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-simd -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-simd -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-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 --check-prefix=CHECK6
17 // 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 --check-prefix=CHECK7
18 // 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
19 // 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 --check-prefix=CHECK8
20 #ifdef CK1
21
22 template <typename T, int X, long long Y>
23 struct SS{
24 T a[X];
25 float b;
fooSS26 int foo(void) {
27
28 #pragma omp target
29 #pragma omp teams distribute parallel for simd
30 for(int i = 0; i < X; i++) {
31 a[i] = (T)0;
32 }
33 #pragma omp target
34 #pragma omp teams distribute parallel for simd dist_schedule(static)
35 for(int i = 0; i < X; i++) {
36 a[i] = (T)0;
37 }
38 #pragma omp target
39 #pragma omp teams distribute parallel for simd dist_schedule(static, X/2)
40 for(int i = 0; i < X; i++) {
41 a[i] = (T)0;
42 }
43
44
45
46
47
48
49
50
51
52
53 return a[0];
54 }
55 };
56
teams_template_struct(void)57 int teams_template_struct(void) {
58 SS<int, 123, 456> V;
59 return V.foo();
60
61 }
62
63
64 #endif // CK1
65
66 // Test host codegen.
67 // 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=CHECK9
68 // 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
69 // 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=CHECK10
70 // 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=CHECK11
71 // 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
72 // 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=CHECK12
73
74 // 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 --check-prefix=CHECK13
75 // 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
76 // 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 --check-prefix=CHECK14
77 // 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 --check-prefix=CHECK15
78 // 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
79 // 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 --check-prefix=CHECK16
80 #ifdef CK2
81
82 template <typename T, int n>
tmain(T argc)83 int tmain(T argc) {
84 T a[n];
85 int m = 10;
86 #pragma omp target
87 #pragma omp teams distribute parallel for simd
88 for(int i = 0; i < n; i++) {
89 a[i] = (T)0;
90 }
91 #pragma omp target
92 #pragma omp teams distribute parallel for simd dist_schedule(static)
93 for(int i = 0; i < n; i++) {
94 a[i] = (T)0;
95 }
96 #pragma omp target
97 #pragma omp teams distribute parallel for simd dist_schedule(static, m)
98 for(int i = 0; i < n; i++) {
99 a[i] = (T)0;
100 }
101 return 0;
102 }
103
main(int argc,char ** argv)104 int main (int argc, char **argv) {
105 int n = 100;
106 int a[n];
107 int m = 10;
108 #pragma omp target
109 #pragma omp teams distribute parallel for simd
110 for(int i = 0; i < n; i++) {
111 a[i] = 0;
112 }
113 #pragma omp target
114 #pragma omp teams distribute parallel for simd dist_schedule(static)
115 for(int i = 0; i < n; i++) {
116 a[i] = 0;
117 }
118 #pragma omp target
119 #pragma omp teams distribute parallel for simd dist_schedule(static, m)
120 for(int i = 0; i < n; i++) {
121 a[i] = 0;
122 }
123 return tmain<int, 10>(argc);
124 }
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147 #endif // CK2
148 #endif // #ifndef HEADER
149 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
150 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
151 // CHECK1-NEXT: entry:
152 // CHECK1-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
153 // CHECK1-NEXT: [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
154 // CHECK1-NEXT: ret i32 [[CALL]]
155 //
156 //
157 // CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
158 // CHECK1-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
159 // CHECK1-NEXT: entry:
160 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
161 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
162 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
163 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
164 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
165 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
166 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
167 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
168 // CHECK1-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
169 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 8
170 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 8
171 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 8
172 // CHECK1-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
173 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
174 // CHECK1-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
175 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
176 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
177 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
178 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
179 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
180 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
181 // CHECK1-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
182 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
183 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
184 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
185 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
186 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
187 // 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_l28.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)
188 // CHECK1-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
189 // CHECK1-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
190 // CHECK1: omp_offload.failed:
191 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
192 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
193 // CHECK1: omp_offload.cont:
194 // CHECK1-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
195 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
196 // CHECK1-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
197 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 8
198 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
199 // CHECK1-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
200 // CHECK1-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 8
201 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
202 // CHECK1-NEXT: store i8* null, i8** [[TMP13]], align 8
203 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
204 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
205 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
206 // 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_l33.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)
207 // CHECK1-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
208 // CHECK1-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
209 // CHECK1: omp_offload.failed7:
210 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33(%struct.SS* [[THIS1]]) #[[ATTR2]]
211 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT8]]
212 // CHECK1: omp_offload.cont8:
213 // CHECK1-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
214 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
215 // CHECK1-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
216 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
217 // CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
218 // CHECK1-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
219 // CHECK1-NEXT: store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 8
220 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 0
221 // CHECK1-NEXT: store i8* null, i8** [[TMP22]], align 8
222 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
223 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
224 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
225 // 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_l38.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)
226 // CHECK1-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
227 // CHECK1-NEXT: br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
228 // CHECK1: omp_offload.failed14:
229 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38(%struct.SS* [[THIS1]]) #[[ATTR2]]
230 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT15]]
231 // CHECK1: omp_offload.cont15:
232 // CHECK1-NEXT: [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
233 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A16]], i64 0, i64 0
234 // CHECK1-NEXT: [[TMP27:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
235 // CHECK1-NEXT: ret i32 [[TMP27]]
236 //
237 //
238 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
239 // CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
240 // CHECK1-NEXT: entry:
241 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
242 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
243 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
244 // 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]])
245 // CHECK1-NEXT: ret void
246 //
247 //
248 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
249 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
250 // CHECK1-NEXT: entry:
251 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
252 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
253 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
254 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
255 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
256 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
257 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
258 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
259 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
260 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
261 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
262 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
263 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
264 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
265 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
266 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
267 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
268 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
269 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
