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
20301 // CHECK18-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
20302 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
20303 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
20304 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20305 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
20306 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
20307 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
20308 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
20309 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
20310 // CHECK18-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
20311 // CHECK18-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
20312 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
20313 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
20314 // CHECK18-NEXT: [[I4:%.*]] = alloca i32, align 4
20315 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
20316 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
20317 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
20318 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
20319 // CHECK18-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
20320 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20321 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20322 // CHECK18-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
20323 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20324 // CHECK18-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20325 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
20326 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
20327 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20328 // CHECK18-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
20329 // CHECK18-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
20330 // CHECK18-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
20331 // CHECK18-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
20332 // CHECK18-NEXT: store i32 0, i32* [[I]], align 4
20333 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20334 // CHECK18-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
20335 // CHECK18-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
20336 // CHECK18: omp.precond.then:
20337 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
20338 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20339 // CHECK18-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
20340 // CHECK18-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
20341 // CHECK18-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
20342 // CHECK18-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
20343 // CHECK18-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
20344 // CHECK18-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
20345 // CHECK18-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
20346 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
20347 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
20348 // CHECK18-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20349 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
20350 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
20351 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20352 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20353 // CHECK18-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
20354 // CHECK18-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
20355 // CHECK18: cond.true:
20356 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20357 // CHECK18-NEXT: br label [[COND_END:%.*]]
20358 // CHECK18: cond.false:
20359 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20360 // CHECK18-NEXT: br label [[COND_END]]
20361 // CHECK18: cond.end:
20362 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
20363 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
20364 // CHECK18-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
20365 // CHECK18-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
20366 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
20367 // CHECK18: omp.inner.for.cond:
20368 // CHECK18-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20369 // CHECK18-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20370 // CHECK18-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
20371 // CHECK18-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
20372 // CHECK18: omp.inner.for.body:
20373 // CHECK18-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20374 // CHECK18-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
20375 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
20376 // CHECK18-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
20377 // CHECK18-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
20378 // CHECK18-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
20379 // CHECK18-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
20380 // CHECK18-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
20381 // CHECK18-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
20382 // CHECK18: omp.body.continue:
20383 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
20384 // CHECK18: omp.inner.for.inc:
20385 // CHECK18-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20386 // CHECK18-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
20387 // CHECK18-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
20388 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
20389 // CHECK18: omp.inner.for.end:
20390 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
20391 // CHECK18: omp.loop.exit:
20392 // CHECK18-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20393 // CHECK18-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
20394 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
20395 // CHECK18-NEXT: br label [[OMP_PRECOND_END]]
20396 // CHECK18: omp.precond.end:
20397 // CHECK18-NEXT: ret void
20398 //
20399 //
20400 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158
20401 // CHECK18-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
20402 // CHECK18-NEXT: entry:
20403 // CHECK18-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
20404 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
20405 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
20406 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20407 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
20408 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
20409 // CHECK18-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
20410 // CHECK18-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
20411 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20412 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20413 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
20414 // CHECK18-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
20415 // CHECK18-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20416 // CHECK18-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20417 // CHECK18-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
20418 // CHECK18-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
20419 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20420 // CHECK18-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
20421 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
20422 // CHECK18-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
20423 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
20424 // CHECK18-NEXT: ret void
20425 //
20426 //
20427 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..5
20428 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
20429 // CHECK18-NEXT: entry:
20430 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
20431 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
20432 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
20433 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
20434 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20435 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
20436 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
20437 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
20438 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
20439 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
20440 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
20441 // CHECK18-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
20442 // CHECK18-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
20443 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
20444 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
20445 // CHECK18-NEXT: [[I4:%.*]] = alloca i32, align 4
20446 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
20447 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
20448 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
20449 // CHECK18-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
20450 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20451 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20452 // CHECK18-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
20453 // CHECK18-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
20454 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20455 // CHECK18-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20456 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
20457 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
20458 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
20459 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20460 // CHECK18-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
20461 // CHECK18-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
20462 // CHECK18-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
20463 // CHECK18-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
20464 // CHECK18-NEXT: store i32 0, i32* [[I]], align 4
20465 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20466 // CHECK18-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
20467 // CHECK18-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
20468 // CHECK18: omp.precond.then:
20469 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
20470 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20471 // CHECK18-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
20472 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
20473 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
20474 // CHECK18-NEXT: [[TMP7:%.*]] = load i32, i32* [[CONV]], align 8
20475 // CHECK18-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20476 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
20477 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
20478 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20479 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20480 // CHECK18-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
20481 // CHECK18-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
20482 // CHECK18: cond.true:
20483 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20484 // CHECK18-NEXT: br label [[COND_END:%.*]]
20485 // CHECK18: cond.false:
20486 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20487 // CHECK18-NEXT: br label [[COND_END]]
20488 // CHECK18: cond.end:
20489 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
20490 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
20491 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
20492 // CHECK18-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
20493 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
20494 // CHECK18: omp.inner.for.cond:
20495 // CHECK18-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20496 // CHECK18-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20497 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
20498 // CHECK18-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
20499 // CHECK18-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
20500 // CHECK18: omp.inner.for.body:
20501 // CHECK18-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
20502 // CHECK18-NEXT: [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
20503 // CHECK18-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20504 // CHECK18-NEXT: [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
20505 // CHECK18-NEXT: [[TMP21:%.*]] = load i32, i32* [[CONV]], align 8
20506 // CHECK18-NEXT: [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
20507 // CHECK18-NEXT: store i32 [[TMP21]], i32* [[CONV7]], align 4
20508 // CHECK18-NEXT: [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
20509 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]])
20510 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
20511 // CHECK18: omp.inner.for.inc:
20512 // CHECK18-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20513 // CHECK18-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
20514 // CHECK18-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
20515 // CHECK18-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
20516 // CHECK18-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
20517 // CHECK18-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
20518 // CHECK18-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
20519 // CHECK18-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4
20520 // CHECK18-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20521 // CHECK18-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
20522 // CHECK18-NEXT: [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
20523 // CHECK18-NEXT: store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4
20524 // CHECK18-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20525 // CHECK18-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20526 // CHECK18-NEXT: [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
20527 // CHECK18-NEXT: br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
20528 // CHECK18: cond.true12:
20529 // CHECK18-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20530 // CHECK18-NEXT: br label [[COND_END14:%.*]]
20531 // CHECK18: cond.false13:
20532 // CHECK18-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20533 // CHECK18-NEXT: br label [[COND_END14]]
20534 // CHECK18: cond.end14:
20535 // CHECK18-NEXT: [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
20536 // CHECK18-NEXT: store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4
20537 // CHECK18-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
20538 // CHECK18-NEXT: store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4
20539 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
20540 // CHECK18: omp.inner.for.end:
20541 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
20542 // CHECK18: omp.loop.exit:
20543 // CHECK18-NEXT: [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20544 // CHECK18-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
20545 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
20546 // CHECK18-NEXT: br label [[OMP_PRECOND_END]]
20547 // CHECK18: omp.precond.end:
20548 // CHECK18-NEXT: ret void
20549 //
20550 //
20551 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..6
20552 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
20553 // CHECK18-NEXT: entry:
20554 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
20555 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
20556 // CHECK18-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
20557 // CHECK18-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
20558 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
20559 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
20560 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20561 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
20562 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
20563 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
20564 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
20565 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
20566 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
20567 // CHECK18-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
20568 // CHECK18-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
20569 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
20570 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
20571 // CHECK18-NEXT: [[I6:%.*]] = alloca i32, align 4
20572 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
20573 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
20574 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
20575 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
20576 // CHECK18-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
20577 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20578 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20579 // CHECK18-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
20580 // CHECK18-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
20581 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20582 // CHECK18-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20583 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
20584 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
20585 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
20586 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20587 // CHECK18-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
20588 // CHECK18-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
20589 // CHECK18-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
20590 // CHECK18-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
20591 // CHECK18-NEXT: store i32 0, i32* [[I]], align 4
20592 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20593 // CHECK18-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
20594 // CHECK18-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
20595 // CHECK18: omp.precond.then:
20596 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
20597 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20598 // CHECK18-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
20599 // CHECK18-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
20600 // CHECK18-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
20601 // CHECK18-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
20602 // CHECK18-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
20603 // CHECK18-NEXT: store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
20604 // CHECK18-NEXT: store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
20605 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
20606 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
20607 // CHECK18-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20608 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
20609 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
20610 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20611 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20612 // CHECK18-NEXT: [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
20613 // CHECK18-NEXT: br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
20614 // CHECK18: cond.true:
20615 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20616 // CHECK18-NEXT: br label [[COND_END:%.*]]
20617 // CHECK18: cond.false:
20618 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20619 // CHECK18-NEXT: br label [[COND_END]]
20620 // CHECK18: cond.end:
20621 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
20622 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
20623 // CHECK18-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
20624 // CHECK18-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
20625 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
20626 // CHECK18: omp.inner.for.cond:
20627 // CHECK18-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20628 // CHECK18-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20629 // CHECK18-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
20630 // CHECK18-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
20631 // CHECK18: omp.inner.for.body:
20632 // CHECK18-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20633 // CHECK18-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
20634 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
20635 // CHECK18-NEXT: store i32 [[ADD]], i32* [[I6]], align 4
20636 // CHECK18-NEXT: [[TMP19:%.*]] = load i32, i32* [[I6]], align 4
20637 // CHECK18-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
20638 // CHECK18-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
20639 // CHECK18-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
20640 // CHECK18-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
20641 // CHECK18: omp.body.continue:
20642 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
20643 // CHECK18: omp.inner.for.inc:
20644 // CHECK18-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20645 // CHECK18-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
20646 // CHECK18-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
20647 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
20648 // CHECK18: omp.inner.for.end:
20649 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
20650 // CHECK18: omp.loop.exit:
20651 // CHECK18-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20652 // CHECK18-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
20653 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
20654 // CHECK18-NEXT: br label [[OMP_PRECOND_END]]
20655 // CHECK18: omp.precond.end:
20656 // CHECK18-NEXT: ret void
20657 //
20658 //
20659 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163
20660 // CHECK18-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
20661 // CHECK18-NEXT: entry:
20662 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
20663 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
20664 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20665 // CHECK18-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
20666 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20667 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20668 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
20669 // CHECK18-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20670 // CHECK18-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20671 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
20672 // CHECK18-NEXT: ret void
20673 //
20674 //
20675 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..8
20676 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
20677 // CHECK18-NEXT: entry:
20678 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
20679 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
20680 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
20681 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
20682 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20683 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
20684 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
20685 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
20686 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
20687 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
20688 // CHECK18-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
20689 // CHECK18-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
20690 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
20691 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
20692 // CHECK18-NEXT: [[I3:%.*]] = alloca i32, align 4
20693 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
20694 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
20695 // CHECK18-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
20696 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20697 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20698 // CHECK18-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
20699 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20700 // CHECK18-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20701 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
20702 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
20703 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20704 // CHECK18-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
20705 // CHECK18-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
20706 // CHECK18-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
20707 // CHECK18-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
20708 // CHECK18-NEXT: store i32 0, i32* [[I]], align 4
20709 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20710 // CHECK18-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
20711 // CHECK18-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
20712 // CHECK18: omp.precond.then:
20713 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
20714 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20715 // CHECK18-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
20716 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
20717 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
20718 // CHECK18-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20719 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
20720 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
20721 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20722 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20723 // CHECK18-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
20724 // CHECK18-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
20725 // CHECK18: cond.true:
20726 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20727 // CHECK18-NEXT: br label [[COND_END:%.*]]
20728 // CHECK18: cond.false:
20729 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20730 // CHECK18-NEXT: br label [[COND_END]]
20731 // CHECK18: cond.end:
20732 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
20733 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
20734 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
20735 // CHECK18-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
20736 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
20737 // CHECK18: omp.inner.for.cond:
20738 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20739 // CHECK18-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20740 // CHECK18-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
20741 // CHECK18-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
20742 // CHECK18: omp.inner.for.body:
20743 // CHECK18-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
20744 // CHECK18-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
20745 // CHECK18-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20746 // CHECK18-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
20747 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
20748 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
20749 // CHECK18: omp.inner.for.inc:
20750 // CHECK18-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20751 // CHECK18-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
20752 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
20753 // CHECK18-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
20754 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
20755 // CHECK18: omp.inner.for.end:
20756 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
20757 // CHECK18: omp.loop.exit:
20758 // CHECK18-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20759 // CHECK18-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
20760 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
20761 // CHECK18-NEXT: br label [[OMP_PRECOND_END]]
20762 // CHECK18: omp.precond.end:
20763 // CHECK18-NEXT: ret void
20764 //
20765 //
20766 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..9
20767 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
20768 // CHECK18-NEXT: entry:
20769 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
20770 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
20771 // CHECK18-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
20772 // CHECK18-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
20773 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
20774 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
20775 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20776 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
20777 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
20778 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
20779 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
20780 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
20781 // CHECK18-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
20782 // CHECK18-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
20783 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
20784 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
20785 // CHECK18-NEXT: [[I4:%.*]] = alloca i32, align 4
20786 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
20787 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
20788 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
20789 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
20790 // CHECK18-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
20791 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20792 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20793 // CHECK18-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
20794 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20795 // CHECK18-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20796 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
20797 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
20798 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20799 // CHECK18-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
20800 // CHECK18-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
20801 // CHECK18-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
20802 // CHECK18-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
20803 // CHECK18-NEXT: store i32 0, i32* [[I]], align 4
20804 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20805 // CHECK18-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
20806 // CHECK18-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
20807 // CHECK18: omp.precond.then:
20808 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
20809 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20810 // CHECK18-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
20811 // CHECK18-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
20812 // CHECK18-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
20813 // CHECK18-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
20814 // CHECK18-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
20815 // CHECK18-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
20816 // CHECK18-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
20817 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
20818 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
20819 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
20820 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
20821 // CHECK18-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20822 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
20823 // CHECK18-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 35, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
20824 // CHECK18-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
20825 // CHECK18: omp.dispatch.cond:
20826 // CHECK18-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20827 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
20828 // CHECK18-NEXT: [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
20829 // CHECK18-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
20830 // CHECK18-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
20831 // CHECK18: omp.dispatch.body:
20832 // CHECK18-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
20833 // CHECK18-NEXT: store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
20834 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
20835 // CHECK18: omp.inner.for.cond:
20836 // CHECK18-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
20837 // CHECK18-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !15
20838 // CHECK18-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
20839 // CHECK18-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
20840 // CHECK18: omp.inner.for.body:
20841 // CHECK18-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
20842 // CHECK18-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
20843 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
20844 // CHECK18-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !15
20845 // CHECK18-NEXT: [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !15
20846 // CHECK18-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
20847 // CHECK18-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
20848 // CHECK18-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !15
20849 // CHECK18-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
20850 // CHECK18: omp.body.continue:
20851 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
20852 // CHECK18: omp.inner.for.inc:
20853 // CHECK18-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
20854 // CHECK18-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP21]], 1
20855 // CHECK18-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
20856 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
20857 // CHECK18: omp.inner.for.end:
20858 // CHECK18-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
20859 // CHECK18: omp.dispatch.inc:
20860 // CHECK18-NEXT: br label [[OMP_DISPATCH_COND]]
20861 // CHECK18: omp.dispatch.end:
20862 // CHECK18-NEXT: br label [[OMP_PRECOND_END]]
20863 // CHECK18: omp.precond.end:
20864 // CHECK18-NEXT: ret void
20865 //
20866 //
20867 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168
20868 // CHECK18-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
20869 // CHECK18-NEXT: entry:
20870 // CHECK18-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
20871 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
20872 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
20873 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20874 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
20875 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
20876 // CHECK18-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
20877 // CHECK18-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
20878 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20879 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20880 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
20881 // CHECK18-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
20882 // CHECK18-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20883 // CHECK18-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20884 // CHECK18-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
20885 // CHECK18-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
20886 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
20887 // CHECK18-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
20888 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
20889 // CHECK18-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
20890 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
20891 // CHECK18-NEXT: ret void
20892 //
20893 //
20894 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..11
20895 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
20896 // CHECK18-NEXT: entry:
20897 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
20898 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
20899 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
20900 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
20901 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
20902 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
20903 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
20904 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
20905 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
20906 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
20907 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
20908 // CHECK18-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
20909 // CHECK18-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
20910 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
20911 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
20912 // CHECK18-NEXT: [[I4:%.*]] = alloca i32, align 4
20913 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
20914 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
20915 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
20916 // CHECK18-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
20917 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
20918 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
20919 // CHECK18-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
20920 // CHECK18-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
20921 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
20922 // CHECK18-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
20923 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
20924 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
20925 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
20926 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20927 // CHECK18-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
20928 // CHECK18-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
20929 // CHECK18-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
20930 // CHECK18-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
20931 // CHECK18-NEXT: store i32 0, i32* [[I]], align 4
20932 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
20933 // CHECK18-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
20934 // CHECK18-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
20935 // CHECK18: omp.precond.then:
20936 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
20937 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20938 // CHECK18-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
20939 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
20940 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
20941 // CHECK18-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20942 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
20943 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
20944 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20945 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20946 // CHECK18-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
20947 // CHECK18-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
20948 // CHECK18: cond.true:
20949 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
20950 // CHECK18-NEXT: br label [[COND_END:%.*]]
20951 // CHECK18: cond.false:
20952 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20953 // CHECK18-NEXT: br label [[COND_END]]
20954 // CHECK18: cond.end:
20955 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
20956 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
20957 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
20958 // CHECK18-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
20959 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
20960 // CHECK18: omp.inner.for.cond:
20961 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20962 // CHECK18-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20963 // CHECK18-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
20964 // CHECK18-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
20965 // CHECK18: omp.inner.for.body:
20966 // CHECK18-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
20967 // CHECK18-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
20968 // CHECK18-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
20969 // CHECK18-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
20970 // CHECK18-NEXT: [[TMP20:%.*]] = load i32, i32* [[CONV]], align 8
20971 // CHECK18-NEXT: [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
20972 // CHECK18-NEXT: store i32 [[TMP20]], i32* [[CONV7]], align 4
20973 // CHECK18-NEXT: [[TMP21:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
20974 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP21]])
20975 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
20976 // CHECK18: omp.inner.for.inc:
20977 // CHECK18-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
20978 // CHECK18-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
20979 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
20980 // CHECK18-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
20981 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
20982 // CHECK18: omp.inner.for.end:
20983 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
20984 // CHECK18: omp.loop.exit:
20985 // CHECK18-NEXT: [[TMP24:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
20986 // CHECK18-NEXT: [[TMP25:%.*]] = load i32, i32* [[TMP24]], align 4
20987 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP25]])
20988 // CHECK18-NEXT: br label [[OMP_PRECOND_END]]
20989 // CHECK18: omp.precond.end:
20990 // CHECK18-NEXT: ret void
20991 //
20992 //
20993 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..12
20994 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
20995 // CHECK18-NEXT: entry:
20996 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
20997 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
20998 // CHECK18-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
20999 // CHECK18-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
21000 // CHECK18-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
21001 // CHECK18-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
21002 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
21003 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
21004 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
21005 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
21006 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
21007 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
21008 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
21009 // CHECK18-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
21010 // CHECK18-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
21011 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
21012 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
21013 // CHECK18-NEXT: [[I6:%.*]] = alloca i32, align 4
21014 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
21015 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
21016 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
21017 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21018 // CHECK18-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
21019 // CHECK18-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
21020 // CHECK18-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
21021 // CHECK18-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
21022 // CHECK18-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
21023 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
21024 // CHECK18-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
21025 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
21026 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
21027 // CHECK18-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
21028 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
21029 // CHECK18-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
21030 // CHECK18-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
21031 // CHECK18-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
21032 // CHECK18-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
21033 // CHECK18-NEXT: store i32 0, i32* [[I]], align 4
21034 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
21035 // CHECK18-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
21036 // CHECK18-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
21037 // CHECK18: omp.precond.then:
21038 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
21039 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
21040 // CHECK18-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
21041 // CHECK18-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
21042 // CHECK18-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
21043 // CHECK18-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21044 // CHECK18-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