270 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
271 // 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)
272 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
273 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
274 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
275 // CHECK1: cond.true:
276 // CHECK1-NEXT: br label [[COND_END:%.*]]
277 // CHECK1: cond.false:
278 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
279 // CHECK1-NEXT: br label [[COND_END]]
280 // CHECK1: cond.end:
281 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
282 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
283 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
284 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
285 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
286 // CHECK1: omp.inner.for.cond:
287 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
288 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
289 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
290 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
291 // CHECK1: omp.inner.for.body:
292 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
293 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
294 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
295 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
296 // 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]])
297 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
298 // CHECK1: omp.inner.for.inc:
299 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
300 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
301 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
302 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
303 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
304 // CHECK1: omp.inner.for.end:
305 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
306 // CHECK1: omp.loop.exit:
307 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
308 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
309 // CHECK1-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
310 // CHECK1-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
311 // CHECK1: .omp.final.then:
312 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
313 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
314 // CHECK1: .omp.final.done:
315 // CHECK1-NEXT: ret void
316 //
317 //
318 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
319 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
320 // CHECK1-NEXT: entry:
321 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
322 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
323 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
324 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
325 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
326 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
327 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
328 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
329 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
330 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
331 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
332 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
333 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
334 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
335 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
336 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
337 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
338 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
339 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
340 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
341 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
342 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
343 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
344 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
345 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
346 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
347 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
348 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
349 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
350 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
351 // 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)
352 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
353 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
354 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
355 // CHECK1: cond.true:
356 // CHECK1-NEXT: br label [[COND_END:%.*]]
357 // CHECK1: cond.false:
358 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
359 // CHECK1-NEXT: br label [[COND_END]]
360 // CHECK1: cond.end:
361 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
362 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
363 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
364 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
365 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
366 // CHECK1: omp.inner.for.cond:
367 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
368 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
369 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
370 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
371 // CHECK1: omp.inner.for.body:
372 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
373 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
374 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
375 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
376 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
377 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
378 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
379 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
380 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
381 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
382 // CHECK1: omp.body.continue:
383 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
384 // CHECK1: omp.inner.for.inc:
385 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
386 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
387 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
388 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
389 // CHECK1: omp.inner.for.end:
390 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
391 // CHECK1: omp.loop.exit:
392 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
393 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
394 // CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
395 // CHECK1-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
396 // CHECK1: .omp.final.then:
397 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
398 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
399 // CHECK1: .omp.final.done:
400 // CHECK1-NEXT: ret void
401 //
402 //
403 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33
404 // CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
405 // CHECK1-NEXT: entry:
406 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
407 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
408 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
409 // 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]])
410 // CHECK1-NEXT: ret void
411 //
412 //
413 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
414 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
415 // CHECK1-NEXT: entry:
416 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
417 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
418 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
419 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
420 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
421 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
422 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
423 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
424 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
425 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
426 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
427 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
428 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
429 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
430 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
431 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
432 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
433 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
434 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
435 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
436 // 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)
437 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
438 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
439 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
440 // CHECK1: cond.true:
441 // CHECK1-NEXT: br label [[COND_END:%.*]]
442 // CHECK1: cond.false:
443 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
444 // CHECK1-NEXT: br label [[COND_END]]
445 // CHECK1: cond.end:
446 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
447 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
448 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
449 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
450 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
451 // CHECK1: omp.inner.for.cond:
452 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
453 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
454 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
455 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
456 // CHECK1: omp.inner.for.body:
457 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
458 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
459 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
460 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
461 // 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]])
462 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
463 // CHECK1: omp.inner.for.inc:
464 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
465 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
466 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
467 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
468 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
469 // CHECK1: omp.inner.for.end:
470 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
471 // CHECK1: omp.loop.exit:
472 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
473 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
474 // CHECK1-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
475 // CHECK1-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
476 // CHECK1: .omp.final.then:
477 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
478 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
479 // CHECK1: .omp.final.done:
480 // CHECK1-NEXT: ret void
481 //
482 //
483 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
484 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
485 // CHECK1-NEXT: entry:
486 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
487 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
488 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
489 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
490 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
491 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
492 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
493 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
494 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
495 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
496 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
497 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
498 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
499 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
500 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
501 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
502 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
503 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
504 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
505 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
506 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
507 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
508 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
509 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
510 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
511 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
512 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
513 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
514 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
515 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
516 // 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)
517 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
518 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
519 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
520 // CHECK1: cond.true:
521 // CHECK1-NEXT: br label [[COND_END:%.*]]
522 // CHECK1: cond.false:
523 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
524 // CHECK1-NEXT: br label [[COND_END]]
525 // CHECK1: cond.end:
526 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
527 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
528 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
529 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
530 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
531 // CHECK1: omp.inner.for.cond:
532 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
533 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
534 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
535 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
536 // CHECK1: omp.inner.for.body:
537 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
538 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
539 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
540 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
541 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
542 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
543 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
544 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
545 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
546 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
547 // CHECK1: omp.body.continue:
548 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
549 // CHECK1: omp.inner.for.inc:
550 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
551 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
552 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
553 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
554 // CHECK1: omp.inner.for.end:
555 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
556 // CHECK1: omp.loop.exit:
557 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
558 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
559 // CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
560 // CHECK1-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
561 // CHECK1: .omp.final.then:
562 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
563 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
564 // CHECK1: .omp.final.done:
565 // CHECK1-NEXT: ret void
566 //
567 //
568 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38
569 // CHECK1-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
570 // CHECK1-NEXT: entry:
571 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
572 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
573 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
574 // 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]])
575 // CHECK1-NEXT: ret void
576 //
577 //
578 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
579 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
580 // CHECK1-NEXT: entry:
581 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
582 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
583 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
584 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
585 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
586 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
587 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
588 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
589 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
590 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
591 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
592 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
593 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
594 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
595 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
596 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
597 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
598 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
599 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
600 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
601 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
602 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
603 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
604 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
605 // CHECK1: cond.true:
606 // CHECK1-NEXT: br label [[COND_END:%.*]]
607 // CHECK1: cond.false:
608 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
609 // CHECK1-NEXT: br label [[COND_END]]
610 // CHECK1: cond.end:
611 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
612 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
613 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
614 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
615 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
616 // CHECK1: omp.inner.for.cond:
617 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
618 // CHECK1-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 123
619 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
620 // CHECK1: omp.inner.for.body:
621 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
622 // CHECK1-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
623 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
624 // CHECK1-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
625 // 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 [[TMP8]], i64 [[TMP10]], %struct.SS* [[TMP0]])
626 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
627 // CHECK1: omp.inner.for.inc:
628 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
629 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
630 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
631 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
632 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
633 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
634 // CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
635 // CHECK1-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4
636 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
637 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
638 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
639 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4
640 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
641 // CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP17]], 122
642 // CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
643 // CHECK1: cond.true5:
644 // CHECK1-NEXT: br label [[COND_END7:%.*]]
645 // CHECK1: cond.false6:
646 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
647 // CHECK1-NEXT: br label [[COND_END7]]
648 // CHECK1: cond.end7:
649 // CHECK1-NEXT: [[COND8:%.*]] = phi i32 [ 122, [[COND_TRUE5]] ], [ [[TMP18]], [[COND_FALSE6]] ]
650 // CHECK1-NEXT: store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4
651 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
652 // CHECK1-NEXT: store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4
653 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
654 // CHECK1: omp.inner.for.end:
655 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
656 // CHECK1: omp.loop.exit:
657 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
658 // CHECK1-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
659 // CHECK1-NEXT: [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
660 // CHECK1-NEXT: br i1 [[TMP21]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
661 // CHECK1: .omp.final.then:
662 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
663 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