21045 // CHECK18-NEXT: store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
21046 // CHECK18-NEXT: store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
21047 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
21048 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
21049 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[CONV]], align 8
21050 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
21051 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21052 // CHECK18-NEXT: [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
21053 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
21054 // CHECK18-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 35, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
21055 // CHECK18-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
21056 // CHECK18: omp.dispatch.cond:
21057 // CHECK18-NEXT: [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
21058 // CHECK18-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
21059 // CHECK18-NEXT: [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
21060 // CHECK18-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
21061 // CHECK18-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
21062 // CHECK18: omp.dispatch.body:
21063 // CHECK18-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
21064 // CHECK18-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
21065 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
21066 // CHECK18: omp.inner.for.cond:
21067 // CHECK18-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
21068 // CHECK18-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
21069 // CHECK18-NEXT: [[CMP7:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
21070 // CHECK18-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
21071 // CHECK18: omp.inner.for.body:
21072 // CHECK18-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
21073 // CHECK18-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
21074 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
21075 // CHECK18-NEXT: store i32 [[ADD]], i32* [[I6]], align 4, !llvm.access.group !18
21076 // CHECK18-NEXT: [[TMP21:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !18
21077 // CHECK18-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP21]] to i64
21078 // CHECK18-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
21079 // CHECK18-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !18
21080 // CHECK18-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
21081 // CHECK18: omp.body.continue:
21082 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
21083 // CHECK18: omp.inner.for.inc:
21084 // CHECK18-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
21085 // CHECK18-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP22]], 1
21086 // CHECK18-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
21087 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
21088 // CHECK18: omp.inner.for.end:
21089 // CHECK18-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
21090 // CHECK18: omp.dispatch.inc:
21091 // CHECK18-NEXT: br label [[OMP_DISPATCH_COND]]
21092 // CHECK18: omp.dispatch.end:
21093 // CHECK18-NEXT: br label [[OMP_PRECOND_END]]
21094 // CHECK18: omp.precond.end:
21095 // CHECK18-NEXT: ret void
21096 //
21097 //
21098 // CHECK18-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
21099 // CHECK18-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
21100 // CHECK18-NEXT: entry:
21101 // CHECK18-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
21102 // CHECK18-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
21103 // CHECK18-NEXT: [[M:%.*]] = alloca i32, align 4
21104 // CHECK18-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
21105 // CHECK18-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
21106 // CHECK18-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
21107 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
21108 // CHECK18-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
21109 // CHECK18-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
21110 // CHECK18-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
21111 // CHECK18-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
21112 // CHECK18-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
21113 // CHECK18-NEXT: [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
21114 // CHECK18-NEXT: [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
21115 // CHECK18-NEXT: [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
21116 // CHECK18-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
21117 // CHECK18-NEXT: [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 8
21118 // CHECK18-NEXT: [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 8
21119 // CHECK18-NEXT: [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 8
21120 // CHECK18-NEXT: [[_TMP16:%.*]] = alloca i32, align 4
21121 // CHECK18-NEXT: [[M_CASTED19:%.*]] = alloca i64, align 8
21122 // CHECK18-NEXT: [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [2 x i8*], align 8
21123 // CHECK18-NEXT: [[DOTOFFLOAD_PTRS22:%.*]] = alloca [2 x i8*], align 8
21124 // CHECK18-NEXT: [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [2 x i8*], align 8
21125 // CHECK18-NEXT: [[_TMP24:%.*]] = alloca i32, align 4
21126 // CHECK18-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
21127 // CHECK18-NEXT: store i32 10, i32* [[M]], align 4
21128 // CHECK18-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
21129 // CHECK18-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
21130 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
21131 // CHECK18-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
21132 // CHECK18-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
21133 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
21134 // CHECK18-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
21135 // CHECK18-NEXT: store i8* null, i8** [[TMP4]], align 8
21136 // CHECK18-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
21137 // CHECK18-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
21138 // CHECK18-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
21139 // CHECK18-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
21140 // CHECK18-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
21141 // CHECK18-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
21142 // CHECK18: omp_offload.failed:
21143 // CHECK18-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116([10 x i32]* [[A]]) #[[ATTR3]]
21144 // CHECK18-NEXT: br label [[OMP_OFFLOAD_CONT]]
21145 // CHECK18: omp_offload.cont:
21146 // CHECK18-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
21147 // CHECK18-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
21148 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 8
21149 // CHECK18-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
21150 // CHECK18-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
21151 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 8
21152 // CHECK18-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
21153 // CHECK18-NEXT: store i8* null, i8** [[TMP13]], align 8
21154 // CHECK18-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
21155 // CHECK18-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
21156 // CHECK18-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
21157 // CHECK18-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
21158 // CHECK18-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
21159 // CHECK18-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
21160 // CHECK18: omp_offload.failed5:
21161 // CHECK18-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121([10 x i32]* [[A]]) #[[ATTR3]]
21162 // CHECK18-NEXT: br label [[OMP_OFFLOAD_CONT6]]
21163 // CHECK18: omp_offload.cont6:
21164 // CHECK18-NEXT: [[TMP18:%.*]] = load i32, i32* [[M]], align 4
21165 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
21166 // CHECK18-NEXT: store i32 [[TMP18]], i32* [[CONV]], align 4
21167 // CHECK18-NEXT: [[TMP19:%.*]] = load i64, i64* [[M_CASTED]], align 8
21168 // CHECK18-NEXT: [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
21169 // CHECK18-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
21170 // CHECK18-NEXT: store i64 [[TMP19]], i64* [[TMP21]], align 8
21171 // CHECK18-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
21172 // CHECK18-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
21173 // CHECK18-NEXT: store i64 [[TMP19]], i64* [[TMP23]], align 8
21174 // CHECK18-NEXT: [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
21175 // CHECK18-NEXT: store i8* null, i8** [[TMP24]], align 8
21176 // CHECK18-NEXT: [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
21177 // CHECK18-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
21178 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 8
21179 // CHECK18-NEXT: [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
21180 // CHECK18-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
21181 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 8
21182 // CHECK18-NEXT: [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
21183 // CHECK18-NEXT: store i8* null, i8** [[TMP29]], align 8
21184 // CHECK18-NEXT: [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
21185 // CHECK18-NEXT: [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
21186 // CHECK18-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
21187 // CHECK18-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
21188 // CHECK18-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
21189 // CHECK18-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
21190 // CHECK18: omp_offload.failed11:
21191 // CHECK18-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126(i64 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
21192 // CHECK18-NEXT: br label [[OMP_OFFLOAD_CONT12]]
21193 // CHECK18: omp_offload.cont12:
21194 // CHECK18-NEXT: [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
21195 // CHECK18-NEXT: [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
21196 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 8
21197 // CHECK18-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
21198 // CHECK18-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
21199 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 8
21200 // CHECK18-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i64 0, i64 0
21201 // CHECK18-NEXT: store i8* null, i8** [[TMP38]], align 8
21202 // CHECK18-NEXT: [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
21203 // CHECK18-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
21204 // CHECK18-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
21205 // CHECK18-NEXT: [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
21206 // CHECK18-NEXT: [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
21207 // CHECK18-NEXT: br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
21208 // CHECK18: omp_offload.failed17:
21209 // CHECK18-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131([10 x i32]* [[A]]) #[[ATTR3]]
21210 // CHECK18-NEXT: br label [[OMP_OFFLOAD_CONT18]]
21211 // CHECK18: omp_offload.cont18:
21212 // CHECK18-NEXT: [[TMP43:%.*]] = load i32, i32* [[M]], align 4
21213 // CHECK18-NEXT: [[CONV20:%.*]] = bitcast i64* [[M_CASTED19]] to i32*
21214 // CHECK18-NEXT: store i32 [[TMP43]], i32* [[CONV20]], align 4
21215 // CHECK18-NEXT: [[TMP44:%.*]] = load i64, i64* [[M_CASTED19]], align 8
21216 // CHECK18-NEXT: [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
21217 // CHECK18-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
21218 // CHECK18-NEXT: store i64 [[TMP44]], i64* [[TMP46]], align 8
21219 // CHECK18-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
21220 // CHECK18-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
21221 // CHECK18-NEXT: store i64 [[TMP44]], i64* [[TMP48]], align 8
21222 // CHECK18-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
21223 // CHECK18-NEXT: store i8* null, i8** [[TMP49]], align 8
21224 // CHECK18-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
21225 // CHECK18-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
21226 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 8
21227 // CHECK18-NEXT: [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
21228 // CHECK18-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
21229 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 8
21230 // CHECK18-NEXT: [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
21231 // CHECK18-NEXT: store i8* null, i8** [[TMP54]], align 8
21232 // CHECK18-NEXT: [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
21233 // CHECK18-NEXT: [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
21234 // CHECK18-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
21235 // CHECK18-NEXT: [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
21236 // CHECK18-NEXT: [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
21237 // CHECK18-NEXT: br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED25:%.*]], label [[OMP_OFFLOAD_CONT26:%.*]]
21238 // CHECK18: omp_offload.failed25:
21239 // CHECK18-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136(i64 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
21240 // CHECK18-NEXT: br label [[OMP_OFFLOAD_CONT26]]
21241 // CHECK18: omp_offload.cont26:
21242 // CHECK18-NEXT: ret i32 0
21243 //
21244 //
21245 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116
21246 // CHECK18-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
21247 // CHECK18-NEXT: entry:
21248 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21249 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21250 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21251 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
21252 // CHECK18-NEXT: ret void
21253 //
21254 //
21255 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..14
21256 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
21257 // CHECK18-NEXT: entry:
21258 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
21259 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
21260 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21261 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
21262 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
21263 // CHECK18-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
21264 // CHECK18-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
21265 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
21266 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
21267 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
21268 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
21269 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
21270 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21271 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21272 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
21273 // CHECK18-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
21274 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
21275 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
21276 // CHECK18-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
21277 // CHECK18-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
21278 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
21279 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21280 // CHECK18-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
21281 // CHECK18-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
21282 // CHECK18: cond.true:
21283 // CHECK18-NEXT: br label [[COND_END:%.*]]
21284 // CHECK18: cond.false:
21285 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21286 // CHECK18-NEXT: br label [[COND_END]]
21287 // CHECK18: cond.end:
21288 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
21289 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
21290 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
21291 // CHECK18-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
21292 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
21293 // CHECK18: omp.inner.for.cond:
21294 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21295 // CHECK18-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21296 // CHECK18-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
21297 // CHECK18-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
21298 // CHECK18: omp.inner.for.body:
21299 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
21300 // CHECK18-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
21301 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21302 // CHECK18-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
21303 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
21304 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
21305 // CHECK18: omp.inner.for.inc:
21306 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21307 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
21308 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
21309 // CHECK18-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
21310 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
21311 // CHECK18: omp.inner.for.end:
21312 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
21313 // CHECK18: omp.loop.exit:
21314 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
21315 // CHECK18-NEXT: ret void
21316 //
21317 //
21318 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..15
21319 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
21320 // CHECK18-NEXT: entry:
21321 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
21322 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
21323 // CHECK18-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
21324 // CHECK18-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
21325 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21326 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
21327 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
21328 // CHECK18-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
21329 // CHECK18-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
21330 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
21331 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
21332 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
21333 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
21334 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
21335 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
21336 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21337 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21338 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21339 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
21340 // CHECK18-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
21341 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
21342 // CHECK18-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
21343 // CHECK18-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21344 // CHECK18-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
21345 // CHECK18-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
21346 // CHECK18-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
21347 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
21348 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
21349 // CHECK18-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
21350 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
21351 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
21352 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21353 // CHECK18-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
21354 // CHECK18-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
21355 // CHECK18: cond.true:
21356 // CHECK18-NEXT: br label [[COND_END:%.*]]
21357 // CHECK18: cond.false:
21358 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21359 // CHECK18-NEXT: br label [[COND_END]]
21360 // CHECK18: cond.end:
21361 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
21362 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
21363 // CHECK18-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
21364 // CHECK18-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
21365 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
21366 // CHECK18: omp.inner.for.cond:
21367 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21368 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21369 // CHECK18-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
21370 // CHECK18-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
21371 // CHECK18: omp.inner.for.body:
21372 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21373 // CHECK18-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
21374 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
21375 // CHECK18-NEXT: store i32 [[ADD]], i32* [[I]], align 4
21376 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
21377 // CHECK18-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
21378 // CHECK18-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
21379 // CHECK18-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
21380 // CHECK18-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
21381 // CHECK18: omp.body.continue:
21382 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
21383 // CHECK18: omp.inner.for.inc:
21384 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21385 // CHECK18-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
21386 // CHECK18-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
21387 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
21388 // CHECK18: omp.inner.for.end:
21389 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
21390 // CHECK18: omp.loop.exit:
21391 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
21392 // CHECK18-NEXT: ret void
21393 //
21394 //
21395 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121
21396 // CHECK18-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
21397 // CHECK18-NEXT: entry:
21398 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21399 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21400 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21401 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
21402 // CHECK18-NEXT: ret void
21403 //
21404 //
21405 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..17
21406 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
21407 // CHECK18-NEXT: entry:
21408 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
21409 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
21410 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21411 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
21412 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
21413 // CHECK18-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
21414 // CHECK18-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
21415 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
21416 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
21417 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
21418 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
21419 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
21420 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21421 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21422 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
21423 // CHECK18-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
21424 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
21425 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
21426 // CHECK18-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
21427 // CHECK18-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
21428 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
21429 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21430 // CHECK18-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
21431 // CHECK18-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
21432 // CHECK18: cond.true:
21433 // CHECK18-NEXT: br label [[COND_END:%.*]]
21434 // CHECK18: cond.false:
21435 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21436 // CHECK18-NEXT: br label [[COND_END]]
21437 // CHECK18: cond.end:
21438 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
21439 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
21440 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
21441 // CHECK18-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
21442 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
21443 // CHECK18: omp.inner.for.cond:
21444 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21445 // CHECK18-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21446 // CHECK18-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
21447 // CHECK18-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
21448 // CHECK18: omp.inner.for.body:
21449 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
21450 // CHECK18-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
21451 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21452 // CHECK18-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
21453 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
21454 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
21455 // CHECK18: omp.inner.for.inc:
21456 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21457 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
21458 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
21459 // CHECK18-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
21460 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
21461 // CHECK18: omp.inner.for.end:
21462 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
21463 // CHECK18: omp.loop.exit:
21464 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
21465 // CHECK18-NEXT: ret void
21466 //
21467 //
21468 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..18
21469 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
21470 // CHECK18-NEXT: entry:
21471 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
21472 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
21473 // CHECK18-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
21474 // CHECK18-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
21475 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21476 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
21477 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
21478 // CHECK18-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
21479 // CHECK18-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
21480 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
21481 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
21482 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
21483 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
21484 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
21485 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
21486 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21487 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21488 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21489 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
21490 // CHECK18-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
21491 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
21492 // CHECK18-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
21493 // CHECK18-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21494 // CHECK18-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
21495 // CHECK18-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
21496 // CHECK18-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
21497 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
21498 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
21499 // CHECK18-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
21500 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
21501 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
21502 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21503 // CHECK18-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
21504 // CHECK18-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
21505 // CHECK18: cond.true:
21506 // CHECK18-NEXT: br label [[COND_END:%.*]]
21507 // CHECK18: cond.false:
21508 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21509 // CHECK18-NEXT: br label [[COND_END]]
21510 // CHECK18: cond.end:
21511 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
21512 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
21513 // CHECK18-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
21514 // CHECK18-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
21515 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
21516 // CHECK18: omp.inner.for.cond:
21517 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21518 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21519 // CHECK18-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
21520 // CHECK18-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
21521 // CHECK18: omp.inner.for.body:
21522 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21523 // CHECK18-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
21524 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
21525 // CHECK18-NEXT: store i32 [[ADD]], i32* [[I]], align 4
21526 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
21527 // CHECK18-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
21528 // CHECK18-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
21529 // CHECK18-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
21530 // CHECK18-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
21531 // CHECK18: omp.body.continue:
21532 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
21533 // CHECK18: omp.inner.for.inc:
21534 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21535 // CHECK18-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
21536 // CHECK18-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
21537 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
21538 // CHECK18: omp.inner.for.end:
21539 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
21540 // CHECK18: omp.loop.exit:
21541 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
21542 // CHECK18-NEXT: ret void
21543 //
21544 //
21545 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126
21546 // CHECK18-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
21547 // CHECK18-NEXT: entry:
21548 // CHECK18-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
21549 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21550 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
21551 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
21552 // CHECK18-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
21553 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21554 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
21555 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21556 // CHECK18-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
21557 // CHECK18-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
21558 // CHECK18-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
21559 // CHECK18-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
21560 // CHECK18-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
21561 // CHECK18-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
21562 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
21563 // CHECK18-NEXT: ret void
21564 //
21565 //
21566 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..21
21567 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
21568 // CHECK18-NEXT: entry:
21569 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
21570 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
21571 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21572 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
21573 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
21574 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
21575 // CHECK18-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
21576 // CHECK18-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
21577 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
21578 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
21579 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
21580 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
21581 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
21582 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
21583 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21584 // CHECK18-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
21585 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21586 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
21587 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
21588 // CHECK18-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
21589 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
21590 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
21591 // CHECK18-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
21592 // CHECK18-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
21593 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
21594 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21595 // CHECK18-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
21596 // CHECK18-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
21597 // CHECK18: cond.true:
21598 // CHECK18-NEXT: br label [[COND_END:%.*]]
21599 // CHECK18: cond.false:
21600 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21601 // CHECK18-NEXT: br label [[COND_END]]
21602 // CHECK18: cond.end:
21603 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
21604 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
21605 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
21606 // CHECK18-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
21607 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
21608 // CHECK18: omp.inner.for.cond:
21609 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21610 // CHECK18-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21611 // CHECK18-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
21612 // CHECK18-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
21613 // CHECK18: omp.inner.for.body:
21614 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
21615 // CHECK18-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
21616 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21617 // CHECK18-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
21618 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8
21619 // CHECK18-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
21620 // CHECK18-NEXT: store i32 [[TMP12]], i32* [[CONV2]], align 4
21621 // CHECK18-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
21622 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
21623 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
21624 // CHECK18: omp.inner.for.inc:
21625 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21626 // CHECK18-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
21627 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
21628 // CHECK18-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
21629 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
21630 // CHECK18: omp.inner.for.end:
21631 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
21632 // CHECK18: omp.loop.exit:
21633 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
21634 // CHECK18-NEXT: ret void
21635 //
21636 //
21637 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..22
21638 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
21639 // CHECK18-NEXT: entry:
21640 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
21641 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
21642 // CHECK18-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
21643 // CHECK18-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
21644 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21645 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
21646 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
21647 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
21648 // CHECK18-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
21649 // CHECK18-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
21650 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
21651 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
21652 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
21653 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
21654 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
21655 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
21656 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21657 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21658 // CHECK18-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
21659 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21660 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
21661 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
21662 // CHECK18-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
21663 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
21664 // CHECK18-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
21665 // CHECK18-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21666 // CHECK18-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
21667 // CHECK18-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
21668 // CHECK18-NEXT: store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
21669 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
21670 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
21671 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
21672 // CHECK18-NEXT: [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
21673 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
21674 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
21675 // CHECK18-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
21676 // CHECK18: omp.dispatch.cond:
21677 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21678 // CHECK18-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21679 // CHECK18-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP7]] to i32
21680 // CHECK18-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[CONV3]]
21681 // CHECK18-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
21682 // CHECK18: cond.true:
21683 // CHECK18-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21684 // CHECK18-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP8]] to i32
21685 // CHECK18-NEXT: br label [[COND_END:%.*]]
21686 // CHECK18: cond.false:
21687 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21688 // CHECK18-NEXT: br label [[COND_END]]
21689 // CHECK18: cond.end:
21690 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ [[CONV4]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
21691 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
21692 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
21693 // CHECK18-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
21694 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21695 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21696 // CHECK18-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
21697 // CHECK18-NEXT: br i1 [[CMP5]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
21698 // CHECK18: omp.dispatch.body:
21699 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
21700 // CHECK18: omp.inner.for.cond:
21701 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21702 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21703 // CHECK18-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
21704 // CHECK18-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
21705 // CHECK18: omp.inner.for.body:
21706 // CHECK18-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21707 // CHECK18-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
21708 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
21709 // CHECK18-NEXT: store i32 [[ADD]], i32* [[I]], align 4
21710 // CHECK18-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4
21711 // CHECK18-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
21712 // CHECK18-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
21713 // CHECK18-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
21714 // CHECK18-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
21715 // CHECK18: omp.body.continue:
21716 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
21717 // CHECK18: omp.inner.for.inc:
21718 // CHECK18-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21719 // CHECK18-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP17]], 1
21720 // CHECK18-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
21721 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
21722 // CHECK18: omp.inner.for.end:
21723 // CHECK18-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
21724 // CHECK18: omp.dispatch.inc:
21725 // CHECK18-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
21726 // CHECK18-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
21727 // CHECK18-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
21728 // CHECK18-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_LB]], align 4
21729 // CHECK18-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21730 // CHECK18-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
21731 // CHECK18-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
21732 // CHECK18-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_UB]], align 4
21733 // CHECK18-NEXT: br label [[OMP_DISPATCH_COND]]
21734 // CHECK18: omp.dispatch.end:
21735 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
21736 // CHECK18-NEXT: ret void
21737 //
21738 //
21739 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
21740 // CHECK18-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
21741 // CHECK18-NEXT: entry:
21742 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21743 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21744 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21745 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
21746 // CHECK18-NEXT: ret void
21747 //
21748 //
21749 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..25
21750 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
21751 // CHECK18-NEXT: entry:
21752 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
21753 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
21754 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21755 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
21756 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
21757 // CHECK18-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
21758 // CHECK18-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
21759 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
21760 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
21761 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
21762 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
21763 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
21764 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21765 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21766 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
21767 // CHECK18-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
21768 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
21769 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
21770 // CHECK18-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
21771 // CHECK18-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
21772 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
21773 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21774 // CHECK18-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
21775 // CHECK18-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
21776 // CHECK18: cond.true:
21777 // CHECK18-NEXT: br label [[COND_END:%.*]]
21778 // CHECK18: cond.false:
21779 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21780 // CHECK18-NEXT: br label [[COND_END]]
21781 // CHECK18: cond.end:
21782 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
21783 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
21784 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
21785 // CHECK18-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
21786 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
21787 // CHECK18: omp.inner.for.cond:
21788 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21789 // CHECK18-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21790 // CHECK18-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
21791 // CHECK18-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
21792 // CHECK18: omp.inner.for.body:
21793 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
21794 // CHECK18-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
21795 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21796 // CHECK18-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
21797 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
21798 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
21799 // CHECK18: omp.inner.for.inc:
21800 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21801 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
21802 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
21803 // CHECK18-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
21804 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
21805 // CHECK18: omp.inner.for.end:
21806 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
21807 // CHECK18: omp.loop.exit:
21808 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
21809 // CHECK18-NEXT: ret void
21810 //
21811 //
21812 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..26
21813 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
21814 // CHECK18-NEXT: entry:
21815 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
21816 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
21817 // CHECK18-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
21818 // CHECK18-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
21819 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21820 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
21821 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
21822 // CHECK18-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
21823 // CHECK18-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
21824 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
21825 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
21826 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
21827 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
21828 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
21829 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
21830 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21831 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21832 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21833 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
21834 // CHECK18-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
21835 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
21836 // CHECK18-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
21837 // CHECK18-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21838 // CHECK18-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
21839 // CHECK18-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
21840 // CHECK18-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
21841 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
21842 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
21843 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
21844 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
21845 // CHECK18-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
21846 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
21847 // CHECK18-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
21848 // CHECK18-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
21849 // CHECK18: omp.dispatch.cond:
21850 // CHECK18-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
21851 // CHECK18-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
21852 // CHECK18-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
21853 // CHECK18: omp.dispatch.body:
21854 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
21855 // CHECK18-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
21856 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
21857 // CHECK18: omp.inner.for.cond:
21858 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
21859 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !21
21860 // CHECK18-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
21861 // CHECK18-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
21862 // CHECK18: omp.inner.for.body:
21863 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
21864 // CHECK18-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
21865 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
21866 // CHECK18-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !21
21867 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !21
21868 // CHECK18-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP12]] to i64
21869 // CHECK18-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
21870 // CHECK18-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !21
21871 // CHECK18-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
21872 // CHECK18: omp.body.continue:
21873 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
21874 // CHECK18: omp.inner.for.inc:
21875 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
21876 // CHECK18-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], 1
21877 // CHECK18-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !21
21878 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
21879 // CHECK18: omp.inner.for.end:
21880 // CHECK18-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
21881 // CHECK18: omp.dispatch.inc:
21882 // CHECK18-NEXT: br label [[OMP_DISPATCH_COND]]
21883 // CHECK18: omp.dispatch.end:
21884 // CHECK18-NEXT: ret void
21885 //
21886 //
21887 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136
21888 // CHECK18-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
21889 // CHECK18-NEXT: entry:
21890 // CHECK18-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
21891 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21892 // CHECK18-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
21893 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
21894 // CHECK18-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
21895 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21896 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
21897 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21898 // CHECK18-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
21899 // CHECK18-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
21900 // CHECK18-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
21901 // CHECK18-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
21902 // CHECK18-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
21903 // CHECK18-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
21904 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
21905 // CHECK18-NEXT: ret void
21906 //
21907 //
21908 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..29
21909 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
21910 // CHECK18-NEXT: entry:
21911 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
21912 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
21913 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21914 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
21915 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
21916 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
21917 // CHECK18-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
21918 // CHECK18-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
21919 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
21920 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
21921 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
21922 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
21923 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
21924 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
21925 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
21926 // CHECK18-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
21927 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
21928 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
21929 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
21930 // CHECK18-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
21931 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
21932 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
21933 // CHECK18-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
21934 // CHECK18-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
21935 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
21936 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21937 // CHECK18-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
21938 // CHECK18-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
21939 // CHECK18: cond.true:
21940 // CHECK18-NEXT: br label [[COND_END:%.*]]
21941 // CHECK18: cond.false:
21942 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21943 // CHECK18-NEXT: br label [[COND_END]]
21944 // CHECK18: cond.end:
21945 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
21946 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
21947 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
21948 // CHECK18-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
21949 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
21950 // CHECK18: omp.inner.for.cond:
21951 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21952 // CHECK18-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21953 // CHECK18-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
21954 // CHECK18-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
21955 // CHECK18: omp.inner.for.body:
21956 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
21957 // CHECK18-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
21958 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
21959 // CHECK18-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
21960 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8
21961 // CHECK18-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
21962 // CHECK18-NEXT: store i32 [[TMP12]], i32* [[CONV2]], align 4
21963 // CHECK18-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
21964 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
21965 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
21966 // CHECK18: omp.inner.for.inc:
21967 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
21968 // CHECK18-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
21969 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
21970 // CHECK18-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
21971 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
21972 // CHECK18: omp.inner.for.end:
21973 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
21974 // CHECK18: omp.loop.exit:
21975 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
21976 // CHECK18-NEXT: ret void
21977 //
21978 //
21979 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined..30
21980 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
21981 // CHECK18-NEXT: entry:
21982 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
21983 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
21984 // CHECK18-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
21985 // CHECK18-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
21986 // CHECK18-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
21987 // CHECK18-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
21988 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
21989 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
21990 // CHECK18-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
21991 // CHECK18-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
21992 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
21993 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
21994 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
21995 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
21996 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
21997 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
21998 // CHECK18-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
21999 // CHECK18-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
22000 // CHECK18-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
22001 // CHECK18-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
22002 // CHECK18-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
22003 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
22004 // CHECK18-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
22005 // CHECK18-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
22006 // CHECK18-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
22007 // CHECK18-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
22008 // CHECK18-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
22009 // CHECK18-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
22010 // CHECK18-NEXT: store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
22011 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
22012 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
22013 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV]], align 8
22014 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
22015 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
22016 // CHECK18-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
22017 // CHECK18-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
22018 // CHECK18-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 35, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
22019 // CHECK18-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
22020 // CHECK18: omp.dispatch.cond:
22021 // CHECK18-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
22022 // CHECK18-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
22023 // CHECK18-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
22024 // CHECK18: omp.dispatch.body:
22025 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
22026 // CHECK18-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
22027 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
22028 // CHECK18: omp.inner.for.cond:
22029 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
22030 // CHECK18-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !24
22031 // CHECK18-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
22032 // CHECK18-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
22033 // CHECK18: omp.inner.for.body:
22034 // CHECK18-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
22035 // CHECK18-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
22036 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
22037 // CHECK18-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !24
22038 // CHECK18-NEXT: [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !24
22039 // CHECK18-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
22040 // CHECK18-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
22041 // CHECK18-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !24
22042 // CHECK18-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
22043 // CHECK18: omp.body.continue:
22044 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
22045 // CHECK18: omp.inner.for.inc:
22046 // CHECK18-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
22047 // CHECK18-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
22048 // CHECK18-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
22049 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
22050 // CHECK18: omp.inner.for.end:
22051 // CHECK18-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
22052 // CHECK18: omp.dispatch.inc:
22053 // CHECK18-NEXT: br label [[OMP_DISPATCH_COND]]
22054 // CHECK18: omp.dispatch.end:
22055 // CHECK18-NEXT: ret void
22056 //
22057 //
22058 // CHECK18-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
22059 // CHECK18-SAME: () #[[ATTR5:[0-9]+]] {
22060 // CHECK18-NEXT: entry:
22061 // CHECK18-NEXT: call void @__tgt_register_requires(i64 1)
22062 // CHECK18-NEXT: ret void
22063 //
22064 //
22065 // CHECK19-LABEL: define {{[^@]+}}@main
22066 // CHECK19-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
22067 // CHECK19-NEXT: entry:
22068 // CHECK19-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
22069 // CHECK19-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
22070 // CHECK19-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
22071 // CHECK19-NEXT: [[N:%.*]] = alloca i32, align 4
22072 // CHECK19-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
22073 // CHECK19-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
22074 // CHECK19-NEXT: [[M:%.*]] = alloca i32, align 4
22075 // CHECK19-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
22076 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
22077 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
22078 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
22079 // CHECK19-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
22080 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
22081 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
22082 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
22083 // CHECK19-NEXT: [[N_CASTED3:%.*]] = alloca i32, align 4
22084 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
22085 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
22086 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
22087 // CHECK19-NEXT: [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
22088 // CHECK19-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
22089 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
22090 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
22091 // CHECK19-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
22092 // CHECK19-NEXT: [[N_CASTED17:%.*]] = alloca i32, align 4
22093 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS18:%.*]] = alloca [4 x i8*], align 4
22094 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS19:%.*]] = alloca [4 x i8*], align 4
22095 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS20:%.*]] = alloca [4 x i8*], align 4
22096 // CHECK19-NEXT: [[DOTOFFLOAD_SIZES21:%.*]] = alloca [4 x i64], align 4
22097 // CHECK19-NEXT: [[_TMP22:%.*]] = alloca i32, align 4
22098 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_23:%.*]] = alloca i32, align 4
22099 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
22100 // CHECK19-NEXT: [[N_CASTED31:%.*]] = alloca i32, align 4
22101 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS32:%.*]] = alloca [3 x i8*], align 4
22102 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS33:%.*]] = alloca [3 x i8*], align 4
22103 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS34:%.*]] = alloca [3 x i8*], align 4
22104 // CHECK19-NEXT: [[DOTOFFLOAD_SIZES35:%.*]] = alloca [3 x i64], align 4
22105 // CHECK19-NEXT: [[_TMP36:%.*]] = alloca i32, align 4
22106 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_37:%.*]] = alloca i32, align 4
22107 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
22108 // CHECK19-NEXT: [[M_CASTED45:%.*]] = alloca i32, align 4
22109 // CHECK19-NEXT: [[N_CASTED46:%.*]] = alloca i32, align 4
22110 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS47:%.*]] = alloca [4 x i8*], align 4
22111 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS48:%.*]] = alloca [4 x i8*], align 4
22112 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS49:%.*]] = alloca [4 x i8*], align 4
22113 // CHECK19-NEXT: [[DOTOFFLOAD_SIZES50:%.*]] = alloca [4 x i64], align 4
22114 // CHECK19-NEXT: [[_TMP51:%.*]] = alloca i32, align 4
22115 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_52:%.*]] = alloca i32, align 4
22116 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_53:%.*]] = alloca i32, align 4
22117 // CHECK19-NEXT: store i32 0, i32* [[RETVAL]], align 4
22118 // CHECK19-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
22119 // CHECK19-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
22120 // CHECK19-NEXT: store i32 100, i32* [[N]], align 4
22121 // CHECK19-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
22122 // CHECK19-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
22123 // CHECK19-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
22124 // CHECK19-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
22125 // CHECK19-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
22126 // CHECK19-NEXT: store i32 10, i32* [[M]], align 4
22127 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
22128 // CHECK19-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
22129 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
22130 // CHECK19-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
22131 // CHECK19-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
22132 // CHECK19-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
22133 // CHECK19-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
22134 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[TMP7]], align 4
22135 // CHECK19-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
22136 // CHECK19-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
22137 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[TMP9]], align 4
22138 // CHECK19-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
22139 // CHECK19-NEXT: store i64 4, i64* [[TMP10]], align 4
22140 // CHECK19-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
22141 // CHECK19-NEXT: store i8* null, i8** [[TMP11]], align 4
22142 // CHECK19-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
22143 // CHECK19-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
22144 // CHECK19-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
22145 // CHECK19-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
22146 // CHECK19-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
22147 // CHECK19-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
22148 // CHECK19-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
22149 // CHECK19-NEXT: store i64 4, i64* [[TMP16]], align 4
22150 // CHECK19-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
22151 // CHECK19-NEXT: store i8* null, i8** [[TMP17]], align 4
22152 // CHECK19-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
22153 // CHECK19-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
22154 // CHECK19-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 4
22155 // CHECK19-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
22156 // CHECK19-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
22157 // CHECK19-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 4
22158 // CHECK19-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
22159 // CHECK19-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 4
22160 // CHECK19-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
22161 // CHECK19-NEXT: store i8* null, i8** [[TMP23]], align 4
22162 // CHECK19-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
22163 // CHECK19-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
22164 // CHECK19-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
22165 // CHECK19-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
22166 // CHECK19-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
22167 // CHECK19-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
22168 // CHECK19-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
22169 // CHECK19-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
22170 // CHECK19-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
22171 // CHECK19-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
22172 // CHECK19-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22173 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
22174 // CHECK19-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
22175 // CHECK19-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
22176 // CHECK19-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
22177 // CHECK19-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
22178 // CHECK19-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
22179 // CHECK19: omp_offload.failed:
22180 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
22181 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT]]
22182 // CHECK19: omp_offload.cont:
22183 // CHECK19-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
22184 // CHECK19-NEXT: store i32 [[TMP33]], i32* [[N_CASTED3]], align 4
22185 // CHECK19-NEXT: [[TMP34:%.*]] = load i32, i32* [[N_CASTED3]], align 4
22186 // CHECK19-NEXT: [[TMP35:%.*]] = mul nuw i32 [[TMP0]], 4
22187 // CHECK19-NEXT: [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
22188 // CHECK19-NEXT: [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
22189 // CHECK19-NEXT: [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
22190 // CHECK19-NEXT: store i32 [[TMP34]], i32* [[TMP38]], align 4
22191 // CHECK19-NEXT: [[TMP39:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
22192 // CHECK19-NEXT: [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
22193 // CHECK19-NEXT: store i32 [[TMP34]], i32* [[TMP40]], align 4
22194 // CHECK19-NEXT: [[TMP41:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
22195 // CHECK19-NEXT: store i64 4, i64* [[TMP41]], align 4
22196 // CHECK19-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
22197 // CHECK19-NEXT: store i8* null, i8** [[TMP42]], align 4
22198 // CHECK19-NEXT: [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
22199 // CHECK19-NEXT: [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i32*
22200 // CHECK19-NEXT: store i32 [[TMP0]], i32* [[TMP44]], align 4
22201 // CHECK19-NEXT: [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
22202 // CHECK19-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
22203 // CHECK19-NEXT: store i32 [[TMP0]], i32* [[TMP46]], align 4
22204 // CHECK19-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 1
22205 // CHECK19-NEXT: store i64 4, i64* [[TMP47]], align 4
22206 // CHECK19-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
22207 // CHECK19-NEXT: store i8* null, i8** [[TMP48]], align 4
22208 // CHECK19-NEXT: [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
22209 // CHECK19-NEXT: [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32**
22210 // CHECK19-NEXT: store i32* [[VLA]], i32** [[TMP50]], align 4
22211 // CHECK19-NEXT: [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
22212 // CHECK19-NEXT: [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32**
22213 // CHECK19-NEXT: store i32* [[VLA]], i32** [[TMP52]], align 4
22214 // CHECK19-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
22215 // CHECK19-NEXT: store i64 [[TMP36]], i64* [[TMP53]], align 4
22216 // CHECK19-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
22217 // CHECK19-NEXT: store i8* null, i8** [[TMP54]], align 4
22218 // CHECK19-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
22219 // CHECK19-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
22220 // CHECK19-NEXT: [[TMP57:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
22221 // CHECK19-NEXT: [[TMP58:%.*]] = load i32, i32* [[N]], align 4
22222 // CHECK19-NEXT: store i32 [[TMP58]], i32* [[DOTCAPTURE_EXPR_9]], align 4
22223 // CHECK19-NEXT: [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
22224 // CHECK19-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP59]], 0
22225 // CHECK19-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
22226 // CHECK19-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
22227 // CHECK19-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
22228 // CHECK19-NEXT: [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
22229 // CHECK19-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP60]], 1
22230 // CHECK19-NEXT: [[TMP61:%.*]] = zext i32 [[ADD14]] to i64
22231 // CHECK19-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP61]])
22232 // CHECK19-NEXT: [[TMP62:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153.region_id, i32 3, i8** [[TMP55]], i8** [[TMP56]], i64* [[TMP57]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
22233 // CHECK19-NEXT: [[TMP63:%.*]] = icmp ne i32 [[TMP62]], 0
22234 // CHECK19-NEXT: br i1 [[TMP63]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
22235 // CHECK19: omp_offload.failed15:
22236 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153(i32 [[TMP34]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
22237 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT16]]
22238 // CHECK19: omp_offload.cont16:
22239 // CHECK19-NEXT: [[TMP64:%.*]] = load i32, i32* [[M]], align 4
22240 // CHECK19-NEXT: store i32 [[TMP64]], i32* [[M_CASTED]], align 4
22241 // CHECK19-NEXT: [[TMP65:%.*]] = load i32, i32* [[M_CASTED]], align 4
22242 // CHECK19-NEXT: [[TMP66:%.*]] = load i32, i32* [[N]], align 4
22243 // CHECK19-NEXT: store i32 [[TMP66]], i32* [[N_CASTED17]], align 4
22244 // CHECK19-NEXT: [[TMP67:%.*]] = load i32, i32* [[N_CASTED17]], align 4
22245 // CHECK19-NEXT: [[TMP68:%.*]] = mul nuw i32 [[TMP0]], 4
22246 // CHECK19-NEXT: [[TMP69:%.*]] = sext i32 [[TMP68]] to i64
22247 // CHECK19-NEXT: [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
22248 // CHECK19-NEXT: [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i32*
22249 // CHECK19-NEXT: store i32 [[TMP65]], i32* [[TMP71]], align 4
22250 // CHECK19-NEXT: [[TMP72:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
22251 // CHECK19-NEXT: [[TMP73:%.*]] = bitcast i8** [[TMP72]] to i32*
22252 // CHECK19-NEXT: store i32 [[TMP65]], i32* [[TMP73]], align 4
22253 // CHECK19-NEXT: [[TMP74:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
22254 // CHECK19-NEXT: store i64 4, i64* [[TMP74]], align 4
22255 // CHECK19-NEXT: [[TMP75:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 0
22256 // CHECK19-NEXT: store i8* null, i8** [[TMP75]], align 4
22257 // CHECK19-NEXT: [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 1
22258 // CHECK19-NEXT: [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i32*
22259 // CHECK19-NEXT: store i32 [[TMP67]], i32* [[TMP77]], align 4
22260 // CHECK19-NEXT: [[TMP78:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 1
22261 // CHECK19-NEXT: [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
22262 // CHECK19-NEXT: store i32 [[TMP67]], i32* [[TMP79]], align 4
22263 // CHECK19-NEXT: [[TMP80:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 1
22264 // CHECK19-NEXT: store i64 4, i64* [[TMP80]], align 4
22265 // CHECK19-NEXT: [[TMP81:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 1
22266 // CHECK19-NEXT: store i8* null, i8** [[TMP81]], align 4
22267 // CHECK19-NEXT: [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 2
22268 // CHECK19-NEXT: [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i32*
22269 // CHECK19-NEXT: store i32 [[TMP0]], i32* [[TMP83]], align 4
22270 // CHECK19-NEXT: [[TMP84:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 2
22271 // CHECK19-NEXT: [[TMP85:%.*]] = bitcast i8** [[TMP84]] to i32*
22272 // CHECK19-NEXT: store i32 [[TMP0]], i32* [[TMP85]], align 4
22273 // CHECK19-NEXT: [[TMP86:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 2
22274 // CHECK19-NEXT: store i64 4, i64* [[TMP86]], align 4
22275 // CHECK19-NEXT: [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 2
22276 // CHECK19-NEXT: store i8* null, i8** [[TMP87]], align 4
22277 // CHECK19-NEXT: [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 3
22278 // CHECK19-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
22279 // CHECK19-NEXT: store i32* [[VLA]], i32** [[TMP89]], align 4
22280 // CHECK19-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 3
22281 // CHECK19-NEXT: [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i32**
22282 // CHECK19-NEXT: store i32* [[VLA]], i32** [[TMP91]], align 4
22283 // CHECK19-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 3
22284 // CHECK19-NEXT: store i64 [[TMP69]], i64* [[TMP92]], align 4
22285 // CHECK19-NEXT: [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 3
22286 // CHECK19-NEXT: store i8* null, i8** [[TMP93]], align 4
22287 // CHECK19-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
22288 // CHECK19-NEXT: [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
22289 // CHECK19-NEXT: [[TMP96:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
22290 // CHECK19-NEXT: [[TMP97:%.*]] = load i32, i32* [[N]], align 4
22291 // CHECK19-NEXT: store i32 [[TMP97]], i32* [[DOTCAPTURE_EXPR_23]], align 4
22292 // CHECK19-NEXT: [[TMP98:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_23]], align 4
22293 // CHECK19-NEXT: [[SUB25:%.*]] = sub nsw i32 [[TMP98]], 0
22294 // CHECK19-NEXT: [[DIV26:%.*]] = sdiv i32 [[SUB25]], 1
22295 // CHECK19-NEXT: [[SUB27:%.*]] = sub nsw i32 [[DIV26]], 1
22296 // CHECK19-NEXT: store i32 [[SUB27]], i32* [[DOTCAPTURE_EXPR_24]], align 4
22297 // CHECK19-NEXT: [[TMP99:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
22298 // CHECK19-NEXT: [[ADD28:%.*]] = add nsw i32 [[TMP99]], 1
22299 // CHECK19-NEXT: [[TMP100:%.*]] = zext i32 [[ADD28]] to i64
22300 // CHECK19-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP100]])
22301 // CHECK19-NEXT: [[TMP101:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158.region_id, i32 4, i8** [[TMP94]], i8** [[TMP95]], i64* [[TMP96]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
22302 // CHECK19-NEXT: [[TMP102:%.*]] = icmp ne i32 [[TMP101]], 0
22303 // CHECK19-NEXT: br i1 [[TMP102]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
22304 // CHECK19: omp_offload.failed29:
22305 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158(i32 [[TMP65]], i32 [[TMP67]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
22306 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT30]]
22307 // CHECK19: omp_offload.cont30:
22308 // CHECK19-NEXT: [[TMP103:%.*]] = load i32, i32* [[N]], align 4
22309 // CHECK19-NEXT: store i32 [[TMP103]], i32* [[N_CASTED31]], align 4
22310 // CHECK19-NEXT: [[TMP104:%.*]] = load i32, i32* [[N_CASTED31]], align 4
22311 // CHECK19-NEXT: [[TMP105:%.*]] = mul nuw i32 [[TMP0]], 4
22312 // CHECK19-NEXT: [[TMP106:%.*]] = sext i32 [[TMP105]] to i64
22313 // CHECK19-NEXT: [[TMP107:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
22314 // CHECK19-NEXT: [[TMP108:%.*]] = bitcast i8** [[TMP107]] to i32*
22315 // CHECK19-NEXT: store i32 [[TMP104]], i32* [[TMP108]], align 4
22316 // CHECK19-NEXT: [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
22317 // CHECK19-NEXT: [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i32*
22318 // CHECK19-NEXT: store i32 [[TMP104]], i32* [[TMP110]], align 4
22319 // CHECK19-NEXT: [[TMP111:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
22320 // CHECK19-NEXT: store i64 4, i64* [[TMP111]], align 4
22321 // CHECK19-NEXT: [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 0
22322 // CHECK19-NEXT: store i8* null, i8** [[TMP112]], align 4
22323 // CHECK19-NEXT: [[TMP113:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 1
22324 // CHECK19-NEXT: [[TMP114:%.*]] = bitcast i8** [[TMP113]] to i32*
22325 // CHECK19-NEXT: store i32 [[TMP0]], i32* [[TMP114]], align 4
22326 // CHECK19-NEXT: [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 1
22327 // CHECK19-NEXT: [[TMP116:%.*]] = bitcast i8** [[TMP115]] to i32*
22328 // CHECK19-NEXT: store i32 [[TMP0]], i32* [[TMP116]], align 4
22329 // CHECK19-NEXT: [[TMP117:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 1
22330 // CHECK19-NEXT: store i64 4, i64* [[TMP117]], align 4
22331 // CHECK19-NEXT: [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 1
22332 // CHECK19-NEXT: store i8* null, i8** [[TMP118]], align 4
22333 // CHECK19-NEXT: [[TMP119:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 2
22334 // CHECK19-NEXT: [[TMP120:%.*]] = bitcast i8** [[TMP119]] to i32**
22335 // CHECK19-NEXT: store i32* [[VLA]], i32** [[TMP120]], align 4
22336 // CHECK19-NEXT: [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 2
22337 // CHECK19-NEXT: [[TMP122:%.*]] = bitcast i8** [[TMP121]] to i32**
22338 // CHECK19-NEXT: store i32* [[VLA]], i32** [[TMP122]], align 4
22339 // CHECK19-NEXT: [[TMP123:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 2
22340 // CHECK19-NEXT: store i64 [[TMP106]], i64* [[TMP123]], align 4
22341 // CHECK19-NEXT: [[TMP124:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 2
22342 // CHECK19-NEXT: store i8* null, i8** [[TMP124]], align 4
22343 // CHECK19-NEXT: [[TMP125:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
22344 // CHECK19-NEXT: [[TMP126:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
22345 // CHECK19-NEXT: [[TMP127:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
22346 // CHECK19-NEXT: [[TMP128:%.*]] = load i32, i32* [[N]], align 4
22347 // CHECK19-NEXT: store i32 [[TMP128]], i32* [[DOTCAPTURE_EXPR_37]], align 4
22348 // CHECK19-NEXT: [[TMP129:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_37]], align 4
22349 // CHECK19-NEXT: [[SUB39:%.*]] = sub nsw i32 [[TMP129]], 0
22350 // CHECK19-NEXT: [[DIV40:%.*]] = sdiv i32 [[SUB39]], 1
22351 // CHECK19-NEXT: [[SUB41:%.*]] = sub nsw i32 [[DIV40]], 1
22352 // CHECK19-NEXT: store i32 [[SUB41]], i32* [[DOTCAPTURE_EXPR_38]], align 4
22353 // CHECK19-NEXT: [[TMP130:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_38]], align 4
22354 // CHECK19-NEXT: [[ADD42:%.*]] = add nsw i32 [[TMP130]], 1
22355 // CHECK19-NEXT: [[TMP131:%.*]] = zext i32 [[ADD42]] to i64
22356 // CHECK19-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP131]])
22357 // CHECK19-NEXT: [[TMP132:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163.region_id, i32 3, i8** [[TMP125]], i8** [[TMP126]], i64* [[TMP127]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
22358 // CHECK19-NEXT: [[TMP133:%.*]] = icmp ne i32 [[TMP132]], 0
22359 // CHECK19-NEXT: br i1 [[TMP133]], label [[OMP_OFFLOAD_FAILED43:%.*]], label [[OMP_OFFLOAD_CONT44:%.*]]
22360 // CHECK19: omp_offload.failed43:
22361 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163(i32 [[TMP104]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
22362 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT44]]
22363 // CHECK19: omp_offload.cont44:
22364 // CHECK19-NEXT: [[TMP134:%.*]] = load i32, i32* [[M]], align 4
22365 // CHECK19-NEXT: store i32 [[TMP134]], i32* [[M_CASTED45]], align 4
22366 // CHECK19-NEXT: [[TMP135:%.*]] = load i32, i32* [[M_CASTED45]], align 4
22367 // CHECK19-NEXT: [[TMP136:%.*]] = load i32, i32* [[N]], align 4
22368 // CHECK19-NEXT: store i32 [[TMP136]], i32* [[N_CASTED46]], align 4
22369 // CHECK19-NEXT: [[TMP137:%.*]] = load i32, i32* [[N_CASTED46]], align 4
22370 // CHECK19-NEXT: [[TMP138:%.*]] = mul nuw i32 [[TMP0]], 4
22371 // CHECK19-NEXT: [[TMP139:%.*]] = sext i32 [[TMP138]] to i64
22372 // CHECK19-NEXT: [[TMP140:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
22373 // CHECK19-NEXT: [[TMP141:%.*]] = bitcast i8** [[TMP140]] to i32*
22374 // CHECK19-NEXT: store i32 [[TMP135]], i32* [[TMP141]], align 4
22375 // CHECK19-NEXT: [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
22376 // CHECK19-NEXT: [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i32*
22377 // CHECK19-NEXT: store i32 [[TMP135]], i32* [[TMP143]], align 4
22378 // CHECK19-NEXT: [[TMP144:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
22379 // CHECK19-NEXT: store i64 4, i64* [[TMP144]], align 4
22380 // CHECK19-NEXT: [[TMP145:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 0
22381 // CHECK19-NEXT: store i8* null, i8** [[TMP145]], align 4
22382 // CHECK19-NEXT: [[TMP146:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 1
22383 // CHECK19-NEXT: [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i32*
22384 // CHECK19-NEXT: store i32 [[TMP137]], i32* [[TMP147]], align 4
22385 // CHECK19-NEXT: [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 1
22386 // CHECK19-NEXT: [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
22387 // CHECK19-NEXT: store i32 [[TMP137]], i32* [[TMP149]], align 4
22388 // CHECK19-NEXT: [[TMP150:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 1
22389 // CHECK19-NEXT: store i64 4, i64* [[TMP150]], align 4
22390 // CHECK19-NEXT: [[TMP151:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 1
22391 // CHECK19-NEXT: store i8* null, i8** [[TMP151]], align 4
22392 // CHECK19-NEXT: [[TMP152:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 2
22393 // CHECK19-NEXT: [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i32*
22394 // CHECK19-NEXT: store i32 [[TMP0]], i32* [[TMP153]], align 4
22395 // CHECK19-NEXT: [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 2
22396 // CHECK19-NEXT: [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32*
22397 // CHECK19-NEXT: store i32 [[TMP0]], i32* [[TMP155]], align 4
22398 // CHECK19-NEXT: [[TMP156:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 2
22399 // CHECK19-NEXT: store i64 4, i64* [[TMP156]], align 4
22400 // CHECK19-NEXT: [[TMP157:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 2
22401 // CHECK19-NEXT: store i8* null, i8** [[TMP157]], align 4
22402 // CHECK19-NEXT: [[TMP158:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 3
22403 // CHECK19-NEXT: [[TMP159:%.*]] = bitcast i8** [[TMP158]] to i32**
22404 // CHECK19-NEXT: store i32* [[VLA]], i32** [[TMP159]], align 4
22405 // CHECK19-NEXT: [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 3
22406 // CHECK19-NEXT: [[TMP161:%.*]] = bitcast i8** [[TMP160]] to i32**
22407 // CHECK19-NEXT: store i32* [[VLA]], i32** [[TMP161]], align 4
22408 // CHECK19-NEXT: [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 3
22409 // CHECK19-NEXT: store i64 [[TMP139]], i64* [[TMP162]], align 4
22410 // CHECK19-NEXT: [[TMP163:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 3
22411 // CHECK19-NEXT: store i8* null, i8** [[TMP163]], align 4
22412 // CHECK19-NEXT: [[TMP164:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
22413 // CHECK19-NEXT: [[TMP165:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
22414 // CHECK19-NEXT: [[TMP166:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
22415 // CHECK19-NEXT: [[TMP167:%.*]] = load i32, i32* [[N]], align 4
22416 // CHECK19-NEXT: store i32 [[TMP167]], i32* [[DOTCAPTURE_EXPR_52]], align 4
22417 // CHECK19-NEXT: [[TMP168:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_52]], align 4
22418 // CHECK19-NEXT: [[SUB54:%.*]] = sub nsw i32 [[TMP168]], 0
22419 // CHECK19-NEXT: [[DIV55:%.*]] = sdiv i32 [[SUB54]], 1
22420 // CHECK19-NEXT: [[SUB56:%.*]] = sub nsw i32 [[DIV55]], 1
22421 // CHECK19-NEXT: store i32 [[SUB56]], i32* [[DOTCAPTURE_EXPR_53]], align 4
22422 // CHECK19-NEXT: [[TMP169:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_53]], align 4
22423 // CHECK19-NEXT: [[ADD57:%.*]] = add nsw i32 [[TMP169]], 1
22424 // CHECK19-NEXT: [[TMP170:%.*]] = zext i32 [[ADD57]] to i64
22425 // CHECK19-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP170]])
22426 // CHECK19-NEXT: [[TMP171:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168.region_id, i32 4, i8** [[TMP164]], i8** [[TMP165]], i64* [[TMP166]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
22427 // CHECK19-NEXT: [[TMP172:%.*]] = icmp ne i32 [[TMP171]], 0
22428 // CHECK19-NEXT: br i1 [[TMP172]], label [[OMP_OFFLOAD_FAILED58:%.*]], label [[OMP_OFFLOAD_CONT59:%.*]]
22429 // CHECK19: omp_offload.failed58:
22430 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168(i32 [[TMP135]], i32 [[TMP137]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
22431 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT59]]
22432 // CHECK19: omp_offload.cont59:
22433 // CHECK19-NEXT: [[TMP173:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
22434 // CHECK19-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP173]])
22435 // CHECK19-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
22436 // CHECK19-NEXT: [[TMP174:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
22437 // CHECK19-NEXT: call void @llvm.stackrestore(i8* [[TMP174]])
22438 // CHECK19-NEXT: [[TMP175:%.*]] = load i32, i32* [[RETVAL]], align 4
22439 // CHECK19-NEXT: ret i32 [[TMP175]]
22440 //
22441 //
22442 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148
22443 // CHECK19-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
22444 // CHECK19-NEXT: entry:
22445 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
22446 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
22447 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
22448 // CHECK19-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
22449 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
22450 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
22451 // CHECK19-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
22452 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
22453 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
22454 // CHECK19-NEXT: ret void
22455 //
22456 //
22457 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined.
22458 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
22459 // CHECK19-NEXT: entry:
22460 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
22461 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
22462 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
22463 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
22464 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
22465 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
22466 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
22467 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
22468 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
22469 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
22470 // CHECK19-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
22471 // CHECK19-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
22472 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
22473 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
22474 // CHECK19-NEXT: [[I3:%.*]] = alloca i32, align 4
22475 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
22476 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
22477 // CHECK19-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
22478 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
22479 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
22480 // CHECK19-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
22481 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
22482 // CHECK19-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
22483 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
22484 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
22485 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
22486 // CHECK19-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
22487 // CHECK19-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
22488 // CHECK19-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
22489 // CHECK19-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
22490 // CHECK19-NEXT: store i32 0, i32* [[I]], align 4
22491 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
22492 // CHECK19-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
22493 // CHECK19-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
22494 // CHECK19: omp.precond.then:
22495 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
22496 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22497 // CHECK19-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
22498 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
22499 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
22500 // CHECK19-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
22501 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
22502 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
22503 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22504 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22505 // CHECK19-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
22506 // CHECK19-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
22507 // CHECK19: cond.true:
22508 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22509 // CHECK19-NEXT: br label [[COND_END:%.*]]
22510 // CHECK19: cond.false:
22511 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22512 // CHECK19-NEXT: br label [[COND_END]]
22513 // CHECK19: cond.end:
22514 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
22515 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
22516 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
22517 // CHECK19-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
22518 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
22519 // CHECK19: omp.inner.for.cond:
22520 // CHECK19-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
22521 // CHECK19-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22522 // CHECK19-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
22523 // CHECK19-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
22524 // CHECK19: omp.inner.for.body:
22525 // CHECK19-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
22526 // CHECK19-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22527 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
22528 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
22529 // CHECK19: omp.inner.for.inc:
22530 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
22531 // CHECK19-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
22532 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
22533 // CHECK19-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
22534 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
22535 // CHECK19: omp.inner.for.end:
22536 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
22537 // CHECK19: omp.loop.exit:
22538 // CHECK19-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
22539 // CHECK19-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
22540 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
22541 // CHECK19-NEXT: br label [[OMP_PRECOND_END]]
22542 // CHECK19: omp.precond.end:
22543 // CHECK19-NEXT: ret void
22544 //
22545 //
22546 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..1
22547 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
22548 // CHECK19-NEXT: entry:
22549 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
22550 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
22551 // CHECK19-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
22552 // CHECK19-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
22553 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
22554 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
22555 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
22556 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
22557 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
22558 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
22559 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
22560 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
22561 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
22562 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
22563 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
22564 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
22565 // CHECK19-NEXT: [[I3:%.*]] = alloca i32, align 4
22566 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
22567 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
22568 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
22569 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
22570 // CHECK19-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
22571 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
22572 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
22573 // CHECK19-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
22574 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
22575 // CHECK19-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
22576 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
22577 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
22578 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
22579 // CHECK19-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
22580 // CHECK19-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
22581 // CHECK19-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
22582 // CHECK19-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
22583 // CHECK19-NEXT: store i32 0, i32* [[I]], align 4
22584 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
22585 // CHECK19-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
22586 // CHECK19-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
22587 // CHECK19: omp.precond.then:
22588 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
22589 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22590 // CHECK19-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
22591 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
22592 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
22593 // CHECK19-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
22594 // CHECK19-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
22595 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
22596 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
22597 // CHECK19-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
22598 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
22599 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
22600 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
22601 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22602 // CHECK19-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
22603 // CHECK19-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
22604 // CHECK19: cond.true:
22605 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22606 // CHECK19-NEXT: br label [[COND_END:%.*]]
22607 // CHECK19: cond.false:
22608 // CHECK19-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
22609 // CHECK19-NEXT: br label [[COND_END]]
22610 // CHECK19: cond.end:
22611 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
22612 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
22613 // CHECK19-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
22614 // CHECK19-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
22615 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
22616 // CHECK19: omp.inner.for.cond:
22617 // CHECK19-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
22618 // CHECK19-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
22619 // CHECK19-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
22620 // CHECK19-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
22621 // CHECK19: omp.inner.for.body:
22622 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
22623 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
22624 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
22625 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
22626 // CHECK19-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
22627 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
22628 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
22629 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
22630 // CHECK19: omp.body.continue:
22631 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
22632 // CHECK19: omp.inner.for.inc:
22633 // CHECK19-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
22634 // CHECK19-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
22635 // CHECK19-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
22636 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
22637 // CHECK19: omp.inner.for.end:
22638 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
22639 // CHECK19: omp.loop.exit:
22640 // CHECK19-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
22641 // CHECK19-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
22642 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
22643 // CHECK19-NEXT: br label [[OMP_PRECOND_END]]
22644 // CHECK19: omp.precond.end:
22645 // CHECK19-NEXT: ret void
22646 //
22647 //
22648 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153
22649 // CHECK19-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
22650 // CHECK19-NEXT: entry:
22651 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
22652 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
22653 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
22654 // CHECK19-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
22655 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
22656 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
22657 // CHECK19-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
22658 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
22659 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
22660 // CHECK19-NEXT: ret void
22661 //
22662 //
22663 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..2
22664 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
22665 // CHECK19-NEXT: entry:
22666 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
22667 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
22668 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
22669 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
22670 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
22671 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
22672 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
22673 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
22674 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
22675 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
22676 // CHECK19-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
22677 // CHECK19-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
22678 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
22679 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
22680 // CHECK19-NEXT: [[I3:%.*]] = alloca i32, align 4
22681 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
22682 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
22683 // CHECK19-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
22684 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
22685 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
22686 // CHECK19-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
22687 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
22688 // CHECK19-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
22689 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
22690 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
22691 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
22692 // CHECK19-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
22693 // CHECK19-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
22694 // CHECK19-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
22695 // CHECK19-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
22696 // CHECK19-NEXT: store i32 0, i32* [[I]], align 4
22697 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
22698 // CHECK19-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
22699 // CHECK19-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
22700 // CHECK19: omp.precond.then:
22701 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
22702 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22703 // CHECK19-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
22704 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
22705 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
22706 // CHECK19-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
22707 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
22708 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
22709 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22710 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22711 // CHECK19-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
22712 // CHECK19-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
22713 // CHECK19: cond.true:
22714 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22715 // CHECK19-NEXT: br label [[COND_END:%.*]]
22716 // CHECK19: cond.false:
22717 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22718 // CHECK19-NEXT: br label [[COND_END]]
22719 // CHECK19: cond.end:
22720 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
22721 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
22722 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
22723 // CHECK19-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
22724 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
22725 // CHECK19: omp.inner.for.cond:
22726 // CHECK19-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
22727 // CHECK19-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22728 // CHECK19-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
22729 // CHECK19-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
22730 // CHECK19: omp.inner.for.body:
22731 // CHECK19-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
22732 // CHECK19-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22733 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
22734 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
22735 // CHECK19: omp.inner.for.inc:
22736 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
22737 // CHECK19-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
22738 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
22739 // CHECK19-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
22740 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
22741 // CHECK19: omp.inner.for.end:
22742 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
22743 // CHECK19: omp.loop.exit:
22744 // CHECK19-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
22745 // CHECK19-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
22746 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
22747 // CHECK19-NEXT: br label [[OMP_PRECOND_END]]
22748 // CHECK19: omp.precond.end:
22749 // CHECK19-NEXT: ret void
22750 //
22751 //
22752 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..3
22753 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
22754 // CHECK19-NEXT: entry:
22755 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
22756 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
22757 // CHECK19-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
22758 // CHECK19-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
22759 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
22760 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
22761 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
22762 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
22763 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
22764 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
22765 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
22766 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
22767 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
22768 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
22769 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
22770 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
22771 // CHECK19-NEXT: [[I3:%.*]] = alloca i32, align 4
22772 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
22773 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
22774 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
22775 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
22776 // CHECK19-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
22777 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
22778 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
22779 // CHECK19-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
22780 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
22781 // CHECK19-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
22782 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
22783 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
22784 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
22785 // CHECK19-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
22786 // CHECK19-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
22787 // CHECK19-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
22788 // CHECK19-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
22789 // CHECK19-NEXT: store i32 0, i32* [[I]], align 4
22790 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
22791 // CHECK19-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
22792 // CHECK19-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
22793 // CHECK19: omp.precond.then:
22794 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
22795 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22796 // CHECK19-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
22797 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
22798 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
22799 // CHECK19-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
22800 // CHECK19-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
22801 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
22802 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
22803 // CHECK19-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
22804 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
22805 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
22806 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
22807 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22808 // CHECK19-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
22809 // CHECK19-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
22810 // CHECK19: cond.true:
22811 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22812 // CHECK19-NEXT: br label [[COND_END:%.*]]
22813 // CHECK19: cond.false:
22814 // CHECK19-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
22815 // CHECK19-NEXT: br label [[COND_END]]
22816 // CHECK19: cond.end:
22817 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
22818 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
22819 // CHECK19-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
22820 // CHECK19-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
22821 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
22822 // CHECK19: omp.inner.for.cond:
22823 // CHECK19-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
22824 // CHECK19-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
22825 // CHECK19-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
22826 // CHECK19-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
22827 // CHECK19: omp.inner.for.body:
22828 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
22829 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
22830 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
22831 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
22832 // CHECK19-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
22833 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
22834 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
22835 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
22836 // CHECK19: omp.body.continue:
22837 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
22838 // CHECK19: omp.inner.for.inc:
22839 // CHECK19-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
22840 // CHECK19-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
22841 // CHECK19-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
22842 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
22843 // CHECK19: omp.inner.for.end:
22844 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
22845 // CHECK19: omp.loop.exit:
22846 // CHECK19-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
22847 // CHECK19-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
22848 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
22849 // CHECK19-NEXT: br label [[OMP_PRECOND_END]]
22850 // CHECK19: omp.precond.end:
22851 // CHECK19-NEXT: ret void
22852 //
22853 //
22854 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158
22855 // CHECK19-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
22856 // CHECK19-NEXT: entry:
22857 // CHECK19-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
22858 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
22859 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
22860 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
22861 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
22862 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
22863 // CHECK19-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
22864 // CHECK19-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
22865 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
22866 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
22867 // CHECK19-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
22868 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
22869 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
22870 // CHECK19-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
22871 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
22872 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
22873 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
22874 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
22875 // CHECK19-NEXT: ret void
22876 //
22877 //
22878 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..5
22879 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
22880 // CHECK19-NEXT: entry:
22881 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
22882 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
22883 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
22884 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
22885 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
22886 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
22887 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
22888 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
22889 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
22890 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
22891 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
22892 // CHECK19-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
22893 // CHECK19-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
22894 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
22895 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
22896 // CHECK19-NEXT: [[I4:%.*]] = alloca i32, align 4
22897 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
22898 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
22899 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
22900 // CHECK19-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
22901 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
22902 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
22903 // CHECK19-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
22904 // CHECK19-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
22905 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
22906 // CHECK19-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
22907 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
22908 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
22909 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22910 // CHECK19-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
22911 // CHECK19-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
22912 // CHECK19-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
22913 // CHECK19-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
22914 // CHECK19-NEXT: store i32 0, i32* [[I]], align 4
22915 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
22916 // CHECK19-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
22917 // CHECK19-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
22918 // CHECK19: omp.precond.then:
22919 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
22920 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
22921 // CHECK19-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
22922 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
22923 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
22924 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
22925 // CHECK19-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
22926 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
22927 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
22928 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22929 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
22930 // CHECK19-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
22931 // CHECK19-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
22932 // CHECK19: cond.true:
22933 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
22934 // CHECK19-NEXT: br label [[COND_END:%.*]]
22935 // CHECK19: cond.false:
22936 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22937 // CHECK19-NEXT: br label [[COND_END]]
22938 // CHECK19: cond.end:
22939 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
22940 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
22941 // CHECK19-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
22942 // CHECK19-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
22943 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
22944 // CHECK19: omp.inner.for.cond:
22945 // CHECK19-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
22946 // CHECK19-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
22947 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
22948 // CHECK19-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
22949 // CHECK19-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
22950 // CHECK19: omp.inner.for.body:
22951 // CHECK19-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
22952 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22953 // CHECK19-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
22954 // CHECK19-NEXT: store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
22955 // CHECK19-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
22956 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]])
22957 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
22958 // CHECK19: omp.inner.for.inc:
22959 // CHECK19-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
22960 // CHECK19-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
22961 // CHECK19-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
22962 // CHECK19-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
22963 // CHECK19-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
22964 // CHECK19-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
22965 // CHECK19-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
22966 // CHECK19-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4
22967 // CHECK19-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22968 // CHECK19-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
22969 // CHECK19-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
22970 // CHECK19-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4
22971 // CHECK19-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22972 // CHECK19-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
22973 // CHECK19-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
22974 // CHECK19-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
22975 // CHECK19: cond.true11:
22976 // CHECK19-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
22977 // CHECK19-NEXT: br label [[COND_END13:%.*]]
22978 // CHECK19: cond.false12:
22979 // CHECK19-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
22980 // CHECK19-NEXT: br label [[COND_END13]]
22981 // CHECK19: cond.end13:
22982 // CHECK19-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
22983 // CHECK19-NEXT: store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4
22984 // CHECK19-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
22985 // CHECK19-NEXT: store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
22986 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
22987 // CHECK19: omp.inner.for.end:
22988 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
22989 // CHECK19: omp.loop.exit:
22990 // CHECK19-NEXT: [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
22991 // CHECK19-NEXT: [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
22992 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
22993 // CHECK19-NEXT: br label [[OMP_PRECOND_END]]
22994 // CHECK19: omp.precond.end:
22995 // CHECK19-NEXT: ret void
22996 //
22997 //
22998 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..6
22999 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
23000 // CHECK19-NEXT: entry:
23001 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
23002 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
23003 // CHECK19-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
23004 // CHECK19-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
23005 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
23006 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
23007 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
23008 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
23009 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
23010 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
23011 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
23012 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
23013 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
23014 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
23015 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
23016 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
23017 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
23018 // CHECK19-NEXT: [[I4:%.*]] = alloca i32, align 4
23019 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
23020 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
23021 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
23022 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
23023 // CHECK19-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
23024 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
23025 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
23026 // CHECK19-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
23027 // CHECK19-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
23028 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
23029 // CHECK19-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
23030 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
23031 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
23032 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
23033 // CHECK19-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
23034 // CHECK19-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
23035 // CHECK19-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
23036 // CHECK19-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
23037 // CHECK19-NEXT: store i32 0, i32* [[I]], align 4
23038 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
23039 // CHECK19-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
23040 // CHECK19-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
23041 // CHECK19: omp.precond.then:
23042 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
23043 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
23044 // CHECK19-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
23045 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
23046 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
23047 // CHECK19-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
23048 // CHECK19-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
23049 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
23050 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
23051 // CHECK19-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23052 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
23053 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
23054 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
23055 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
23056 // CHECK19-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
23057 // CHECK19-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
23058 // CHECK19: cond.true:
23059 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
23060 // CHECK19-NEXT: br label [[COND_END:%.*]]
23061 // CHECK19: cond.false:
23062 // CHECK19-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
23063 // CHECK19-NEXT: br label [[COND_END]]
23064 // CHECK19: cond.end:
23065 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
23066 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
23067 // CHECK19-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
23068 // CHECK19-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
23069 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
23070 // CHECK19: omp.inner.for.cond:
23071 // CHECK19-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23072 // CHECK19-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
23073 // CHECK19-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
23074 // CHECK19-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
23075 // CHECK19: omp.inner.for.body:
23076 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23077 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
23078 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
23079 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
23080 // CHECK19-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
23081 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
23082 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
23083 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
23084 // CHECK19: omp.body.continue:
23085 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
23086 // CHECK19: omp.inner.for.inc:
23087 // CHECK19-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23088 // CHECK19-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
23089 // CHECK19-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
23090 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
23091 // CHECK19: omp.inner.for.end:
23092 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
23093 // CHECK19: omp.loop.exit:
23094 // CHECK19-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23095 // CHECK19-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
23096 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
23097 // CHECK19-NEXT: br label [[OMP_PRECOND_END]]
23098 // CHECK19: omp.precond.end:
23099 // CHECK19-NEXT: ret void
23100 //
23101 //
23102 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163
23103 // CHECK19-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
23104 // CHECK19-NEXT: entry:
23105 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
23106 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
23107 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
23108 // CHECK19-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
23109 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
23110 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
23111 // CHECK19-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
23112 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
23113 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
23114 // CHECK19-NEXT: ret void
23115 //
23116 //
23117 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..8
23118 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
23119 // CHECK19-NEXT: entry:
23120 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
23121 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
23122 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
23123 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
23124 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
23125 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
23126 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
23127 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
23128 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
23129 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
23130 // CHECK19-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
23131 // CHECK19-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
23132 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
23133 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
23134 // CHECK19-NEXT: [[I3:%.*]] = alloca i32, align 4
23135 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
23136 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
23137 // CHECK19-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
23138 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
23139 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
23140 // CHECK19-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
23141 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
23142 // CHECK19-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
23143 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
23144 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
23145 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
23146 // CHECK19-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
23147 // CHECK19-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
23148 // CHECK19-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
23149 // CHECK19-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
23150 // CHECK19-NEXT: store i32 0, i32* [[I]], align 4
23151 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
23152 // CHECK19-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
23153 // CHECK19-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
23154 // CHECK19: omp.precond.then:
23155 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
23156 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
23157 // CHECK19-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
23158 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
23159 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
23160 // CHECK19-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23161 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
23162 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
23163 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23164 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
23165 // CHECK19-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
23166 // CHECK19-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
23167 // CHECK19: cond.true:
23168 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
23169 // CHECK19-NEXT: br label [[COND_END:%.*]]
23170 // CHECK19: cond.false:
23171 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23172 // CHECK19-NEXT: br label [[COND_END]]
23173 // CHECK19: cond.end:
23174 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
23175 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
23176 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
23177 // CHECK19-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
23178 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
23179 // CHECK19: omp.inner.for.cond:
23180 // CHECK19-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23181 // CHECK19-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23182 // CHECK19-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
23183 // CHECK19-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
23184 // CHECK19: omp.inner.for.body:
23185 // CHECK19-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
23186 // CHECK19-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23187 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
23188 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
23189 // CHECK19: omp.inner.for.inc:
23190 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23191 // CHECK19-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
23192 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
23193 // CHECK19-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
23194 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
23195 // CHECK19: omp.inner.for.end:
23196 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
23197 // CHECK19: omp.loop.exit:
23198 // CHECK19-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23199 // CHECK19-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
23200 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
23201 // CHECK19-NEXT: br label [[OMP_PRECOND_END]]
23202 // CHECK19: omp.precond.end:
23203 // CHECK19-NEXT: ret void
23204 //
23205 //
23206 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..9
23207 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
23208 // CHECK19-NEXT: entry:
23209 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
23210 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
23211 // CHECK19-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
23212 // CHECK19-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
23213 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
23214 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
23215 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
23216 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
23217 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
23218 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
23219 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
23220 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
23221 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
23222 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
23223 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
23224 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
23225 // CHECK19-NEXT: [[I3:%.*]] = alloca i32, align 4
23226 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
23227 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
23228 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
23229 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
23230 // CHECK19-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
23231 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
23232 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
23233 // CHECK19-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
23234 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
23235 // CHECK19-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
23236 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
23237 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
23238 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
23239 // CHECK19-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
23240 // CHECK19-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
23241 // CHECK19-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
23242 // CHECK19-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
23243 // CHECK19-NEXT: store i32 0, i32* [[I]], align 4
23244 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
23245 // CHECK19-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
23246 // CHECK19-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
23247 // CHECK19: omp.precond.then:
23248 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
23249 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
23250 // CHECK19-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
23251 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
23252 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
23253 // CHECK19-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
23254 // CHECK19-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
23255 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
23256 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
23257 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
23258 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
23259 // CHECK19-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23260 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
23261 // CHECK19-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 35, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
23262 // CHECK19-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
23263 // CHECK19: omp.dispatch.cond:
23264 // CHECK19-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23265 // CHECK19-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
23266 // CHECK19-NEXT: [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
23267 // CHECK19-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
23268 // CHECK19-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
23269 // CHECK19: omp.dispatch.body:
23270 // CHECK19-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
23271 // CHECK19-NEXT: store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
23272 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
23273 // CHECK19: omp.inner.for.cond:
23274 // CHECK19-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
23275 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !16
23276 // CHECK19-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
23277 // CHECK19-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
23278 // CHECK19: omp.inner.for.body:
23279 // CHECK19-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
23280 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
23281 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
23282 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !16
23283 // CHECK19-NEXT: [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !16
23284 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
23285 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !16
23286 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
23287 // CHECK19: omp.body.continue:
23288 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
23289 // CHECK19: omp.inner.for.inc:
23290 // CHECK19-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
23291 // CHECK19-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP21]], 1
23292 // CHECK19-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
23293 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
23294 // CHECK19: omp.inner.for.end:
23295 // CHECK19-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
23296 // CHECK19: omp.dispatch.inc:
23297 // CHECK19-NEXT: br label [[OMP_DISPATCH_COND]]
23298 // CHECK19: omp.dispatch.end:
23299 // CHECK19-NEXT: br label [[OMP_PRECOND_END]]
23300 // CHECK19: omp.precond.end:
23301 // CHECK19-NEXT: ret void
23302 //
23303 //
23304 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168
23305 // CHECK19-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
23306 // CHECK19-NEXT: entry:
23307 // CHECK19-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
23308 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
23309 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
23310 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
23311 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
23312 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
23313 // CHECK19-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
23314 // CHECK19-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
23315 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
23316 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
23317 // CHECK19-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
23318 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
23319 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
23320 // CHECK19-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
23321 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
23322 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
23323 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
23324 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
23325 // CHECK19-NEXT: ret void
23326 //
23327 //
23328 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..11
23329 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
23330 // CHECK19-NEXT: entry:
23331 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
23332 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
23333 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
23334 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
23335 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
23336 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
23337 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
23338 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
23339 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
23340 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
23341 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
23342 // CHECK19-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
23343 // CHECK19-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
23344 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
23345 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
23346 // CHECK19-NEXT: [[I4:%.*]] = alloca i32, align 4
23347 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
23348 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
23349 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
23350 // CHECK19-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
23351 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
23352 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
23353 // CHECK19-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
23354 // CHECK19-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
23355 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
23356 // CHECK19-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
23357 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
23358 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
23359 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
23360 // CHECK19-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
23361 // CHECK19-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
23362 // CHECK19-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
23363 // CHECK19-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
23364 // CHECK19-NEXT: store i32 0, i32* [[I]], align 4
23365 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
23366 // CHECK19-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
23367 // CHECK19-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
23368 // CHECK19: omp.precond.then:
23369 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
23370 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
23371 // CHECK19-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
23372 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
23373 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
23374 // CHECK19-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23375 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
23376 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
23377 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23378 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
23379 // CHECK19-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
23380 // CHECK19-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
23381 // CHECK19: cond.true:
23382 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
23383 // CHECK19-NEXT: br label [[COND_END:%.*]]
23384 // CHECK19: cond.false:
23385 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23386 // CHECK19-NEXT: br label [[COND_END]]
23387 // CHECK19: cond.end:
23388 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
23389 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
23390 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
23391 // CHECK19-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
23392 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
23393 // CHECK19: omp.inner.for.cond:
23394 // CHECK19-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23395 // CHECK19-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23396 // CHECK19-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
23397 // CHECK19-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
23398 // CHECK19: omp.inner.for.body:
23399 // CHECK19-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
23400 // CHECK19-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23401 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
23402 // CHECK19-NEXT: store i32 [[TMP18]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
23403 // CHECK19-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
23404 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP19]])
23405 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
23406 // CHECK19: omp.inner.for.inc:
23407 // CHECK19-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23408 // CHECK19-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
23409 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
23410 // CHECK19-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
23411 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
23412 // CHECK19: omp.inner.for.end:
23413 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
23414 // CHECK19: omp.loop.exit:
23415 // CHECK19-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23416 // CHECK19-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
23417 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
23418 // CHECK19-NEXT: br label [[OMP_PRECOND_END]]
23419 // CHECK19: omp.precond.end:
23420 // CHECK19-NEXT: ret void
23421 //
23422 //
23423 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..12
23424 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
23425 // CHECK19-NEXT: entry:
23426 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
23427 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
23428 // CHECK19-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
23429 // CHECK19-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
23430 // CHECK19-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
23431 // CHECK19-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
23432 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
23433 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
23434 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
23435 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
23436 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
23437 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
23438 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
23439 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
23440 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
23441 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
23442 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
23443 // CHECK19-NEXT: [[I4:%.*]] = alloca i32, align 4
23444 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
23445 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
23446 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
23447 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
23448 // CHECK19-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
23449 // CHECK19-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
23450 // CHECK19-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
23451 // CHECK19-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
23452 // CHECK19-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
23453 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
23454 // CHECK19-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
23455 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
23456 // CHECK19-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
23457 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
23458 // CHECK19-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
23459 // CHECK19-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
23460 // CHECK19-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
23461 // CHECK19-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
23462 // CHECK19-NEXT: store i32 0, i32* [[I]], align 4
23463 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
23464 // CHECK19-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
23465 // CHECK19-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
23466 // CHECK19: omp.precond.then:
23467 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
23468 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
23469 // CHECK19-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
23470 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
23471 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
23472 // CHECK19-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
23473 // CHECK19-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
23474 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
23475 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
23476 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
23477 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
23478 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
23479 // CHECK19-NEXT: [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23480 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
23481 // CHECK19-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 35, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
23482 // CHECK19-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
23483 // CHECK19: omp.dispatch.cond:
23484 // CHECK19-NEXT: [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23485 // CHECK19-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
23486 // CHECK19-NEXT: [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
23487 // CHECK19-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
23488 // CHECK19-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
23489 // CHECK19: omp.dispatch.body:
23490 // CHECK19-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
23491 // CHECK19-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
23492 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
23493 // CHECK19: omp.inner.for.cond:
23494 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
23495 // CHECK19-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
23496 // CHECK19-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
23497 // CHECK19-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
23498 // CHECK19: omp.inner.for.body:
23499 // CHECK19-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
23500 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
23501 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
23502 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !19
23503 // CHECK19-NEXT: [[TMP21:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !19
23504 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP21]]
23505 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !19
23506 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
23507 // CHECK19: omp.body.continue:
23508 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
23509 // CHECK19: omp.inner.for.inc:
23510 // CHECK19-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
23511 // CHECK19-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
23512 // CHECK19-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
23513 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
23514 // CHECK19: omp.inner.for.end:
23515 // CHECK19-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
23516 // CHECK19: omp.dispatch.inc:
23517 // CHECK19-NEXT: br label [[OMP_DISPATCH_COND]]
23518 // CHECK19: omp.dispatch.end:
23519 // CHECK19-NEXT: br label [[OMP_PRECOND_END]]
23520 // CHECK19: omp.precond.end:
23521 // CHECK19-NEXT: ret void
23522 //
23523 //
23524 // CHECK19-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
23525 // CHECK19-SAME: (i32 [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
23526 // CHECK19-NEXT: entry:
23527 // CHECK19-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
23528 // CHECK19-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
23529 // CHECK19-NEXT: [[M:%.*]] = alloca i32, align 4
23530 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
23531 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
23532 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
23533 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
23534 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
23535 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
23536 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
23537 // CHECK19-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
23538 // CHECK19-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
23539 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 4
23540 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 4
23541 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 4
23542 // CHECK19-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
23543 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 4
23544 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 4
23545 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 4
23546 // CHECK19-NEXT: [[_TMP16:%.*]] = alloca i32, align 4
23547 // CHECK19-NEXT: [[M_CASTED19:%.*]] = alloca i32, align 4
23548 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [2 x i8*], align 4
23549 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS21:%.*]] = alloca [2 x i8*], align 4
23550 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [2 x i8*], align 4
23551 // CHECK19-NEXT: [[_TMP23:%.*]] = alloca i32, align 4
23552 // CHECK19-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
23553 // CHECK19-NEXT: store i32 10, i32* [[M]], align 4
23554 // CHECK19-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
23555 // CHECK19-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
23556 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
23557 // CHECK19-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
23558 // CHECK19-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
23559 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
23560 // CHECK19-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
23561 // CHECK19-NEXT: store i8* null, i8** [[TMP4]], align 4
23562 // CHECK19-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
23563 // CHECK19-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
23564 // CHECK19-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
23565 // CHECK19-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
23566 // CHECK19-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
23567 // CHECK19-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
23568 // CHECK19: omp_offload.failed:
23569 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116([10 x i32]* [[A]]) #[[ATTR3]]
23570 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT]]
23571 // CHECK19: omp_offload.cont:
23572 // CHECK19-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
23573 // CHECK19-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
23574 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 4
23575 // CHECK19-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
23576 // CHECK19-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
23577 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 4
23578 // CHECK19-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
23579 // CHECK19-NEXT: store i8* null, i8** [[TMP13]], align 4
23580 // CHECK19-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
23581 // CHECK19-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
23582 // CHECK19-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
23583 // CHECK19-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
23584 // CHECK19-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
23585 // CHECK19-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
23586 // CHECK19: omp_offload.failed5:
23587 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121([10 x i32]* [[A]]) #[[ATTR3]]
23588 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT6]]
23589 // CHECK19: omp_offload.cont6:
23590 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[M]], align 4
23591 // CHECK19-NEXT: store i32 [[TMP18]], i32* [[M_CASTED]], align 4
23592 // CHECK19-NEXT: [[TMP19:%.*]] = load i32, i32* [[M_CASTED]], align 4
23593 // CHECK19-NEXT: [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
23594 // CHECK19-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
23595 // CHECK19-NEXT: store i32 [[TMP19]], i32* [[TMP21]], align 4
23596 // CHECK19-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
23597 // CHECK19-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
23598 // CHECK19-NEXT: store i32 [[TMP19]], i32* [[TMP23]], align 4
23599 // CHECK19-NEXT: [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 0
23600 // CHECK19-NEXT: store i8* null, i8** [[TMP24]], align 4
23601 // CHECK19-NEXT: [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
23602 // CHECK19-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
23603 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 4
23604 // CHECK19-NEXT: [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
23605 // CHECK19-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
23606 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 4
23607 // CHECK19-NEXT: [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 1
23608 // CHECK19-NEXT: store i8* null, i8** [[TMP29]], align 4
23609 // CHECK19-NEXT: [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
23610 // CHECK19-NEXT: [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
23611 // CHECK19-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
23612 // CHECK19-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
23613 // CHECK19-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
23614 // CHECK19-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
23615 // CHECK19: omp_offload.failed11:
23616 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126(i32 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
23617 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT12]]
23618 // CHECK19: omp_offload.cont12:
23619 // CHECK19-NEXT: [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
23620 // CHECK19-NEXT: [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
23621 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 4
23622 // CHECK19-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
23623 // CHECK19-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
23624 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 4
23625 // CHECK19-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
23626 // CHECK19-NEXT: store i8* null, i8** [[TMP38]], align 4
23627 // CHECK19-NEXT: [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
23628 // CHECK19-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
23629 // CHECK19-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
23630 // CHECK19-NEXT: [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
23631 // CHECK19-NEXT: [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
23632 // CHECK19-NEXT: br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
23633 // CHECK19: omp_offload.failed17:
23634 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131([10 x i32]* [[A]]) #[[ATTR3]]
23635 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT18]]
23636 // CHECK19: omp_offload.cont18:
23637 // CHECK19-NEXT: [[TMP43:%.*]] = load i32, i32* [[M]], align 4
23638 // CHECK19-NEXT: store i32 [[TMP43]], i32* [[M_CASTED19]], align 4
23639 // CHECK19-NEXT: [[TMP44:%.*]] = load i32, i32* [[M_CASTED19]], align 4
23640 // CHECK19-NEXT: [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
23641 // CHECK19-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
23642 // CHECK19-NEXT: store i32 [[TMP44]], i32* [[TMP46]], align 4
23643 // CHECK19-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
23644 // CHECK19-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i32*
23645 // CHECK19-NEXT: store i32 [[TMP44]], i32* [[TMP48]], align 4
23646 // CHECK19-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
23647 // CHECK19-NEXT: store i8* null, i8** [[TMP49]], align 4
23648 // CHECK19-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
23649 // CHECK19-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
23650 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 4
23651 // CHECK19-NEXT: [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
23652 // CHECK19-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
23653 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 4
23654 // CHECK19-NEXT: [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
23655 // CHECK19-NEXT: store i8* null, i8** [[TMP54]], align 4
23656 // CHECK19-NEXT: [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
23657 // CHECK19-NEXT: [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
23658 // CHECK19-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
23659 // CHECK19-NEXT: [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
23660 // CHECK19-NEXT: [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
23661 // CHECK19-NEXT: br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
23662 // CHECK19: omp_offload.failed24:
23663 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136(i32 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
23664 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT25]]
23665 // CHECK19: omp_offload.cont25:
23666 // CHECK19-NEXT: ret i32 0
23667 //
23668 //
23669 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116
23670 // CHECK19-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
23671 // CHECK19-NEXT: entry:
23672 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
23673 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
23674 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
23675 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
23676 // CHECK19-NEXT: ret void
23677 //
23678 //
23679 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..14
23680 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
23681 // CHECK19-NEXT: entry:
23682 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
23683 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
23684 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
23685 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
23686 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
23687 // CHECK19-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
23688 // CHECK19-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
23689 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
23690 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
23691 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
23692 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
23693 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
23694 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
23695 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
23696 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
23697 // CHECK19-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
23698 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
23699 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
23700 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23701 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
23702 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
23703 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23704 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
23705 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
23706 // CHECK19: cond.true:
23707 // CHECK19-NEXT: br label [[COND_END:%.*]]
23708 // CHECK19: cond.false:
23709 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23710 // CHECK19-NEXT: br label [[COND_END]]
23711 // CHECK19: cond.end:
23712 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
23713 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
23714 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
23715 // CHECK19-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
23716 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
23717 // CHECK19: omp.inner.for.cond:
23718 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23719 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23720 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
23721 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
23722 // CHECK19: omp.inner.for.body:
23723 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
23724 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23725 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
23726 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
23727 // CHECK19: omp.inner.for.inc:
23728 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23729 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
23730 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
23731 // CHECK19-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
23732 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
23733 // CHECK19: omp.inner.for.end:
23734 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
23735 // CHECK19: omp.loop.exit:
23736 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
23737 // CHECK19-NEXT: ret void
23738 //
23739 //
23740 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..15
23741 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
23742 // CHECK19-NEXT: entry:
23743 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
23744 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
23745 // CHECK19-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
23746 // CHECK19-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
23747 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
23748 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
23749 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
23750 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
23751 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
23752 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
23753 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
23754 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
23755 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
23756 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
23757 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
23758 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
23759 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
23760 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
23761 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
23762 // CHECK19-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
23763 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
23764 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
23765 // CHECK19-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
23766 // CHECK19-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
23767 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
23768 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
23769 // CHECK19-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23770 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
23771 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
23772 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
23773 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
23774 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
23775 // CHECK19: cond.true:
23776 // CHECK19-NEXT: br label [[COND_END:%.*]]
23777 // CHECK19: cond.false:
23778 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
23779 // CHECK19-NEXT: br label [[COND_END]]
23780 // CHECK19: cond.end:
23781 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
23782 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
23783 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
23784 // CHECK19-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
23785 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
23786 // CHECK19: omp.inner.for.cond:
23787 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23788 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
23789 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
23790 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
23791 // CHECK19: omp.inner.for.body:
23792 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23793 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
23794 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
23795 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I]], align 4
23796 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
23797 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
23798 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
23799 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
23800 // CHECK19: omp.body.continue:
23801 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
23802 // CHECK19: omp.inner.for.inc:
23803 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23804 // CHECK19-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
23805 // CHECK19-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
23806 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
23807 // CHECK19: omp.inner.for.end:
23808 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
23809 // CHECK19: omp.loop.exit:
23810 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
23811 // CHECK19-NEXT: ret void
23812 //
23813 //
23814 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121
23815 // CHECK19-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
23816 // CHECK19-NEXT: entry:
23817 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
23818 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
23819 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
23820 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
23821 // CHECK19-NEXT: ret void
23822 //
23823 //
23824 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..17
23825 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
23826 // CHECK19-NEXT: entry:
23827 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
23828 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
23829 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
23830 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
23831 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
23832 // CHECK19-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
23833 // CHECK19-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
23834 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
23835 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
23836 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
23837 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
23838 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
23839 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
23840 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
23841 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
23842 // CHECK19-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
23843 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
23844 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
23845 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23846 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
23847 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
23848 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23849 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
23850 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
23851 // CHECK19: cond.true:
23852 // CHECK19-NEXT: br label [[COND_END:%.*]]
23853 // CHECK19: cond.false:
23854 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23855 // CHECK19-NEXT: br label [[COND_END]]
23856 // CHECK19: cond.end:
23857 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
23858 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
23859 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
23860 // CHECK19-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
23861 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
23862 // CHECK19: omp.inner.for.cond:
23863 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23864 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23865 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
23866 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
23867 // CHECK19: omp.inner.for.body:
23868 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
23869 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
23870 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
23871 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
23872 // CHECK19: omp.inner.for.inc:
23873 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23874 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
23875 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
23876 // CHECK19-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
23877 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
23878 // CHECK19: omp.inner.for.end:
23879 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
23880 // CHECK19: omp.loop.exit:
23881 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
23882 // CHECK19-NEXT: ret void
23883 //
23884 //
23885 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..18
23886 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
23887 // CHECK19-NEXT: entry:
23888 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
23889 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
23890 // CHECK19-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
23891 // CHECK19-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
23892 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
23893 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
23894 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
23895 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
23896 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
23897 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
23898 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
23899 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
23900 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
23901 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
23902 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
23903 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
23904 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
23905 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
23906 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
23907 // CHECK19-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
23908 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
23909 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
23910 // CHECK19-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
23911 // CHECK19-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
23912 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
23913 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
23914 // CHECK19-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
23915 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
23916 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
23917 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
23918 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
23919 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
23920 // CHECK19: cond.true:
23921 // CHECK19-NEXT: br label [[COND_END:%.*]]
23922 // CHECK19: cond.false:
23923 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
23924 // CHECK19-NEXT: br label [[COND_END]]
23925 // CHECK19: cond.end:
23926 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
23927 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
23928 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
23929 // CHECK19-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
23930 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
23931 // CHECK19: omp.inner.for.cond:
23932 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23933 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
23934 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
23935 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
23936 // CHECK19: omp.inner.for.body:
23937 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23938 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
23939 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
23940 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I]], align 4
23941 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
23942 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
23943 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
23944 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
23945 // CHECK19: omp.body.continue:
23946 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
23947 // CHECK19: omp.inner.for.inc:
23948 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
23949 // CHECK19-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
23950 // CHECK19-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
23951 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
23952 // CHECK19: omp.inner.for.end:
23953 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
23954 // CHECK19: omp.loop.exit:
23955 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
23956 // CHECK19-NEXT: ret void
23957 //
23958 //
23959 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126
23960 // CHECK19-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
23961 // CHECK19-NEXT: entry:
23962 // CHECK19-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
23963 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
23964 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
23965 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
23966 // CHECK19-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
23967 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
23968 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
23969 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
23970 // CHECK19-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
23971 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
23972 // CHECK19-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
23973 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
23974 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
23975 // CHECK19-NEXT: ret void
23976 //
23977 //
23978 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..21
23979 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
23980 // CHECK19-NEXT: entry:
23981 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
23982 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
23983 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
23984 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
23985 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
23986 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
23987 // CHECK19-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
23988 // CHECK19-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
23989 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
23990 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
23991 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
23992 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
23993 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
23994 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
23995 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
23996 // CHECK19-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
23997 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
23998 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
23999 // CHECK19-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
24000 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
24001 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
24002 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
24003 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
24004 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
24005 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24006 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
24007 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
24008 // CHECK19: cond.true:
24009 // CHECK19-NEXT: br label [[COND_END:%.*]]
24010 // CHECK19: cond.false:
24011 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24012 // CHECK19-NEXT: br label [[COND_END]]
24013 // CHECK19: cond.end:
24014 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
24015 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
24016 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
24017 // CHECK19-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
24018 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
24019 // CHECK19: omp.inner.for.cond:
24020 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
24021 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24022 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
24023 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
24024 // CHECK19: omp.inner.for.body:
24025 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
24026 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24027 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
24028 // CHECK19-NEXT: store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
24029 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
24030 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
24031 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
24032 // CHECK19: omp.inner.for.inc:
24033 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
24034 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
24035 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
24036 // CHECK19-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
24037 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
24038 // CHECK19: omp.inner.for.end:
24039 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
24040 // CHECK19: omp.loop.exit:
24041 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
24042 // CHECK19-NEXT: ret void
24043 //
24044 //
24045 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..22
24046 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
24047 // CHECK19-NEXT: entry:
24048 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
24049 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
24050 // CHECK19-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
24051 // CHECK19-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
24052 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
24053 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
24054 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
24055 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
24056 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
24057 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
24058 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
24059 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
24060 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
24061 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
24062 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
24063 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
24064 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
24065 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
24066 // CHECK19-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
24067 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
24068 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
24069 // CHECK19-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
24070 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
24071 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
24072 // CHECK19-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
24073 // CHECK19-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
24074 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
24075 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
24076 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
24077 // CHECK19-NEXT: [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
24078 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
24079 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
24080 // CHECK19-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
24081 // CHECK19: omp.dispatch.cond:
24082 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
24083 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
24084 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
24085 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
24086 // CHECK19: cond.true:
24087 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
24088 // CHECK19-NEXT: br label [[COND_END:%.*]]
24089 // CHECK19: cond.false:
24090 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
24091 // CHECK19-NEXT: br label [[COND_END]]
24092 // CHECK19: cond.end:
24093 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
24094 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
24095 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
24096 // CHECK19-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
24097 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
24098 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
24099 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
24100 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
24101 // CHECK19: omp.dispatch.body:
24102 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
24103 // CHECK19: omp.inner.for.cond:
24104 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
24105 // CHECK19-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
24106 // CHECK19-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
24107 // CHECK19-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
24108 // CHECK19: omp.inner.for.body:
24109 // CHECK19-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
24110 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
24111 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
24112 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I]], align 4
24113 // CHECK19-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4
24114 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP16]]
24115 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
24116 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
24117 // CHECK19: omp.body.continue:
24118 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
24119 // CHECK19: omp.inner.for.inc:
24120 // CHECK19-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
24121 // CHECK19-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
24122 // CHECK19-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
24123 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
24124 // CHECK19: omp.inner.for.end:
24125 // CHECK19-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
24126 // CHECK19: omp.dispatch.inc:
24127 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
24128 // CHECK19-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
24129 // CHECK19-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
24130 // CHECK19-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
24131 // CHECK19-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
24132 // CHECK19-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
24133 // CHECK19-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
24134 // CHECK19-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
24135 // CHECK19-NEXT: br label [[OMP_DISPATCH_COND]]
24136 // CHECK19: omp.dispatch.end:
24137 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
24138 // CHECK19-NEXT: ret void
24139 //
24140 //
24141 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
24142 // CHECK19-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
24143 // CHECK19-NEXT: entry:
24144 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
24145 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
24146 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
24147 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
24148 // CHECK19-NEXT: ret void
24149 //
24150 //
24151 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..25
24152 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
24153 // CHECK19-NEXT: entry:
24154 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
24155 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
24156 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
24157 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
24158 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
24159 // CHECK19-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
24160 // CHECK19-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
24161 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
24162 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
24163 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
24164 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
24165 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
24166 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
24167 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
24168 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
24169 // CHECK19-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
24170 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
24171 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
24172 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
24173 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
24174 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
24175 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24176 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
24177 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
24178 // CHECK19: cond.true:
24179 // CHECK19-NEXT: br label [[COND_END:%.*]]
24180 // CHECK19: cond.false:
24181 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24182 // CHECK19-NEXT: br label [[COND_END]]
24183 // CHECK19: cond.end:
24184 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
24185 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
24186 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
24187 // CHECK19-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
24188 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
24189 // CHECK19: omp.inner.for.cond:
24190 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
24191 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24192 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
24193 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
24194 // CHECK19: omp.inner.for.body:
24195 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
24196 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24197 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
24198 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
24199 // CHECK19: omp.inner.for.inc:
24200 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
24201 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
24202 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
24203 // CHECK19-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
24204 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
24205 // CHECK19: omp.inner.for.end:
24206 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
24207 // CHECK19: omp.loop.exit:
24208 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
24209 // CHECK19-NEXT: ret void
24210 //
24211 //
24212 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..26
24213 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
24214 // CHECK19-NEXT: entry:
24215 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
24216 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
24217 // CHECK19-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
24218 // CHECK19-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
24219 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
24220 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
24221 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
24222 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
24223 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
24224 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
24225 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
24226 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
24227 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
24228 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
24229 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
24230 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
24231 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
24232 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
24233 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
24234 // CHECK19-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
24235 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
24236 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
24237 // CHECK19-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
24238 // CHECK19-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
24239 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
24240 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
24241 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
24242 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
24243 // CHECK19-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
24244 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
24245 // CHECK19-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
24246 // CHECK19-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
24247 // CHECK19: omp.dispatch.cond:
24248 // CHECK19-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
24249 // CHECK19-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
24250 // CHECK19-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
24251 // CHECK19: omp.dispatch.body:
24252 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
24253 // CHECK19-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
24254 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
24255 // CHECK19: omp.inner.for.cond:
24256 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
24257 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
24258 // CHECK19-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
24259 // CHECK19-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
24260 // CHECK19: omp.inner.for.body:
24261 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
24262 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
24263 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
24264 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
24265 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !22
24266 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP12]]
24267 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !22
24268 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
24269 // CHECK19: omp.body.continue:
24270 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
24271 // CHECK19: omp.inner.for.inc:
24272 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
24273 // CHECK19-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
24274 // CHECK19-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
24275 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
24276 // CHECK19: omp.inner.for.end:
24277 // CHECK19-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
24278 // CHECK19: omp.dispatch.inc:
24279 // CHECK19-NEXT: br label [[OMP_DISPATCH_COND]]
24280 // CHECK19: omp.dispatch.end:
24281 // CHECK19-NEXT: ret void
24282 //
24283 //
24284 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136
24285 // CHECK19-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
24286 // CHECK19-NEXT: entry:
24287 // CHECK19-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
24288 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
24289 // CHECK19-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
24290 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
24291 // CHECK19-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
24292 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
24293 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
24294 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
24295 // CHECK19-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
24296 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
24297 // CHECK19-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
24298 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
24299 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
24300 // CHECK19-NEXT: ret void
24301 //
24302 //
24303 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..29
24304 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
24305 // CHECK19-NEXT: entry:
24306 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
24307 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
24308 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
24309 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
24310 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
24311 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
24312 // CHECK19-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
24313 // CHECK19-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
24314 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
24315 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
24316 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
24317 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
24318 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
24319 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
24320 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
24321 // CHECK19-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
24322 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
24323 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
24324 // CHECK19-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
24325 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
24326 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
24327 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
24328 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
24329 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
24330 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24331 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
24332 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
24333 // CHECK19: cond.true:
24334 // CHECK19-NEXT: br label [[COND_END:%.*]]
24335 // CHECK19: cond.false:
24336 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24337 // CHECK19-NEXT: br label [[COND_END]]
24338 // CHECK19: cond.end:
24339 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
24340 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
24341 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
24342 // CHECK19-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
24343 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
24344 // CHECK19: omp.inner.for.cond:
24345 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
24346 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24347 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
24348 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
24349 // CHECK19: omp.inner.for.body:
24350 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
24351 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24352 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
24353 // CHECK19-NEXT: store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
24354 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
24355 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
24356 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
24357 // CHECK19: omp.inner.for.inc:
24358 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
24359 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
24360 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
24361 // CHECK19-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
24362 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
24363 // CHECK19: omp.inner.for.end:
24364 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
24365 // CHECK19: omp.loop.exit:
24366 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
24367 // CHECK19-NEXT: ret void
24368 //
24369 //
24370 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined..30
24371 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
24372 // CHECK19-NEXT: entry:
24373 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
24374 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
24375 // CHECK19-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
24376 // CHECK19-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
24377 // CHECK19-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
24378 // CHECK19-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
24379 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
24380 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
24381 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
24382 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
24383 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
24384 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
24385 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
24386 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
24387 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
24388 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
24389 // CHECK19-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
24390 // CHECK19-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
24391 // CHECK19-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
24392 // CHECK19-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
24393 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
24394 // CHECK19-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
24395 // CHECK19-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
24396 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
24397 // CHECK19-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
24398 // CHECK19-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
24399 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
24400 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
24401 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
24402 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
24403 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
24404 // CHECK19-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
24405 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
24406 // CHECK19-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 35, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
24407 // CHECK19-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
24408 // CHECK19: omp.dispatch.cond:
24409 // CHECK19-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
24410 // CHECK19-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
24411 // CHECK19-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
24412 // CHECK19: omp.dispatch.body:
24413 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
24414 // CHECK19-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
24415 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
24416 // CHECK19: omp.inner.for.cond:
24417 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
24418 // CHECK19-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
24419 // CHECK19-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
24420 // CHECK19-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
24421 // CHECK19: omp.inner.for.body:
24422 // CHECK19-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
24423 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
24424 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
24425 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !25
24426 // CHECK19-NEXT: [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !25
24427 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
24428 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
24429 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
24430 // CHECK19: omp.body.continue:
24431 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
24432 // CHECK19: omp.inner.for.inc:
24433 // CHECK19-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
24434 // CHECK19-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP14]], 1
24435 // CHECK19-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
24436 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
24437 // CHECK19: omp.inner.for.end:
24438 // CHECK19-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
24439 // CHECK19: omp.dispatch.inc:
24440 // CHECK19-NEXT: br label [[OMP_DISPATCH_COND]]
24441 // CHECK19: omp.dispatch.end:
24442 // CHECK19-NEXT: ret void
24443 //
24444 //
24445 // CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
24446 // CHECK19-SAME: () #[[ATTR5:[0-9]+]] {
24447 // CHECK19-NEXT: entry:
24448 // CHECK19-NEXT: call void @__tgt_register_requires(i64 1)
24449 // CHECK19-NEXT: ret void
24450 //
24451 //
24452 // CHECK20-LABEL: define {{[^@]+}}@main
24453 // CHECK20-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
24454 // CHECK20-NEXT: entry:
24455 // CHECK20-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
24456 // CHECK20-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
24457 // CHECK20-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
24458 // CHECK20-NEXT: [[N:%.*]] = alloca i32, align 4
24459 // CHECK20-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
24460 // CHECK20-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
24461 // CHECK20-NEXT: [[M:%.*]] = alloca i32, align 4
24462 // CHECK20-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
24463 // CHECK20-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
24464 // CHECK20-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
24465 // CHECK20-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
24466 // CHECK20-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
24467 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
24468 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
24469 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
24470 // CHECK20-NEXT: [[N_CASTED3:%.*]] = alloca i32, align 4
24471 // CHECK20-NEXT: [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
24472 // CHECK20-NEXT: [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
24473 // CHECK20-NEXT: [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
24474 // CHECK20-NEXT: [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
24475 // CHECK20-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
24476 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
24477 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
24478 // CHECK20-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
24479 // CHECK20-NEXT: [[N_CASTED17:%.*]] = alloca i32, align 4
24480 // CHECK20-NEXT: [[DOTOFFLOAD_BASEPTRS18:%.*]] = alloca [4 x i8*], align 4
24481 // CHECK20-NEXT: [[DOTOFFLOAD_PTRS19:%.*]] = alloca [4 x i8*], align 4
24482 // CHECK20-NEXT: [[DOTOFFLOAD_MAPPERS20:%.*]] = alloca [4 x i8*], align 4
24483 // CHECK20-NEXT: [[DOTOFFLOAD_SIZES21:%.*]] = alloca [4 x i64], align 4
24484 // CHECK20-NEXT: [[_TMP22:%.*]] = alloca i32, align 4
24485 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_23:%.*]] = alloca i32, align 4
24486 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
24487 // CHECK20-NEXT: [[N_CASTED31:%.*]] = alloca i32, align 4
24488 // CHECK20-NEXT: [[DOTOFFLOAD_BASEPTRS32:%.*]] = alloca [3 x i8*], align 4
24489 // CHECK20-NEXT: [[DOTOFFLOAD_PTRS33:%.*]] = alloca [3 x i8*], align 4
24490 // CHECK20-NEXT: [[DOTOFFLOAD_MAPPERS34:%.*]] = alloca [3 x i8*], align 4
24491 // CHECK20-NEXT: [[DOTOFFLOAD_SIZES35:%.*]] = alloca [3 x i64], align 4
24492 // CHECK20-NEXT: [[_TMP36:%.*]] = alloca i32, align 4
24493 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_37:%.*]] = alloca i32, align 4
24494 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
24495 // CHECK20-NEXT: [[M_CASTED45:%.*]] = alloca i32, align 4
24496 // CHECK20-NEXT: [[N_CASTED46:%.*]] = alloca i32, align 4
24497 // CHECK20-NEXT: [[DOTOFFLOAD_BASEPTRS47:%.*]] = alloca [4 x i8*], align 4
24498 // CHECK20-NEXT: [[DOTOFFLOAD_PTRS48:%.*]] = alloca [4 x i8*], align 4
24499 // CHECK20-NEXT: [[DOTOFFLOAD_MAPPERS49:%.*]] = alloca [4 x i8*], align 4
24500 // CHECK20-NEXT: [[DOTOFFLOAD_SIZES50:%.*]] = alloca [4 x i64], align 4
24501 // CHECK20-NEXT: [[_TMP51:%.*]] = alloca i32, align 4
24502 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_52:%.*]] = alloca i32, align 4
24503 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_53:%.*]] = alloca i32, align 4
24504 // CHECK20-NEXT: store i32 0, i32* [[RETVAL]], align 4
24505 // CHECK20-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
24506 // CHECK20-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
24507 // CHECK20-NEXT: store i32 100, i32* [[N]], align 4
24508 // CHECK20-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
24509 // CHECK20-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
24510 // CHECK20-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
24511 // CHECK20-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
24512 // CHECK20-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
24513 // CHECK20-NEXT: store i32 10, i32* [[M]], align 4
24514 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
24515 // CHECK20-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
24516 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
24517 // CHECK20-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
24518 // CHECK20-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
24519 // CHECK20-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
24520 // CHECK20-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
24521 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[TMP7]], align 4
24522 // CHECK20-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
24523 // CHECK20-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
24524 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[TMP9]], align 4
24525 // CHECK20-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
24526 // CHECK20-NEXT: store i64 4, i64* [[TMP10]], align 4
24527 // CHECK20-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
24528 // CHECK20-NEXT: store i8* null, i8** [[TMP11]], align 4
24529 // CHECK20-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
24530 // CHECK20-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
24531 // CHECK20-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
24532 // CHECK20-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
24533 // CHECK20-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
24534 // CHECK20-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
24535 // CHECK20-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
24536 // CHECK20-NEXT: store i64 4, i64* [[TMP16]], align 4
24537 // CHECK20-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
24538 // CHECK20-NEXT: store i8* null, i8** [[TMP17]], align 4
24539 // CHECK20-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
24540 // CHECK20-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
24541 // CHECK20-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 4
24542 // CHECK20-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
24543 // CHECK20-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
24544 // CHECK20-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 4
24545 // CHECK20-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
24546 // CHECK20-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 4
24547 // CHECK20-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
24548 // CHECK20-NEXT: store i8* null, i8** [[TMP23]], align 4
24549 // CHECK20-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
24550 // CHECK20-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
24551 // CHECK20-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
24552 // CHECK20-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
24553 // CHECK20-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
24554 // CHECK20-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
24555 // CHECK20-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
24556 // CHECK20-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
24557 // CHECK20-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
24558 // CHECK20-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
24559 // CHECK20-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
24560 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
24561 // CHECK20-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
24562 // CHECK20-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
24563 // CHECK20-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148.region_id, i32 3, i8** [[TMP24]], i8** [[TMP25]], i64* [[TMP26]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
24564 // CHECK20-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
24565 // CHECK20-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
24566 // CHECK20: omp_offload.failed:
24567 // CHECK20-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
24568 // CHECK20-NEXT: br label [[OMP_OFFLOAD_CONT]]
24569 // CHECK20: omp_offload.cont:
24570 // CHECK20-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
24571 // CHECK20-NEXT: store i32 [[TMP33]], i32* [[N_CASTED3]], align 4
24572 // CHECK20-NEXT: [[TMP34:%.*]] = load i32, i32* [[N_CASTED3]], align 4
24573 // CHECK20-NEXT: [[TMP35:%.*]] = mul nuw i32 [[TMP0]], 4
24574 // CHECK20-NEXT: [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
24575 // CHECK20-NEXT: [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
24576 // CHECK20-NEXT: [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
24577 // CHECK20-NEXT: store i32 [[TMP34]], i32* [[TMP38]], align 4
24578 // CHECK20-NEXT: [[TMP39:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
24579 // CHECK20-NEXT: [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
24580 // CHECK20-NEXT: store i32 [[TMP34]], i32* [[TMP40]], align 4
24581 // CHECK20-NEXT: [[TMP41:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
24582 // CHECK20-NEXT: store i64 4, i64* [[TMP41]], align 4
24583 // CHECK20-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
24584 // CHECK20-NEXT: store i8* null, i8** [[TMP42]], align 4
24585 // CHECK20-NEXT: [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
24586 // CHECK20-NEXT: [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i32*
24587 // CHECK20-NEXT: store i32 [[TMP0]], i32* [[TMP44]], align 4
24588 // CHECK20-NEXT: [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
24589 // CHECK20-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
24590 // CHECK20-NEXT: store i32 [[TMP0]], i32* [[TMP46]], align 4
24591 // CHECK20-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 1
24592 // CHECK20-NEXT: store i64 4, i64* [[TMP47]], align 4
24593 // CHECK20-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
24594 // CHECK20-NEXT: store i8* null, i8** [[TMP48]], align 4
24595 // CHECK20-NEXT: [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
24596 // CHECK20-NEXT: [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32**
24597 // CHECK20-NEXT: store i32* [[VLA]], i32** [[TMP50]], align 4
24598 // CHECK20-NEXT: [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
24599 // CHECK20-NEXT: [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32**
24600 // CHECK20-NEXT: store i32* [[VLA]], i32** [[TMP52]], align 4
24601 // CHECK20-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
24602 // CHECK20-NEXT: store i64 [[TMP36]], i64* [[TMP53]], align 4
24603 // CHECK20-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
24604 // CHECK20-NEXT: store i8* null, i8** [[TMP54]], align 4
24605 // CHECK20-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
24606 // CHECK20-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
24607 // CHECK20-NEXT: [[TMP57:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
24608 // CHECK20-NEXT: [[TMP58:%.*]] = load i32, i32* [[N]], align 4
24609 // CHECK20-NEXT: store i32 [[TMP58]], i32* [[DOTCAPTURE_EXPR_9]], align 4
24610 // CHECK20-NEXT: [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
24611 // CHECK20-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP59]], 0
24612 // CHECK20-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
24613 // CHECK20-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
24614 // CHECK20-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
24615 // CHECK20-NEXT: [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
24616 // CHECK20-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP60]], 1
24617 // CHECK20-NEXT: [[TMP61:%.*]] = zext i32 [[ADD14]] to i64
24618 // CHECK20-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP61]])
24619 // CHECK20-NEXT: [[TMP62:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153.region_id, i32 3, i8** [[TMP55]], i8** [[TMP56]], i64* [[TMP57]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.4, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
24620 // CHECK20-NEXT: [[TMP63:%.*]] = icmp ne i32 [[TMP62]], 0
24621 // CHECK20-NEXT: br i1 [[TMP63]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
24622 // CHECK20: omp_offload.failed15:
24623 // CHECK20-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153(i32 [[TMP34]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
24624 // CHECK20-NEXT: br label [[OMP_OFFLOAD_CONT16]]
24625 // CHECK20: omp_offload.cont16:
24626 // CHECK20-NEXT: [[TMP64:%.*]] = load i32, i32* [[M]], align 4
24627 // CHECK20-NEXT: store i32 [[TMP64]], i32* [[M_CASTED]], align 4
24628 // CHECK20-NEXT: [[TMP65:%.*]] = load i32, i32* [[M_CASTED]], align 4
24629 // CHECK20-NEXT: [[TMP66:%.*]] = load i32, i32* [[N]], align 4
24630 // CHECK20-NEXT: store i32 [[TMP66]], i32* [[N_CASTED17]], align 4
24631 // CHECK20-NEXT: [[TMP67:%.*]] = load i32, i32* [[N_CASTED17]], align 4
24632 // CHECK20-NEXT: [[TMP68:%.*]] = mul nuw i32 [[TMP0]], 4
24633 // CHECK20-NEXT: [[TMP69:%.*]] = sext i32 [[TMP68]] to i64
24634 // CHECK20-NEXT: [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
24635 // CHECK20-NEXT: [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i32*
24636 // CHECK20-NEXT: store i32 [[TMP65]], i32* [[TMP71]], align 4
24637 // CHECK20-NEXT: [[TMP72:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
24638 // CHECK20-NEXT: [[TMP73:%.*]] = bitcast i8** [[TMP72]] to i32*
24639 // CHECK20-NEXT: store i32 [[TMP65]], i32* [[TMP73]], align 4
24640 // CHECK20-NEXT: [[TMP74:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
24641 // CHECK20-NEXT: store i64 4, i64* [[TMP74]], align 4
24642 // CHECK20-NEXT: [[TMP75:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 0
24643 // CHECK20-NEXT: store i8* null, i8** [[TMP75]], align 4
24644 // CHECK20-NEXT: [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 1
24645 // CHECK20-NEXT: [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i32*
24646 // CHECK20-NEXT: store i32 [[TMP67]], i32* [[TMP77]], align 4
24647 // CHECK20-NEXT: [[TMP78:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 1
24648 // CHECK20-NEXT: [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
24649 // CHECK20-NEXT: store i32 [[TMP67]], i32* [[TMP79]], align 4
24650 // CHECK20-NEXT: [[TMP80:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 1
24651 // CHECK20-NEXT: store i64 4, i64* [[TMP80]], align 4
24652 // CHECK20-NEXT: [[TMP81:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 1
24653 // CHECK20-NEXT: store i8* null, i8** [[TMP81]], align 4
24654 // CHECK20-NEXT: [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 2
24655 // CHECK20-NEXT: [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i32*
24656 // CHECK20-NEXT: store i32 [[TMP0]], i32* [[TMP83]], align 4
24657 // CHECK20-NEXT: [[TMP84:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 2
24658 // CHECK20-NEXT: [[TMP85:%.*]] = bitcast i8** [[TMP84]] to i32*
24659 // CHECK20-NEXT: store i32 [[TMP0]], i32* [[TMP85]], align 4
24660 // CHECK20-NEXT: [[TMP86:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 2
24661 // CHECK20-NEXT: store i64 4, i64* [[TMP86]], align 4
24662 // CHECK20-NEXT: [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 2
24663 // CHECK20-NEXT: store i8* null, i8** [[TMP87]], align 4
24664 // CHECK20-NEXT: [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 3
24665 // CHECK20-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
24666 // CHECK20-NEXT: store i32* [[VLA]], i32** [[TMP89]], align 4
24667 // CHECK20-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 3
24668 // CHECK20-NEXT: [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i32**
24669 // CHECK20-NEXT: store i32* [[VLA]], i32** [[TMP91]], align 4
24670 // CHECK20-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 3
24671 // CHECK20-NEXT: store i64 [[TMP69]], i64* [[TMP92]], align 4
24672 // CHECK20-NEXT: [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 3
24673 // CHECK20-NEXT: store i8* null, i8** [[TMP93]], align 4
24674 // CHECK20-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
24675 // CHECK20-NEXT: [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
24676 // CHECK20-NEXT: [[TMP96:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
24677 // CHECK20-NEXT: [[TMP97:%.*]] = load i32, i32* [[N]], align 4
24678 // CHECK20-NEXT: store i32 [[TMP97]], i32* [[DOTCAPTURE_EXPR_23]], align 4
24679 // CHECK20-NEXT: [[TMP98:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_23]], align 4
24680 // CHECK20-NEXT: [[SUB25:%.*]] = sub nsw i32 [[TMP98]], 0
24681 // CHECK20-NEXT: [[DIV26:%.*]] = sdiv i32 [[SUB25]], 1
24682 // CHECK20-NEXT: [[SUB27:%.*]] = sub nsw i32 [[DIV26]], 1
24683 // CHECK20-NEXT: store i32 [[SUB27]], i32* [[DOTCAPTURE_EXPR_24]], align 4
24684 // CHECK20-NEXT: [[TMP99:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
24685 // CHECK20-NEXT: [[ADD28:%.*]] = add nsw i32 [[TMP99]], 1
24686 // CHECK20-NEXT: [[TMP100:%.*]] = zext i32 [[ADD28]] to i64
24687 // CHECK20-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP100]])
24688 // CHECK20-NEXT: [[TMP101:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158.region_id, i32 4, i8** [[TMP94]], i8** [[TMP95]], i64* [[TMP96]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.7, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
24689 // CHECK20-NEXT: [[TMP102:%.*]] = icmp ne i32 [[TMP101]], 0
24690 // CHECK20-NEXT: br i1 [[TMP102]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
24691 // CHECK20: omp_offload.failed29:
24692 // CHECK20-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158(i32 [[TMP65]], i32 [[TMP67]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
24693 // CHECK20-NEXT: br label [[OMP_OFFLOAD_CONT30]]
24694 // CHECK20: omp_offload.cont30:
24695 // CHECK20-NEXT: [[TMP103:%.*]] = load i32, i32* [[N]], align 4
24696 // CHECK20-NEXT: store i32 [[TMP103]], i32* [[N_CASTED31]], align 4
24697 // CHECK20-NEXT: [[TMP104:%.*]] = load i32, i32* [[N_CASTED31]], align 4
24698 // CHECK20-NEXT: [[TMP105:%.*]] = mul nuw i32 [[TMP0]], 4
24699 // CHECK20-NEXT: [[TMP106:%.*]] = sext i32 [[TMP105]] to i64
24700 // CHECK20-NEXT: [[TMP107:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
24701 // CHECK20-NEXT: [[TMP108:%.*]] = bitcast i8** [[TMP107]] to i32*
24702 // CHECK20-NEXT: store i32 [[TMP104]], i32* [[TMP108]], align 4
24703 // CHECK20-NEXT: [[TMP109:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
24704 // CHECK20-NEXT: [[TMP110:%.*]] = bitcast i8** [[TMP109]] to i32*
24705 // CHECK20-NEXT: store i32 [[TMP104]], i32* [[TMP110]], align 4
24706 // CHECK20-NEXT: [[TMP111:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
24707 // CHECK20-NEXT: store i64 4, i64* [[TMP111]], align 4
24708 // CHECK20-NEXT: [[TMP112:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 0
24709 // CHECK20-NEXT: store i8* null, i8** [[TMP112]], align 4
24710 // CHECK20-NEXT: [[TMP113:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 1
24711 // CHECK20-NEXT: [[TMP114:%.*]] = bitcast i8** [[TMP113]] to i32*
24712 // CHECK20-NEXT: store i32 [[TMP0]], i32* [[TMP114]], align 4
24713 // CHECK20-NEXT: [[TMP115:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 1
24714 // CHECK20-NEXT: [[TMP116:%.*]] = bitcast i8** [[TMP115]] to i32*
24715 // CHECK20-NEXT: store i32 [[TMP0]], i32* [[TMP116]], align 4
24716 // CHECK20-NEXT: [[TMP117:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 1
24717 // CHECK20-NEXT: store i64 4, i64* [[TMP117]], align 4
24718 // CHECK20-NEXT: [[TMP118:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 1
24719 // CHECK20-NEXT: store i8* null, i8** [[TMP118]], align 4
24720 // CHECK20-NEXT: [[TMP119:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 2
24721 // CHECK20-NEXT: [[TMP120:%.*]] = bitcast i8** [[TMP119]] to i32**
24722 // CHECK20-NEXT: store i32* [[VLA]], i32** [[TMP120]], align 4
24723 // CHECK20-NEXT: [[TMP121:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 2
24724 // CHECK20-NEXT: [[TMP122:%.*]] = bitcast i8** [[TMP121]] to i32**
24725 // CHECK20-NEXT: store i32* [[VLA]], i32** [[TMP122]], align 4
24726 // CHECK20-NEXT: [[TMP123:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 2
24727 // CHECK20-NEXT: store i64 [[TMP106]], i64* [[TMP123]], align 4
24728 // CHECK20-NEXT: [[TMP124:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS34]], i32 0, i32 2
24729 // CHECK20-NEXT: store i8* null, i8** [[TMP124]], align 4
24730 // CHECK20-NEXT: [[TMP125:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS32]], i32 0, i32 0
24731 // CHECK20-NEXT: [[TMP126:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS33]], i32 0, i32 0
24732 // CHECK20-NEXT: [[TMP127:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES35]], i32 0, i32 0
24733 // CHECK20-NEXT: [[TMP128:%.*]] = load i32, i32* [[N]], align 4
24734 // CHECK20-NEXT: store i32 [[TMP128]], i32* [[DOTCAPTURE_EXPR_37]], align 4
24735 // CHECK20-NEXT: [[TMP129:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_37]], align 4
24736 // CHECK20-NEXT: [[SUB39:%.*]] = sub nsw i32 [[TMP129]], 0
24737 // CHECK20-NEXT: [[DIV40:%.*]] = sdiv i32 [[SUB39]], 1
24738 // CHECK20-NEXT: [[SUB41:%.*]] = sub nsw i32 [[DIV40]], 1
24739 // CHECK20-NEXT: store i32 [[SUB41]], i32* [[DOTCAPTURE_EXPR_38]], align 4
24740 // CHECK20-NEXT: [[TMP130:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_38]], align 4
24741 // CHECK20-NEXT: [[ADD42:%.*]] = add nsw i32 [[TMP130]], 1
24742 // CHECK20-NEXT: [[TMP131:%.*]] = zext i32 [[ADD42]] to i64
24743 // CHECK20-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP131]])
24744 // CHECK20-NEXT: [[TMP132:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163.region_id, i32 3, i8** [[TMP125]], i8** [[TMP126]], i64* [[TMP127]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
24745 // CHECK20-NEXT: [[TMP133:%.*]] = icmp ne i32 [[TMP132]], 0
24746 // CHECK20-NEXT: br i1 [[TMP133]], label [[OMP_OFFLOAD_FAILED43:%.*]], label [[OMP_OFFLOAD_CONT44:%.*]]
24747 // CHECK20: omp_offload.failed43:
24748 // CHECK20-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163(i32 [[TMP104]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
24749 // CHECK20-NEXT: br label [[OMP_OFFLOAD_CONT44]]
24750 // CHECK20: omp_offload.cont44:
24751 // CHECK20-NEXT: [[TMP134:%.*]] = load i32, i32* [[M]], align 4
24752 // CHECK20-NEXT: store i32 [[TMP134]], i32* [[M_CASTED45]], align 4
24753 // CHECK20-NEXT: [[TMP135:%.*]] = load i32, i32* [[M_CASTED45]], align 4
24754 // CHECK20-NEXT: [[TMP136:%.*]] = load i32, i32* [[N]], align 4
24755 // CHECK20-NEXT: store i32 [[TMP136]], i32* [[N_CASTED46]], align 4
24756 // CHECK20-NEXT: [[TMP137:%.*]] = load i32, i32* [[N_CASTED46]], align 4
24757 // CHECK20-NEXT: [[TMP138:%.*]] = mul nuw i32 [[TMP0]], 4
24758 // CHECK20-NEXT: [[TMP139:%.*]] = sext i32 [[TMP138]] to i64
24759 // CHECK20-NEXT: [[TMP140:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
24760 // CHECK20-NEXT: [[TMP141:%.*]] = bitcast i8** [[TMP140]] to i32*
24761 // CHECK20-NEXT: store i32 [[TMP135]], i32* [[TMP141]], align 4
24762 // CHECK20-NEXT: [[TMP142:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
24763 // CHECK20-NEXT: [[TMP143:%.*]] = bitcast i8** [[TMP142]] to i32*
24764 // CHECK20-NEXT: store i32 [[TMP135]], i32* [[TMP143]], align 4
24765 // CHECK20-NEXT: [[TMP144:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
24766 // CHECK20-NEXT: store i64 4, i64* [[TMP144]], align 4
24767 // CHECK20-NEXT: [[TMP145:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 0
24768 // CHECK20-NEXT: store i8* null, i8** [[TMP145]], align 4
24769 // CHECK20-NEXT: [[TMP146:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 1
24770 // CHECK20-NEXT: [[TMP147:%.*]] = bitcast i8** [[TMP146]] to i32*
24771 // CHECK20-NEXT: store i32 [[TMP137]], i32* [[TMP147]], align 4
24772 // CHECK20-NEXT: [[TMP148:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 1
24773 // CHECK20-NEXT: [[TMP149:%.*]] = bitcast i8** [[TMP148]] to i32*
24774 // CHECK20-NEXT: store i32 [[TMP137]], i32* [[TMP149]], align 4
24775 // CHECK20-NEXT: [[TMP150:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 1
24776 // CHECK20-NEXT: store i64 4, i64* [[TMP150]], align 4
24777 // CHECK20-NEXT: [[TMP151:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 1
24778 // CHECK20-NEXT: store i8* null, i8** [[TMP151]], align 4
24779 // CHECK20-NEXT: [[TMP152:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 2
24780 // CHECK20-NEXT: [[TMP153:%.*]] = bitcast i8** [[TMP152]] to i32*
24781 // CHECK20-NEXT: store i32 [[TMP0]], i32* [[TMP153]], align 4
24782 // CHECK20-NEXT: [[TMP154:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 2
24783 // CHECK20-NEXT: [[TMP155:%.*]] = bitcast i8** [[TMP154]] to i32*
24784 // CHECK20-NEXT: store i32 [[TMP0]], i32* [[TMP155]], align 4
24785 // CHECK20-NEXT: [[TMP156:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 2
24786 // CHECK20-NEXT: store i64 4, i64* [[TMP156]], align 4
24787 // CHECK20-NEXT: [[TMP157:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 2
24788 // CHECK20-NEXT: store i8* null, i8** [[TMP157]], align 4
24789 // CHECK20-NEXT: [[TMP158:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 3
24790 // CHECK20-NEXT: [[TMP159:%.*]] = bitcast i8** [[TMP158]] to i32**
24791 // CHECK20-NEXT: store i32* [[VLA]], i32** [[TMP159]], align 4
24792 // CHECK20-NEXT: [[TMP160:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 3
24793 // CHECK20-NEXT: [[TMP161:%.*]] = bitcast i8** [[TMP160]] to i32**
24794 // CHECK20-NEXT: store i32* [[VLA]], i32** [[TMP161]], align 4
24795 // CHECK20-NEXT: [[TMP162:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 3
24796 // CHECK20-NEXT: store i64 [[TMP139]], i64* [[TMP162]], align 4
24797 // CHECK20-NEXT: [[TMP163:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS49]], i32 0, i32 3
24798 // CHECK20-NEXT: store i8* null, i8** [[TMP163]], align 4
24799 // CHECK20-NEXT: [[TMP164:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS47]], i32 0, i32 0
24800 // CHECK20-NEXT: [[TMP165:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS48]], i32 0, i32 0
24801 // CHECK20-NEXT: [[TMP166:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES50]], i32 0, i32 0
24802 // CHECK20-NEXT: [[TMP167:%.*]] = load i32, i32* [[N]], align 4
24803 // CHECK20-NEXT: store i32 [[TMP167]], i32* [[DOTCAPTURE_EXPR_52]], align 4
24804 // CHECK20-NEXT: [[TMP168:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_52]], align 4
24805 // CHECK20-NEXT: [[SUB54:%.*]] = sub nsw i32 [[TMP168]], 0
24806 // CHECK20-NEXT: [[DIV55:%.*]] = sdiv i32 [[SUB54]], 1
24807 // CHECK20-NEXT: [[SUB56:%.*]] = sub nsw i32 [[DIV55]], 1
24808 // CHECK20-NEXT: store i32 [[SUB56]], i32* [[DOTCAPTURE_EXPR_53]], align 4
24809 // CHECK20-NEXT: [[TMP169:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_53]], align 4
24810 // CHECK20-NEXT: [[ADD57:%.*]] = add nsw i32 [[TMP169]], 1
24811 // CHECK20-NEXT: [[TMP170:%.*]] = zext i32 [[ADD57]] to i64
24812 // CHECK20-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP170]])
24813 // CHECK20-NEXT: [[TMP171:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168.region_id, i32 4, i8** [[TMP164]], i8** [[TMP165]], i64* [[TMP166]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.13, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
24814 // CHECK20-NEXT: [[TMP172:%.*]] = icmp ne i32 [[TMP171]], 0
24815 // CHECK20-NEXT: br i1 [[TMP172]], label [[OMP_OFFLOAD_FAILED58:%.*]], label [[OMP_OFFLOAD_CONT59:%.*]]
24816 // CHECK20: omp_offload.failed58:
24817 // CHECK20-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168(i32 [[TMP135]], i32 [[TMP137]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
24818 // CHECK20-NEXT: br label [[OMP_OFFLOAD_CONT59]]
24819 // CHECK20: omp_offload.cont59:
24820 // CHECK20-NEXT: [[TMP173:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
24821 // CHECK20-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP173]])
24822 // CHECK20-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
24823 // CHECK20-NEXT: [[TMP174:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
24824 // CHECK20-NEXT: call void @llvm.stackrestore(i8* [[TMP174]])
24825 // CHECK20-NEXT: [[TMP175:%.*]] = load i32, i32* [[RETVAL]], align 4
24826 // CHECK20-NEXT: ret i32 [[TMP175]]
24827 //
24828 //
24829 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l148
24830 // CHECK20-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
24831 // CHECK20-NEXT: entry:
24832 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
24833 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
24834 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
24835 // CHECK20-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
24836 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
24837 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
24838 // CHECK20-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
24839 // CHECK20-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
24840 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
24841 // CHECK20-NEXT: ret void
24842 //
24843 //
24844 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined.
24845 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
24846 // CHECK20-NEXT: entry:
24847 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
24848 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
24849 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
24850 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
24851 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
24852 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
24853 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
24854 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
24855 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
24856 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
24857 // CHECK20-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
24858 // CHECK20-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
24859 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
24860 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
24861 // CHECK20-NEXT: [[I3:%.*]] = alloca i32, align 4
24862 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
24863 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
24864 // CHECK20-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
24865 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
24866 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
24867 // CHECK20-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
24868 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
24869 // CHECK20-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
24870 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
24871 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
24872 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
24873 // CHECK20-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
24874 // CHECK20-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
24875 // CHECK20-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
24876 // CHECK20-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
24877 // CHECK20-NEXT: store i32 0, i32* [[I]], align 4
24878 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
24879 // CHECK20-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
24880 // CHECK20-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
24881 // CHECK20: omp.precond.then:
24882 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
24883 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
24884 // CHECK20-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
24885 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
24886 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
24887 // CHECK20-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
24888 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
24889 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
24890 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24891 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
24892 // CHECK20-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
24893 // CHECK20-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
24894 // CHECK20: cond.true:
24895 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
24896 // CHECK20-NEXT: br label [[COND_END:%.*]]
24897 // CHECK20: cond.false:
24898 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24899 // CHECK20-NEXT: br label [[COND_END]]
24900 // CHECK20: cond.end:
24901 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
24902 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
24903 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
24904 // CHECK20-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
24905 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
24906 // CHECK20: omp.inner.for.cond:
24907 // CHECK20-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
24908 // CHECK20-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24909 // CHECK20-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
24910 // CHECK20-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
24911 // CHECK20: omp.inner.for.body:
24912 // CHECK20-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
24913 // CHECK20-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
24914 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
24915 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
24916 // CHECK20: omp.inner.for.inc:
24917 // CHECK20-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
24918 // CHECK20-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
24919 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
24920 // CHECK20-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
24921 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
24922 // CHECK20: omp.inner.for.end:
24923 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
24924 // CHECK20: omp.loop.exit:
24925 // CHECK20-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
24926 // CHECK20-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
24927 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
24928 // CHECK20-NEXT: br label [[OMP_PRECOND_END]]
24929 // CHECK20: omp.precond.end:
24930 // CHECK20-NEXT: ret void
24931 //
24932 //
24933 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..1
24934 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
24935 // CHECK20-NEXT: entry:
24936 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
24937 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
24938 // CHECK20-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
24939 // CHECK20-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
24940 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
24941 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
24942 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
24943 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
24944 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
24945 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
24946 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
24947 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
24948 // CHECK20-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
24949 // CHECK20-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
24950 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
24951 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
24952 // CHECK20-NEXT: [[I3:%.*]] = alloca i32, align 4
24953 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
24954 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
24955 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
24956 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
24957 // CHECK20-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
24958 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
24959 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
24960 // CHECK20-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
24961 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
24962 // CHECK20-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
24963 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
24964 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
24965 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
24966 // CHECK20-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
24967 // CHECK20-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
24968 // CHECK20-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
24969 // CHECK20-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
24970 // CHECK20-NEXT: store i32 0, i32* [[I]], align 4
24971 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
24972 // CHECK20-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
24973 // CHECK20-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
24974 // CHECK20: omp.precond.then:
24975 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
24976 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
24977 // CHECK20-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
24978 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
24979 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
24980 // CHECK20-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
24981 // CHECK20-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
24982 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
24983 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
24984 // CHECK20-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
24985 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
24986 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
24987 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
24988 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
24989 // CHECK20-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
24990 // CHECK20-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
24991 // CHECK20: cond.true:
24992 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
24993 // CHECK20-NEXT: br label [[COND_END:%.*]]
24994 // CHECK20: cond.false:
24995 // CHECK20-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
24996 // CHECK20-NEXT: br label [[COND_END]]
24997 // CHECK20: cond.end:
24998 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
24999 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
25000 // CHECK20-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
25001 // CHECK20-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
25002 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
25003 // CHECK20: omp.inner.for.cond:
25004 // CHECK20-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25005 // CHECK20-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
25006 // CHECK20-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
25007 // CHECK20-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
25008 // CHECK20: omp.inner.for.body:
25009 // CHECK20-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25010 // CHECK20-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
25011 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
25012 // CHECK20-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
25013 // CHECK20-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
25014 // CHECK20-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
25015 // CHECK20-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
25016 // CHECK20-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
25017 // CHECK20: omp.body.continue:
25018 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
25019 // CHECK20: omp.inner.for.inc:
25020 // CHECK20-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25021 // CHECK20-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
25022 // CHECK20-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
25023 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
25024 // CHECK20: omp.inner.for.end:
25025 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
25026 // CHECK20: omp.loop.exit:
25027 // CHECK20-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25028 // CHECK20-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
25029 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
25030 // CHECK20-NEXT: br label [[OMP_PRECOND_END]]
25031 // CHECK20: omp.precond.end:
25032 // CHECK20-NEXT: ret void
25033 //
25034 //
25035 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l153
25036 // CHECK20-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
25037 // CHECK20-NEXT: entry:
25038 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
25039 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
25040 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
25041 // CHECK20-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
25042 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
25043 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
25044 // CHECK20-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
25045 // CHECK20-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
25046 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
25047 // CHECK20-NEXT: ret void
25048 //
25049 //
25050 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..2
25051 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
25052 // CHECK20-NEXT: entry:
25053 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
25054 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
25055 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
25056 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
25057 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
25058 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
25059 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
25060 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
25061 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
25062 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
25063 // CHECK20-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
25064 // CHECK20-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
25065 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
25066 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
25067 // CHECK20-NEXT: [[I3:%.*]] = alloca i32, align 4
25068 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
25069 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
25070 // CHECK20-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
25071 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
25072 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
25073 // CHECK20-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
25074 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
25075 // CHECK20-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
25076 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
25077 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
25078 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
25079 // CHECK20-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
25080 // CHECK20-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
25081 // CHECK20-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
25082 // CHECK20-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
25083 // CHECK20-NEXT: store i32 0, i32* [[I]], align 4
25084 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
25085 // CHECK20-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
25086 // CHECK20-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
25087 // CHECK20: omp.precond.then:
25088 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
25089 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25090 // CHECK20-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
25091 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
25092 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
25093 // CHECK20-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25094 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
25095 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
25096 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25097 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25098 // CHECK20-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
25099 // CHECK20-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
25100 // CHECK20: cond.true:
25101 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25102 // CHECK20-NEXT: br label [[COND_END:%.*]]
25103 // CHECK20: cond.false:
25104 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25105 // CHECK20-NEXT: br label [[COND_END]]
25106 // CHECK20: cond.end:
25107 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
25108 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
25109 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
25110 // CHECK20-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
25111 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
25112 // CHECK20: omp.inner.for.cond:
25113 // CHECK20-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25114 // CHECK20-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25115 // CHECK20-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
25116 // CHECK20-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
25117 // CHECK20: omp.inner.for.body:
25118 // CHECK20-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
25119 // CHECK20-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25120 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
25121 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
25122 // CHECK20: omp.inner.for.inc:
25123 // CHECK20-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25124 // CHECK20-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
25125 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
25126 // CHECK20-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
25127 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
25128 // CHECK20: omp.inner.for.end:
25129 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
25130 // CHECK20: omp.loop.exit:
25131 // CHECK20-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25132 // CHECK20-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
25133 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
25134 // CHECK20-NEXT: br label [[OMP_PRECOND_END]]
25135 // CHECK20: omp.precond.end:
25136 // CHECK20-NEXT: ret void
25137 //
25138 //
25139 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..3
25140 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
25141 // CHECK20-NEXT: entry:
25142 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
25143 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
25144 // CHECK20-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
25145 // CHECK20-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
25146 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
25147 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
25148 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
25149 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
25150 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
25151 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
25152 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
25153 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
25154 // CHECK20-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
25155 // CHECK20-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
25156 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
25157 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
25158 // CHECK20-NEXT: [[I3:%.*]] = alloca i32, align 4
25159 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
25160 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
25161 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
25162 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
25163 // CHECK20-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
25164 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
25165 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
25166 // CHECK20-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
25167 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
25168 // CHECK20-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
25169 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
25170 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
25171 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
25172 // CHECK20-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
25173 // CHECK20-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
25174 // CHECK20-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
25175 // CHECK20-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
25176 // CHECK20-NEXT: store i32 0, i32* [[I]], align 4
25177 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
25178 // CHECK20-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
25179 // CHECK20-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
25180 // CHECK20: omp.precond.then:
25181 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
25182 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25183 // CHECK20-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
25184 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
25185 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
25186 // CHECK20-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
25187 // CHECK20-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
25188 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
25189 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
25190 // CHECK20-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25191 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
25192 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
25193 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
25194 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25195 // CHECK20-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
25196 // CHECK20-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
25197 // CHECK20: cond.true:
25198 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25199 // CHECK20-NEXT: br label [[COND_END:%.*]]
25200 // CHECK20: cond.false:
25201 // CHECK20-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
25202 // CHECK20-NEXT: br label [[COND_END]]
25203 // CHECK20: cond.end:
25204 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
25205 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
25206 // CHECK20-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
25207 // CHECK20-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
25208 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
25209 // CHECK20: omp.inner.for.cond:
25210 // CHECK20-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25211 // CHECK20-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
25212 // CHECK20-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
25213 // CHECK20-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
25214 // CHECK20: omp.inner.for.body:
25215 // CHECK20-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25216 // CHECK20-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
25217 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
25218 // CHECK20-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
25219 // CHECK20-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
25220 // CHECK20-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
25221 // CHECK20-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
25222 // CHECK20-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
25223 // CHECK20: omp.body.continue:
25224 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
25225 // CHECK20: omp.inner.for.inc:
25226 // CHECK20-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25227 // CHECK20-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
25228 // CHECK20-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
25229 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
25230 // CHECK20: omp.inner.for.end:
25231 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
25232 // CHECK20: omp.loop.exit:
25233 // CHECK20-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25234 // CHECK20-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
25235 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
25236 // CHECK20-NEXT: br label [[OMP_PRECOND_END]]
25237 // CHECK20: omp.precond.end:
25238 // CHECK20-NEXT: ret void
25239 //
25240 //
25241 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l158
25242 // CHECK20-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
25243 // CHECK20-NEXT: entry:
25244 // CHECK20-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
25245 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
25246 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
25247 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
25248 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
25249 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
25250 // CHECK20-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
25251 // CHECK20-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
25252 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
25253 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
25254 // CHECK20-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
25255 // CHECK20-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
25256 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
25257 // CHECK20-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
25258 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
25259 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
25260 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
25261 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
25262 // CHECK20-NEXT: ret void
25263 //
25264 //
25265 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..5
25266 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
25267 // CHECK20-NEXT: entry:
25268 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
25269 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
25270 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
25271 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
25272 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
25273 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
25274 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
25275 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
25276 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
25277 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
25278 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
25279 // CHECK20-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
25280 // CHECK20-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
25281 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
25282 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
25283 // CHECK20-NEXT: [[I4:%.*]] = alloca i32, align 4
25284 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
25285 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
25286 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
25287 // CHECK20-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
25288 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
25289 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
25290 // CHECK20-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
25291 // CHECK20-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
25292 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
25293 // CHECK20-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
25294 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
25295 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
25296 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25297 // CHECK20-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
25298 // CHECK20-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
25299 // CHECK20-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
25300 // CHECK20-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
25301 // CHECK20-NEXT: store i32 0, i32* [[I]], align 4
25302 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25303 // CHECK20-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
25304 // CHECK20-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
25305 // CHECK20: omp.precond.then:
25306 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
25307 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25308 // CHECK20-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
25309 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
25310 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
25311 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
25312 // CHECK20-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25313 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
25314 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
25315 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25316 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25317 // CHECK20-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
25318 // CHECK20-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
25319 // CHECK20: cond.true:
25320 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25321 // CHECK20-NEXT: br label [[COND_END:%.*]]
25322 // CHECK20: cond.false:
25323 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25324 // CHECK20-NEXT: br label [[COND_END]]
25325 // CHECK20: cond.end:
25326 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
25327 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
25328 // CHECK20-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
25329 // CHECK20-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
25330 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
25331 // CHECK20: omp.inner.for.cond:
25332 // CHECK20-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25333 // CHECK20-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25334 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
25335 // CHECK20-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
25336 // CHECK20-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
25337 // CHECK20: omp.inner.for.body:
25338 // CHECK20-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
25339 // CHECK20-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25340 // CHECK20-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
25341 // CHECK20-NEXT: store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
25342 // CHECK20-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
25343 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]])
25344 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
25345 // CHECK20: omp.inner.for.inc:
25346 // CHECK20-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25347 // CHECK20-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
25348 // CHECK20-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
25349 // CHECK20-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
25350 // CHECK20-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
25351 // CHECK20-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
25352 // CHECK20-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
25353 // CHECK20-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4
25354 // CHECK20-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25355 // CHECK20-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
25356 // CHECK20-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
25357 // CHECK20-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4
25358 // CHECK20-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25359 // CHECK20-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25360 // CHECK20-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
25361 // CHECK20-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
25362 // CHECK20: cond.true11:
25363 // CHECK20-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25364 // CHECK20-NEXT: br label [[COND_END13:%.*]]
25365 // CHECK20: cond.false12:
25366 // CHECK20-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25367 // CHECK20-NEXT: br label [[COND_END13]]
25368 // CHECK20: cond.end13:
25369 // CHECK20-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
25370 // CHECK20-NEXT: store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4
25371 // CHECK20-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
25372 // CHECK20-NEXT: store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
25373 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
25374 // CHECK20: omp.inner.for.end:
25375 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
25376 // CHECK20: omp.loop.exit:
25377 // CHECK20-NEXT: [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25378 // CHECK20-NEXT: [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
25379 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
25380 // CHECK20-NEXT: br label [[OMP_PRECOND_END]]
25381 // CHECK20: omp.precond.end:
25382 // CHECK20-NEXT: ret void
25383 //
25384 //
25385 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..6
25386 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
25387 // CHECK20-NEXT: entry:
25388 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
25389 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
25390 // CHECK20-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
25391 // CHECK20-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
25392 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
25393 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
25394 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
25395 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
25396 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
25397 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
25398 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
25399 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
25400 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
25401 // CHECK20-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
25402 // CHECK20-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
25403 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
25404 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
25405 // CHECK20-NEXT: [[I4:%.*]] = alloca i32, align 4
25406 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
25407 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
25408 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
25409 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
25410 // CHECK20-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
25411 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
25412 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
25413 // CHECK20-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
25414 // CHECK20-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
25415 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
25416 // CHECK20-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
25417 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
25418 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
25419 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25420 // CHECK20-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
25421 // CHECK20-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
25422 // CHECK20-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
25423 // CHECK20-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
25424 // CHECK20-NEXT: store i32 0, i32* [[I]], align 4
25425 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25426 // CHECK20-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
25427 // CHECK20-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
25428 // CHECK20: omp.precond.then:
25429 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
25430 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25431 // CHECK20-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
25432 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
25433 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
25434 // CHECK20-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
25435 // CHECK20-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
25436 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
25437 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
25438 // CHECK20-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25439 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
25440 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
25441 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
25442 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25443 // CHECK20-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
25444 // CHECK20-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
25445 // CHECK20: cond.true:
25446 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25447 // CHECK20-NEXT: br label [[COND_END:%.*]]
25448 // CHECK20: cond.false:
25449 // CHECK20-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
25450 // CHECK20-NEXT: br label [[COND_END]]
25451 // CHECK20: cond.end:
25452 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
25453 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
25454 // CHECK20-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
25455 // CHECK20-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
25456 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
25457 // CHECK20: omp.inner.for.cond:
25458 // CHECK20-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25459 // CHECK20-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
25460 // CHECK20-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
25461 // CHECK20-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
25462 // CHECK20: omp.inner.for.body:
25463 // CHECK20-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25464 // CHECK20-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
25465 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
25466 // CHECK20-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
25467 // CHECK20-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
25468 // CHECK20-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
25469 // CHECK20-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
25470 // CHECK20-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
25471 // CHECK20: omp.body.continue:
25472 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
25473 // CHECK20: omp.inner.for.inc:
25474 // CHECK20-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25475 // CHECK20-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
25476 // CHECK20-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
25477 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
25478 // CHECK20: omp.inner.for.end:
25479 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
25480 // CHECK20: omp.loop.exit:
25481 // CHECK20-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25482 // CHECK20-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
25483 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
25484 // CHECK20-NEXT: br label [[OMP_PRECOND_END]]
25485 // CHECK20: omp.precond.end:
25486 // CHECK20-NEXT: ret void
25487 //
25488 //
25489 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l163
25490 // CHECK20-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
25491 // CHECK20-NEXT: entry:
25492 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
25493 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
25494 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
25495 // CHECK20-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
25496 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
25497 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
25498 // CHECK20-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
25499 // CHECK20-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
25500 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
25501 // CHECK20-NEXT: ret void
25502 //
25503 //
25504 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..8
25505 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
25506 // CHECK20-NEXT: entry:
25507 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
25508 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
25509 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
25510 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
25511 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
25512 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
25513 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
25514 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
25515 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
25516 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
25517 // CHECK20-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
25518 // CHECK20-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
25519 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
25520 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
25521 // CHECK20-NEXT: [[I3:%.*]] = alloca i32, align 4
25522 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
25523 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
25524 // CHECK20-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
25525 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
25526 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
25527 // CHECK20-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
25528 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
25529 // CHECK20-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
25530 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
25531 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
25532 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
25533 // CHECK20-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
25534 // CHECK20-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
25535 // CHECK20-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
25536 // CHECK20-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
25537 // CHECK20-NEXT: store i32 0, i32* [[I]], align 4
25538 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
25539 // CHECK20-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
25540 // CHECK20-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
25541 // CHECK20: omp.precond.then:
25542 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
25543 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25544 // CHECK20-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
25545 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
25546 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
25547 // CHECK20-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25548 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
25549 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
25550 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25551 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25552 // CHECK20-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
25553 // CHECK20-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
25554 // CHECK20: cond.true:
25555 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25556 // CHECK20-NEXT: br label [[COND_END:%.*]]
25557 // CHECK20: cond.false:
25558 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25559 // CHECK20-NEXT: br label [[COND_END]]
25560 // CHECK20: cond.end:
25561 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
25562 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
25563 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
25564 // CHECK20-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
25565 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
25566 // CHECK20: omp.inner.for.cond:
25567 // CHECK20-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25568 // CHECK20-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25569 // CHECK20-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
25570 // CHECK20-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
25571 // CHECK20: omp.inner.for.body:
25572 // CHECK20-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
25573 // CHECK20-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25574 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
25575 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
25576 // CHECK20: omp.inner.for.inc:
25577 // CHECK20-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25578 // CHECK20-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
25579 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
25580 // CHECK20-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
25581 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
25582 // CHECK20: omp.inner.for.end:
25583 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
25584 // CHECK20: omp.loop.exit:
25585 // CHECK20-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25586 // CHECK20-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
25587 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
25588 // CHECK20-NEXT: br label [[OMP_PRECOND_END]]
25589 // CHECK20: omp.precond.end:
25590 // CHECK20-NEXT: ret void
25591 //
25592 //
25593 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..9
25594 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
25595 // CHECK20-NEXT: entry:
25596 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
25597 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
25598 // CHECK20-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
25599 // CHECK20-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
25600 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
25601 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
25602 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
25603 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
25604 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
25605 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
25606 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
25607 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
25608 // CHECK20-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
25609 // CHECK20-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
25610 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
25611 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
25612 // CHECK20-NEXT: [[I3:%.*]] = alloca i32, align 4
25613 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
25614 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
25615 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
25616 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
25617 // CHECK20-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
25618 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
25619 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
25620 // CHECK20-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
25621 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
25622 // CHECK20-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
25623 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
25624 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
25625 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
25626 // CHECK20-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
25627 // CHECK20-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
25628 // CHECK20-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
25629 // CHECK20-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
25630 // CHECK20-NEXT: store i32 0, i32* [[I]], align 4
25631 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
25632 // CHECK20-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
25633 // CHECK20-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
25634 // CHECK20: omp.precond.then:
25635 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
25636 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25637 // CHECK20-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
25638 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
25639 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
25640 // CHECK20-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
25641 // CHECK20-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
25642 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
25643 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
25644 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
25645 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
25646 // CHECK20-NEXT: [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25647 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[TMP11]], align 4
25648 // CHECK20-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP12]], i32 35, i32 [[TMP9]], i32 [[TMP10]], i32 1, i32 1)
25649 // CHECK20-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
25650 // CHECK20: omp.dispatch.cond:
25651 // CHECK20-NEXT: [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25652 // CHECK20-NEXT: [[TMP14:%.*]] = load i32, i32* [[TMP13]], align 4
25653 // CHECK20-NEXT: [[TMP15:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP14]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
25654 // CHECK20-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP15]], 0
25655 // CHECK20-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
25656 // CHECK20: omp.dispatch.body:
25657 // CHECK20-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
25658 // CHECK20-NEXT: store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
25659 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
25660 // CHECK20: omp.inner.for.cond:
25661 // CHECK20-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
25662 // CHECK20-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !16
25663 // CHECK20-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
25664 // CHECK20-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
25665 // CHECK20: omp.inner.for.body:
25666 // CHECK20-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
25667 // CHECK20-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
25668 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
25669 // CHECK20-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !16
25670 // CHECK20-NEXT: [[TMP20:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !16
25671 // CHECK20-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
25672 // CHECK20-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !16
25673 // CHECK20-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
25674 // CHECK20: omp.body.continue:
25675 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
25676 // CHECK20: omp.inner.for.inc:
25677 // CHECK20-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
25678 // CHECK20-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP21]], 1
25679 // CHECK20-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !16
25680 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
25681 // CHECK20: omp.inner.for.end:
25682 // CHECK20-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
25683 // CHECK20: omp.dispatch.inc:
25684 // CHECK20-NEXT: br label [[OMP_DISPATCH_COND]]
25685 // CHECK20: omp.dispatch.end:
25686 // CHECK20-NEXT: br label [[OMP_PRECOND_END]]
25687 // CHECK20: omp.precond.end:
25688 // CHECK20-NEXT: ret void
25689 //
25690 //
25691 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l168
25692 // CHECK20-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
25693 // CHECK20-NEXT: entry:
25694 // CHECK20-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
25695 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
25696 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
25697 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
25698 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
25699 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
25700 // CHECK20-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
25701 // CHECK20-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
25702 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
25703 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
25704 // CHECK20-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
25705 // CHECK20-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
25706 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
25707 // CHECK20-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
25708 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
25709 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
25710 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
25711 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
25712 // CHECK20-NEXT: ret void
25713 //
25714 //
25715 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..11
25716 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
25717 // CHECK20-NEXT: entry:
25718 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
25719 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
25720 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
25721 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
25722 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
25723 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
25724 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
25725 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
25726 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
25727 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
25728 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
25729 // CHECK20-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
25730 // CHECK20-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
25731 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
25732 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
25733 // CHECK20-NEXT: [[I4:%.*]] = alloca i32, align 4
25734 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
25735 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
25736 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
25737 // CHECK20-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
25738 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
25739 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
25740 // CHECK20-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
25741 // CHECK20-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
25742 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
25743 // CHECK20-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
25744 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
25745 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
25746 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25747 // CHECK20-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
25748 // CHECK20-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
25749 // CHECK20-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
25750 // CHECK20-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
25751 // CHECK20-NEXT: store i32 0, i32* [[I]], align 4
25752 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25753 // CHECK20-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
25754 // CHECK20-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
25755 // CHECK20: omp.precond.then:
25756 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
25757 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25758 // CHECK20-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
25759 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
25760 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
25761 // CHECK20-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25762 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
25763 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
25764 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25765 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25766 // CHECK20-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
25767 // CHECK20-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
25768 // CHECK20: cond.true:
25769 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25770 // CHECK20-NEXT: br label [[COND_END:%.*]]
25771 // CHECK20: cond.false:
25772 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25773 // CHECK20-NEXT: br label [[COND_END]]
25774 // CHECK20: cond.end:
25775 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
25776 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
25777 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
25778 // CHECK20-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
25779 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
25780 // CHECK20: omp.inner.for.cond:
25781 // CHECK20-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25782 // CHECK20-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25783 // CHECK20-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
25784 // CHECK20-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
25785 // CHECK20: omp.inner.for.body:
25786 // CHECK20-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
25787 // CHECK20-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
25788 // CHECK20-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
25789 // CHECK20-NEXT: store i32 [[TMP18]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
25790 // CHECK20-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
25791 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP19]])
25792 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
25793 // CHECK20: omp.inner.for.inc:
25794 // CHECK20-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
25795 // CHECK20-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
25796 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
25797 // CHECK20-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
25798 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
25799 // CHECK20: omp.inner.for.end:
25800 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
25801 // CHECK20: omp.loop.exit:
25802 // CHECK20-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25803 // CHECK20-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
25804 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
25805 // CHECK20-NEXT: br label [[OMP_PRECOND_END]]
25806 // CHECK20: omp.precond.end:
25807 // CHECK20-NEXT: ret void
25808 //
25809 //
25810 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..12
25811 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
25812 // CHECK20-NEXT: entry:
25813 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
25814 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
25815 // CHECK20-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
25816 // CHECK20-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
25817 // CHECK20-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
25818 // CHECK20-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
25819 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
25820 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
25821 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
25822 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
25823 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
25824 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
25825 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
25826 // CHECK20-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
25827 // CHECK20-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
25828 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
25829 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
25830 // CHECK20-NEXT: [[I4:%.*]] = alloca i32, align 4
25831 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
25832 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
25833 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
25834 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
25835 // CHECK20-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
25836 // CHECK20-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
25837 // CHECK20-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
25838 // CHECK20-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
25839 // CHECK20-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
25840 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
25841 // CHECK20-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
25842 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
25843 // CHECK20-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
25844 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25845 // CHECK20-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
25846 // CHECK20-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
25847 // CHECK20-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
25848 // CHECK20-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
25849 // CHECK20-NEXT: store i32 0, i32* [[I]], align 4
25850 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
25851 // CHECK20-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
25852 // CHECK20-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
25853 // CHECK20: omp.precond.then:
25854 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
25855 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
25856 // CHECK20-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
25857 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
25858 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
25859 // CHECK20-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
25860 // CHECK20-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
25861 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
25862 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
25863 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
25864 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
25865 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
25866 // CHECK20-NEXT: [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25867 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
25868 // CHECK20-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP13]], i32 35, i32 [[TMP10]], i32 [[TMP11]], i32 1, i32 [[TMP9]])
25869 // CHECK20-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
25870 // CHECK20: omp.dispatch.cond:
25871 // CHECK20-NEXT: [[TMP14:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
25872 // CHECK20-NEXT: [[TMP15:%.*]] = load i32, i32* [[TMP14]], align 4
25873 // CHECK20-NEXT: [[TMP16:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP15]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
25874 // CHECK20-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP16]], 0
25875 // CHECK20-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
25876 // CHECK20: omp.dispatch.body:
25877 // CHECK20-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
25878 // CHECK20-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
25879 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
25880 // CHECK20: omp.inner.for.cond:
25881 // CHECK20-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
25882 // CHECK20-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
25883 // CHECK20-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
25884 // CHECK20-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
25885 // CHECK20: omp.inner.for.body:
25886 // CHECK20-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
25887 // CHECK20-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP20]], 1
25888 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
25889 // CHECK20-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !19
25890 // CHECK20-NEXT: [[TMP21:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !19
25891 // CHECK20-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP21]]
25892 // CHECK20-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !19
25893 // CHECK20-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
25894 // CHECK20: omp.body.continue:
25895 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
25896 // CHECK20: omp.inner.for.inc:
25897 // CHECK20-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
25898 // CHECK20-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP22]], 1
25899 // CHECK20-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
25900 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
25901 // CHECK20: omp.inner.for.end:
25902 // CHECK20-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
25903 // CHECK20: omp.dispatch.inc:
25904 // CHECK20-NEXT: br label [[OMP_DISPATCH_COND]]
25905 // CHECK20: omp.dispatch.end:
25906 // CHECK20-NEXT: br label [[OMP_PRECOND_END]]
25907 // CHECK20: omp.precond.end:
25908 // CHECK20-NEXT: ret void
25909 //
25910 //
25911 // CHECK20-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
25912 // CHECK20-SAME: (i32 [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
25913 // CHECK20-NEXT: entry:
25914 // CHECK20-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
25915 // CHECK20-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
25916 // CHECK20-NEXT: [[M:%.*]] = alloca i32, align 4
25917 // CHECK20-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
25918 // CHECK20-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
25919 // CHECK20-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
25920 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
25921 // CHECK20-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
25922 // CHECK20-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
25923 // CHECK20-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
25924 // CHECK20-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
25925 // CHECK20-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
25926 // CHECK20-NEXT: [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 4
25927 // CHECK20-NEXT: [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 4
25928 // CHECK20-NEXT: [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 4
25929 // CHECK20-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
25930 // CHECK20-NEXT: [[DOTOFFLOAD_BASEPTRS13:%.*]] = alloca [1 x i8*], align 4
25931 // CHECK20-NEXT: [[DOTOFFLOAD_PTRS14:%.*]] = alloca [1 x i8*], align 4
25932 // CHECK20-NEXT: [[DOTOFFLOAD_MAPPERS15:%.*]] = alloca [1 x i8*], align 4
25933 // CHECK20-NEXT: [[_TMP16:%.*]] = alloca i32, align 4
25934 // CHECK20-NEXT: [[M_CASTED19:%.*]] = alloca i32, align 4
25935 // CHECK20-NEXT: [[DOTOFFLOAD_BASEPTRS20:%.*]] = alloca [2 x i8*], align 4
25936 // CHECK20-NEXT: [[DOTOFFLOAD_PTRS21:%.*]] = alloca [2 x i8*], align 4
25937 // CHECK20-NEXT: [[DOTOFFLOAD_MAPPERS22:%.*]] = alloca [2 x i8*], align 4
25938 // CHECK20-NEXT: [[_TMP23:%.*]] = alloca i32, align 4
25939 // CHECK20-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
25940 // CHECK20-NEXT: store i32 10, i32* [[M]], align 4
25941 // CHECK20-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
25942 // CHECK20-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
25943 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
25944 // CHECK20-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
25945 // CHECK20-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
25946 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
25947 // CHECK20-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
25948 // CHECK20-NEXT: store i8* null, i8** [[TMP4]], align 4
25949 // CHECK20-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
25950 // CHECK20-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
25951 // CHECK20-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
25952 // CHECK20-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116.region_id, i32 1, i8** [[TMP5]], i8** [[TMP6]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.16, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
25953 // CHECK20-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
25954 // CHECK20-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
25955 // CHECK20: omp_offload.failed:
25956 // CHECK20-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116([10 x i32]* [[A]]) #[[ATTR3]]
25957 // CHECK20-NEXT: br label [[OMP_OFFLOAD_CONT]]
25958 // CHECK20: omp_offload.cont:
25959 // CHECK20-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
25960 // CHECK20-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
25961 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 4
25962 // CHECK20-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
25963 // CHECK20-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
25964 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 4
25965 // CHECK20-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
25966 // CHECK20-NEXT: store i8* null, i8** [[TMP13]], align 4
25967 // CHECK20-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
25968 // CHECK20-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
25969 // CHECK20-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
25970 // CHECK20-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.19, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.20, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
25971 // CHECK20-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
25972 // CHECK20-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
25973 // CHECK20: omp_offload.failed5:
25974 // CHECK20-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121([10 x i32]* [[A]]) #[[ATTR3]]
25975 // CHECK20-NEXT: br label [[OMP_OFFLOAD_CONT6]]
25976 // CHECK20: omp_offload.cont6:
25977 // CHECK20-NEXT: [[TMP18:%.*]] = load i32, i32* [[M]], align 4
25978 // CHECK20-NEXT: store i32 [[TMP18]], i32* [[M_CASTED]], align 4
25979 // CHECK20-NEXT: [[TMP19:%.*]] = load i32, i32* [[M_CASTED]], align 4
25980 // CHECK20-NEXT: [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
25981 // CHECK20-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
25982 // CHECK20-NEXT: store i32 [[TMP19]], i32* [[TMP21]], align 4
25983 // CHECK20-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
25984 // CHECK20-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
25985 // CHECK20-NEXT: store i32 [[TMP19]], i32* [[TMP23]], align 4
25986 // CHECK20-NEXT: [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 0
25987 // CHECK20-NEXT: store i8* null, i8** [[TMP24]], align 4
25988 // CHECK20-NEXT: [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
25989 // CHECK20-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
25990 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 4
25991 // CHECK20-NEXT: [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
25992 // CHECK20-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
25993 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 4
25994 // CHECK20-NEXT: [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 1
25995 // CHECK20-NEXT: store i8* null, i8** [[TMP29]], align 4
25996 // CHECK20-NEXT: [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
25997 // CHECK20-NEXT: [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
25998 // CHECK20-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
25999 // CHECK20-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.23, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.24, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
26000 // CHECK20-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
26001 // CHECK20-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
26002 // CHECK20: omp_offload.failed11:
26003 // CHECK20-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126(i32 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
26004 // CHECK20-NEXT: br label [[OMP_OFFLOAD_CONT12]]
26005 // CHECK20: omp_offload.cont12:
26006 // CHECK20-NEXT: [[TMP34:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
26007 // CHECK20-NEXT: [[TMP35:%.*]] = bitcast i8** [[TMP34]] to [10 x i32]**
26008 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP35]], align 4
26009 // CHECK20-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
26010 // CHECK20-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
26011 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 4
26012 // CHECK20-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS15]], i32 0, i32 0
26013 // CHECK20-NEXT: store i8* null, i8** [[TMP38]], align 4
26014 // CHECK20-NEXT: [[TMP39:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS13]], i32 0, i32 0
26015 // CHECK20-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS14]], i32 0, i32 0
26016 // CHECK20-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
26017 // CHECK20-NEXT: [[TMP41:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131.region_id, i32 1, i8** [[TMP39]], i8** [[TMP40]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.27, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.28, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
26018 // CHECK20-NEXT: [[TMP42:%.*]] = icmp ne i32 [[TMP41]], 0
26019 // CHECK20-NEXT: br i1 [[TMP42]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
26020 // CHECK20: omp_offload.failed17:
26021 // CHECK20-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131([10 x i32]* [[A]]) #[[ATTR3]]
26022 // CHECK20-NEXT: br label [[OMP_OFFLOAD_CONT18]]
26023 // CHECK20: omp_offload.cont18:
26024 // CHECK20-NEXT: [[TMP43:%.*]] = load i32, i32* [[M]], align 4
26025 // CHECK20-NEXT: store i32 [[TMP43]], i32* [[M_CASTED19]], align 4
26026 // CHECK20-NEXT: [[TMP44:%.*]] = load i32, i32* [[M_CASTED19]], align 4
26027 // CHECK20-NEXT: [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
26028 // CHECK20-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
26029 // CHECK20-NEXT: store i32 [[TMP44]], i32* [[TMP46]], align 4
26030 // CHECK20-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
26031 // CHECK20-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i32*
26032 // CHECK20-NEXT: store i32 [[TMP44]], i32* [[TMP48]], align 4
26033 // CHECK20-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 0
26034 // CHECK20-NEXT: store i8* null, i8** [[TMP49]], align 4
26035 // CHECK20-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 1
26036 // CHECK20-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to [10 x i32]**
26037 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP51]], align 4
26038 // CHECK20-NEXT: [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 1
26039 // CHECK20-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to [10 x i32]**
26040 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP53]], align 4
26041 // CHECK20-NEXT: [[TMP54:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS22]], i32 0, i32 1
26042 // CHECK20-NEXT: store i8* null, i8** [[TMP54]], align 4
26043 // CHECK20-NEXT: [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS20]], i32 0, i32 0
26044 // CHECK20-NEXT: [[TMP56:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS21]], i32 0, i32 0
26045 // CHECK20-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
26046 // CHECK20-NEXT: [[TMP57:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136.region_id, i32 2, i8** [[TMP55]], i8** [[TMP56]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.31, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.32, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
26047 // CHECK20-NEXT: [[TMP58:%.*]] = icmp ne i32 [[TMP57]], 0
26048 // CHECK20-NEXT: br i1 [[TMP58]], label [[OMP_OFFLOAD_FAILED24:%.*]], label [[OMP_OFFLOAD_CONT25:%.*]]
26049 // CHECK20: omp_offload.failed24:
26050 // CHECK20-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136(i32 [[TMP44]], [10 x i32]* [[A]]) #[[ATTR3]]
26051 // CHECK20-NEXT: br label [[OMP_OFFLOAD_CONT25]]
26052 // CHECK20: omp_offload.cont25:
26053 // CHECK20-NEXT: ret i32 0
26054 //
26055 //
26056 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l116
26057 // CHECK20-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
26058 // CHECK20-NEXT: entry:
26059 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26060 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26061 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26062 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
26063 // CHECK20-NEXT: ret void
26064 //
26065 //
26066 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..14
26067 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
26068 // CHECK20-NEXT: entry:
26069 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
26070 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
26071 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26072 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
26073 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
26074 // CHECK20-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
26075 // CHECK20-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
26076 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
26077 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
26078 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
26079 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
26080 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
26081 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26082 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26083 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
26084 // CHECK20-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
26085 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
26086 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
26087 // CHECK20-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
26088 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
26089 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
26090 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26091 // CHECK20-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
26092 // CHECK20-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
26093 // CHECK20: cond.true:
26094 // CHECK20-NEXT: br label [[COND_END:%.*]]
26095 // CHECK20: cond.false:
26096 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26097 // CHECK20-NEXT: br label [[COND_END]]
26098 // CHECK20: cond.end:
26099 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
26100 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
26101 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
26102 // CHECK20-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
26103 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
26104 // CHECK20: omp.inner.for.cond:
26105 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26106 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26107 // CHECK20-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
26108 // CHECK20-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
26109 // CHECK20: omp.inner.for.body:
26110 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
26111 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26112 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
26113 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
26114 // CHECK20: omp.inner.for.inc:
26115 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26116 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
26117 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
26118 // CHECK20-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
26119 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
26120 // CHECK20: omp.inner.for.end:
26121 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
26122 // CHECK20: omp.loop.exit:
26123 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
26124 // CHECK20-NEXT: ret void
26125 //
26126 //
26127 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..15
26128 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
26129 // CHECK20-NEXT: entry:
26130 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
26131 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
26132 // CHECK20-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
26133 // CHECK20-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
26134 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26135 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
26136 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
26137 // CHECK20-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
26138 // CHECK20-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
26139 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
26140 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
26141 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
26142 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
26143 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
26144 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
26145 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
26146 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26147 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26148 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
26149 // CHECK20-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
26150 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
26151 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
26152 // CHECK20-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
26153 // CHECK20-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
26154 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
26155 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
26156 // CHECK20-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
26157 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
26158 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
26159 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26160 // CHECK20-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
26161 // CHECK20-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
26162 // CHECK20: cond.true:
26163 // CHECK20-NEXT: br label [[COND_END:%.*]]
26164 // CHECK20: cond.false:
26165 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26166 // CHECK20-NEXT: br label [[COND_END]]
26167 // CHECK20: cond.end:
26168 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
26169 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
26170 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
26171 // CHECK20-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
26172 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
26173 // CHECK20: omp.inner.for.cond:
26174 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26175 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26176 // CHECK20-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
26177 // CHECK20-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
26178 // CHECK20: omp.inner.for.body:
26179 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26180 // CHECK20-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
26181 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
26182 // CHECK20-NEXT: store i32 [[ADD]], i32* [[I]], align 4
26183 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
26184 // CHECK20-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
26185 // CHECK20-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
26186 // CHECK20-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
26187 // CHECK20: omp.body.continue:
26188 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
26189 // CHECK20: omp.inner.for.inc:
26190 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26191 // CHECK20-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
26192 // CHECK20-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
26193 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
26194 // CHECK20: omp.inner.for.end:
26195 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
26196 // CHECK20: omp.loop.exit:
26197 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
26198 // CHECK20-NEXT: ret void
26199 //
26200 //
26201 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l121
26202 // CHECK20-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
26203 // CHECK20-NEXT: entry:
26204 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26205 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26206 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26207 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..17 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
26208 // CHECK20-NEXT: ret void
26209 //
26210 //
26211 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..17
26212 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
26213 // CHECK20-NEXT: entry:
26214 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
26215 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
26216 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26217 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
26218 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
26219 // CHECK20-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
26220 // CHECK20-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
26221 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
26222 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
26223 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
26224 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
26225 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
26226 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26227 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26228 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
26229 // CHECK20-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
26230 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
26231 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
26232 // CHECK20-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
26233 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
26234 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
26235 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26236 // CHECK20-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
26237 // CHECK20-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
26238 // CHECK20: cond.true:
26239 // CHECK20-NEXT: br label [[COND_END:%.*]]
26240 // CHECK20: cond.false:
26241 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26242 // CHECK20-NEXT: br label [[COND_END]]
26243 // CHECK20: cond.end:
26244 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
26245 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
26246 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
26247 // CHECK20-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
26248 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
26249 // CHECK20: omp.inner.for.cond:
26250 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26251 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26252 // CHECK20-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
26253 // CHECK20-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
26254 // CHECK20: omp.inner.for.body:
26255 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
26256 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26257 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..18 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
26258 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
26259 // CHECK20: omp.inner.for.inc:
26260 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26261 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
26262 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
26263 // CHECK20-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
26264 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
26265 // CHECK20: omp.inner.for.end:
26266 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
26267 // CHECK20: omp.loop.exit:
26268 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
26269 // CHECK20-NEXT: ret void
26270 //
26271 //
26272 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..18
26273 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
26274 // CHECK20-NEXT: entry:
26275 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
26276 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
26277 // CHECK20-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
26278 // CHECK20-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
26279 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26280 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
26281 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
26282 // CHECK20-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
26283 // CHECK20-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
26284 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
26285 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
26286 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
26287 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
26288 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
26289 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
26290 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
26291 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26292 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26293 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
26294 // CHECK20-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
26295 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
26296 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
26297 // CHECK20-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
26298 // CHECK20-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
26299 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
26300 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
26301 // CHECK20-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
26302 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
26303 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
26304 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26305 // CHECK20-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
26306 // CHECK20-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
26307 // CHECK20: cond.true:
26308 // CHECK20-NEXT: br label [[COND_END:%.*]]
26309 // CHECK20: cond.false:
26310 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26311 // CHECK20-NEXT: br label [[COND_END]]
26312 // CHECK20: cond.end:
26313 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
26314 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
26315 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
26316 // CHECK20-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
26317 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
26318 // CHECK20: omp.inner.for.cond:
26319 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26320 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26321 // CHECK20-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
26322 // CHECK20-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
26323 // CHECK20: omp.inner.for.body:
26324 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26325 // CHECK20-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
26326 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
26327 // CHECK20-NEXT: store i32 [[ADD]], i32* [[I]], align 4
26328 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
26329 // CHECK20-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
26330 // CHECK20-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
26331 // CHECK20-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
26332 // CHECK20: omp.body.continue:
26333 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
26334 // CHECK20: omp.inner.for.inc:
26335 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26336 // CHECK20-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
26337 // CHECK20-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
26338 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
26339 // CHECK20: omp.inner.for.end:
26340 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
26341 // CHECK20: omp.loop.exit:
26342 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
26343 // CHECK20-NEXT: ret void
26344 //
26345 //
26346 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l126
26347 // CHECK20-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
26348 // CHECK20-NEXT: entry:
26349 // CHECK20-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
26350 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26351 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
26352 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
26353 // CHECK20-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
26354 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26355 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26356 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
26357 // CHECK20-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
26358 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
26359 // CHECK20-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
26360 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
26361 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..21 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
26362 // CHECK20-NEXT: ret void
26363 //
26364 //
26365 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..21
26366 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
26367 // CHECK20-NEXT: entry:
26368 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
26369 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
26370 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26371 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
26372 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
26373 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
26374 // CHECK20-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
26375 // CHECK20-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
26376 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
26377 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
26378 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
26379 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
26380 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
26381 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
26382 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26383 // CHECK20-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
26384 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26385 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
26386 // CHECK20-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
26387 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
26388 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
26389 // CHECK20-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
26390 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
26391 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
26392 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26393 // CHECK20-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
26394 // CHECK20-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
26395 // CHECK20: cond.true:
26396 // CHECK20-NEXT: br label [[COND_END:%.*]]
26397 // CHECK20: cond.false:
26398 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26399 // CHECK20-NEXT: br label [[COND_END]]
26400 // CHECK20: cond.end:
26401 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
26402 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
26403 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
26404 // CHECK20-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
26405 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
26406 // CHECK20: omp.inner.for.cond:
26407 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26408 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26409 // CHECK20-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
26410 // CHECK20-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
26411 // CHECK20: omp.inner.for.body:
26412 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
26413 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26414 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
26415 // CHECK20-NEXT: store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
26416 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
26417 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..22 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
26418 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
26419 // CHECK20: omp.inner.for.inc:
26420 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26421 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
26422 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
26423 // CHECK20-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
26424 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
26425 // CHECK20: omp.inner.for.end:
26426 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
26427 // CHECK20: omp.loop.exit:
26428 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
26429 // CHECK20-NEXT: ret void
26430 //
26431 //
26432 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..22
26433 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
26434 // CHECK20-NEXT: entry:
26435 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
26436 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
26437 // CHECK20-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
26438 // CHECK20-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
26439 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26440 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
26441 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
26442 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
26443 // CHECK20-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
26444 // CHECK20-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
26445 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
26446 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
26447 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
26448 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
26449 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
26450 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
26451 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
26452 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26453 // CHECK20-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
26454 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26455 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
26456 // CHECK20-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
26457 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
26458 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
26459 // CHECK20-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
26460 // CHECK20-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
26461 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
26462 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
26463 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
26464 // CHECK20-NEXT: [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
26465 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
26466 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP3]])
26467 // CHECK20-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
26468 // CHECK20: omp.dispatch.cond:
26469 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26470 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
26471 // CHECK20-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP6]], [[TMP7]]
26472 // CHECK20-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
26473 // CHECK20: cond.true:
26474 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
26475 // CHECK20-NEXT: br label [[COND_END:%.*]]
26476 // CHECK20: cond.false:
26477 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26478 // CHECK20-NEXT: br label [[COND_END]]
26479 // CHECK20: cond.end:
26480 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ [[TMP8]], [[COND_TRUE]] ], [ [[TMP9]], [[COND_FALSE]] ]
26481 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
26482 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
26483 // CHECK20-NEXT: store i32 [[TMP10]], i32* [[DOTOMP_IV]], align 4
26484 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26485 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26486 // CHECK20-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
26487 // CHECK20-NEXT: br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
26488 // CHECK20: omp.dispatch.body:
26489 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
26490 // CHECK20: omp.inner.for.cond:
26491 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26492 // CHECK20-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26493 // CHECK20-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
26494 // CHECK20-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
26495 // CHECK20: omp.inner.for.body:
26496 // CHECK20-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26497 // CHECK20-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
26498 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
26499 // CHECK20-NEXT: store i32 [[ADD]], i32* [[I]], align 4
26500 // CHECK20-NEXT: [[TMP16:%.*]] = load i32, i32* [[I]], align 4
26501 // CHECK20-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP16]]
26502 // CHECK20-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
26503 // CHECK20-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
26504 // CHECK20: omp.body.continue:
26505 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
26506 // CHECK20: omp.inner.for.inc:
26507 // CHECK20-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26508 // CHECK20-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
26509 // CHECK20-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
26510 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
26511 // CHECK20: omp.inner.for.end:
26512 // CHECK20-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
26513 // CHECK20: omp.dispatch.inc:
26514 // CHECK20-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
26515 // CHECK20-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
26516 // CHECK20-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
26517 // CHECK20-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
26518 // CHECK20-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26519 // CHECK20-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
26520 // CHECK20-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
26521 // CHECK20-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
26522 // CHECK20-NEXT: br label [[OMP_DISPATCH_COND]]
26523 // CHECK20: omp.dispatch.end:
26524 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
26525 // CHECK20-NEXT: ret void
26526 //
26527 //
26528 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l131
26529 // CHECK20-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
26530 // CHECK20-NEXT: entry:
26531 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26532 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26533 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26534 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..25 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
26535 // CHECK20-NEXT: ret void
26536 //
26537 //
26538 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..25
26539 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
26540 // CHECK20-NEXT: entry:
26541 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
26542 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
26543 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26544 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
26545 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
26546 // CHECK20-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
26547 // CHECK20-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
26548 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
26549 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
26550 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
26551 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
26552 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
26553 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26554 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26555 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
26556 // CHECK20-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
26557 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
26558 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
26559 // CHECK20-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
26560 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
26561 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
26562 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26563 // CHECK20-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
26564 // CHECK20-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
26565 // CHECK20: cond.true:
26566 // CHECK20-NEXT: br label [[COND_END:%.*]]
26567 // CHECK20: cond.false:
26568 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26569 // CHECK20-NEXT: br label [[COND_END]]
26570 // CHECK20: cond.end:
26571 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
26572 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
26573 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
26574 // CHECK20-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
26575 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
26576 // CHECK20: omp.inner.for.cond:
26577 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26578 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26579 // CHECK20-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
26580 // CHECK20-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
26581 // CHECK20: omp.inner.for.body:
26582 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
26583 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26584 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..26 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
26585 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
26586 // CHECK20: omp.inner.for.inc:
26587 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26588 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
26589 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
26590 // CHECK20-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
26591 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
26592 // CHECK20: omp.inner.for.end:
26593 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
26594 // CHECK20: omp.loop.exit:
26595 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
26596 // CHECK20-NEXT: ret void
26597 //
26598 //
26599 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..26
26600 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
26601 // CHECK20-NEXT: entry:
26602 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
26603 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
26604 // CHECK20-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
26605 // CHECK20-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
26606 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26607 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
26608 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
26609 // CHECK20-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
26610 // CHECK20-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
26611 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
26612 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
26613 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
26614 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
26615 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
26616 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
26617 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
26618 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26619 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26620 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
26621 // CHECK20-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
26622 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
26623 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
26624 // CHECK20-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
26625 // CHECK20-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
26626 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
26627 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
26628 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
26629 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26630 // CHECK20-NEXT: [[TMP5:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
26631 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 4
26632 // CHECK20-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32 35, i32 [[TMP3]], i32 [[TMP4]], i32 1, i32 1)
26633 // CHECK20-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
26634 // CHECK20: omp.dispatch.cond:
26635 // CHECK20-NEXT: [[TMP7:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP6]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
26636 // CHECK20-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP7]], 0
26637 // CHECK20-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
26638 // CHECK20: omp.dispatch.body:
26639 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
26640 // CHECK20-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
26641 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
26642 // CHECK20: omp.inner.for.cond:
26643 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
26644 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
26645 // CHECK20-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
26646 // CHECK20-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
26647 // CHECK20: omp.inner.for.body:
26648 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
26649 // CHECK20-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
26650 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
26651 // CHECK20-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
26652 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !22
26653 // CHECK20-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP12]]
26654 // CHECK20-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !22
26655 // CHECK20-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
26656 // CHECK20: omp.body.continue:
26657 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
26658 // CHECK20: omp.inner.for.inc:
26659 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
26660 // CHECK20-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP13]], 1
26661 // CHECK20-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
26662 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
26663 // CHECK20: omp.inner.for.end:
26664 // CHECK20-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
26665 // CHECK20: omp.dispatch.inc:
26666 // CHECK20-NEXT: br label [[OMP_DISPATCH_COND]]
26667 // CHECK20: omp.dispatch.end:
26668 // CHECK20-NEXT: ret void
26669 //
26670 //
26671 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l136
26672 // CHECK20-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
26673 // CHECK20-NEXT: entry:
26674 // CHECK20-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
26675 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26676 // CHECK20-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
26677 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
26678 // CHECK20-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
26679 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26680 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26681 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
26682 // CHECK20-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
26683 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
26684 // CHECK20-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
26685 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
26686 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..29 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
26687 // CHECK20-NEXT: ret void
26688 //
26689 //
26690 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..29
26691 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
26692 // CHECK20-NEXT: entry:
26693 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
26694 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
26695 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26696 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
26697 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
26698 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
26699 // CHECK20-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
26700 // CHECK20-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
26701 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
26702 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
26703 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
26704 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
26705 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
26706 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
26707 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26708 // CHECK20-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
26709 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26710 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
26711 // CHECK20-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
26712 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
26713 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
26714 // CHECK20-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
26715 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
26716 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
26717 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26718 // CHECK20-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
26719 // CHECK20-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
26720 // CHECK20: cond.true:
26721 // CHECK20-NEXT: br label [[COND_END:%.*]]
26722 // CHECK20: cond.false:
26723 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26724 // CHECK20-NEXT: br label [[COND_END]]
26725 // CHECK20: cond.end:
26726 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
26727 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
26728 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
26729 // CHECK20-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
26730 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
26731 // CHECK20: omp.inner.for.cond:
26732 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26733 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26734 // CHECK20-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
26735 // CHECK20-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
26736 // CHECK20: omp.inner.for.body:
26737 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
26738 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
26739 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
26740 // CHECK20-NEXT: store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
26741 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
26742 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..30 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
26743 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
26744 // CHECK20: omp.inner.for.inc:
26745 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
26746 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
26747 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
26748 // CHECK20-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
26749 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
26750 // CHECK20: omp.inner.for.end:
26751 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
26752 // CHECK20: omp.loop.exit:
26753 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
26754 // CHECK20-NEXT: ret void
26755 //
26756 //
26757 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined..30
26758 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
26759 // CHECK20-NEXT: entry:
26760 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
26761 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
26762 // CHECK20-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
26763 // CHECK20-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
26764 // CHECK20-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
26765 // CHECK20-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
26766 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
26767 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
26768 // CHECK20-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
26769 // CHECK20-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
26770 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
26771 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
26772 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
26773 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
26774 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
26775 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
26776 // CHECK20-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
26777 // CHECK20-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
26778 // CHECK20-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
26779 // CHECK20-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
26780 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
26781 // CHECK20-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
26782 // CHECK20-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
26783 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
26784 // CHECK20-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
26785 // CHECK20-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
26786 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
26787 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
26788 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
26789 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
26790 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
26791 // CHECK20-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
26792 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
26793 // CHECK20-NEXT: call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32 35, i32 [[TMP4]], i32 [[TMP5]], i32 1, i32 [[TMP3]])
26794 // CHECK20-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
26795 // CHECK20: omp.dispatch.cond:
26796 // CHECK20-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB3]], i32 [[TMP7]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
26797 // CHECK20-NEXT: [[TOBOOL:%.*]] = icmp ne i32 [[TMP8]], 0
26798 // CHECK20-NEXT: br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
26799 // CHECK20: omp.dispatch.body:
26800 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
26801 // CHECK20-NEXT: store i32 [[TMP9]], i32* [[DOTOMP_IV]], align 4
26802 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
26803 // CHECK20: omp.inner.for.cond:
26804 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
26805 // CHECK20-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !25
26806 // CHECK20-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP10]], [[TMP11]]
26807 // CHECK20-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
26808 // CHECK20: omp.inner.for.body:
26809 // CHECK20-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
26810 // CHECK20-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP12]], 1
26811 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
26812 // CHECK20-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !25
26813 // CHECK20-NEXT: [[TMP13:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !25
26814 // CHECK20-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
26815 // CHECK20-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !25
26816 // CHECK20-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
26817 // CHECK20: omp.body.continue:
26818 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
26819 // CHECK20: omp.inner.for.inc:
26820 // CHECK20-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
26821 // CHECK20-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP14]], 1
26822 // CHECK20-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !25
26823 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP26:![0-9]+]]
26824 // CHECK20: omp.inner.for.end:
26825 // CHECK20-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
26826 // CHECK20: omp.dispatch.inc:
26827 // CHECK20-NEXT: br label [[OMP_DISPATCH_COND]]
26828 // CHECK20: omp.dispatch.end:
26829 // CHECK20-NEXT: ret void
26830 //
26831 //
26832 // CHECK20-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
26833 // CHECK20-SAME: () #[[ATTR5:[0-9]+]] {
26834 // CHECK20-NEXT: entry:
26835 // CHECK20-NEXT: call void @__tgt_register_requires(i64 1)
26836 // CHECK20-NEXT: ret void
26837 //
26838