664 // CHECK1: .omp.final.done:
665 // CHECK1-NEXT: ret void
666 //
667 //
668 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
669 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
670 // CHECK1-NEXT: entry:
671 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
672 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
673 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
674 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
675 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
676 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
677 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
678 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
679 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
680 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
681 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
682 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
683 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
684 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
685 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
686 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
687 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
688 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
689 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
690 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
691 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
692 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
693 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
694 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
695 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
696 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
697 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
698 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
699 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
700 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
701 // 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)
702 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
703 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
704 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
705 // CHECK1: cond.true:
706 // CHECK1-NEXT: br label [[COND_END:%.*]]
707 // CHECK1: cond.false:
708 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
709 // CHECK1-NEXT: br label [[COND_END]]
710 // CHECK1: cond.end:
711 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
712 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
713 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
714 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
715 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
716 // CHECK1: omp.inner.for.cond:
717 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
718 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
719 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
720 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
721 // CHECK1: omp.inner.for.body:
722 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
723 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
724 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
725 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
726 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
727 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
728 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
729 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
730 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
731 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
732 // CHECK1: omp.body.continue:
733 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
734 // CHECK1: omp.inner.for.inc:
735 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
736 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
737 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
738 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
739 // CHECK1: omp.inner.for.end:
740 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
741 // CHECK1: omp.loop.exit:
742 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
743 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
744 // CHECK1-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
745 // CHECK1-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
746 // CHECK1: .omp.final.then:
747 // CHECK1-NEXT: store i32 123, i32* [[I]], align 4
748 // CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
749 // CHECK1: .omp.final.done:
750 // CHECK1-NEXT: ret void
751 //
752 //
753 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
754 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
755 // CHECK1-NEXT: entry:
756 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
757 // CHECK1-NEXT: ret void
758 //
759 //
760 // CHECK2-LABEL: define {{[^@]+}}@_Z21teams_template_structv
761 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
762 // CHECK2-NEXT: entry:
763 // CHECK2-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
764 // CHECK2-NEXT: [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
765 // CHECK2-NEXT: ret i32 [[CALL]]
766 //
767 //
768 // CHECK2-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
769 // CHECK2-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
770 // CHECK2-NEXT: entry:
771 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
772 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
773 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
774 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
775 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
776 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
777 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
778 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
779 // CHECK2-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
780 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 8
781 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 8
782 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 8
783 // CHECK2-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
784 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
785 // CHECK2-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
786 // CHECK2-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
787 // CHECK2-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
788 // CHECK2-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
789 // CHECK2-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
790 // CHECK2-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
791 // CHECK2-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
792 // CHECK2-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
793 // CHECK2-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
794 // CHECK2-NEXT: store i8* null, i8** [[TMP4]], align 8
795 // CHECK2-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
796 // CHECK2-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
797 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
798 // 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_l28.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)
799 // CHECK2-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
800 // CHECK2-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
801 // CHECK2: omp_offload.failed:
802 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
803 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT]]
804 // CHECK2: omp_offload.cont:
805 // CHECK2-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
806 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
807 // CHECK2-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
808 // CHECK2-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 8
809 // CHECK2-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
810 // CHECK2-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
811 // CHECK2-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 8
812 // CHECK2-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
813 // CHECK2-NEXT: store i8* null, i8** [[TMP13]], align 8
814 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
815 // CHECK2-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
816 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
817 // 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_l33.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)
818 // CHECK2-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
819 // CHECK2-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
820 // CHECK2: omp_offload.failed7:
821 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33(%struct.SS* [[THIS1]]) #[[ATTR2]]
822 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT8]]
823 // CHECK2: omp_offload.cont8:
824 // CHECK2-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
825 // CHECK2-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
826 // CHECK2-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
827 // CHECK2-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
828 // CHECK2-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
829 // CHECK2-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
830 // CHECK2-NEXT: store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 8
831 // CHECK2-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i64 0, i64 0
832 // CHECK2-NEXT: store i8* null, i8** [[TMP22]], align 8
833 // CHECK2-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
834 // CHECK2-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
835 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
836 // 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_l38.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)
837 // CHECK2-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
838 // CHECK2-NEXT: br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
839 // CHECK2: omp_offload.failed14:
840 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38(%struct.SS* [[THIS1]]) #[[ATTR2]]
841 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT15]]
842 // CHECK2: omp_offload.cont15:
843 // CHECK2-NEXT: [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
844 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A16]], i64 0, i64 0
845 // CHECK2-NEXT: [[TMP27:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
846 // CHECK2-NEXT: ret i32 [[TMP27]]
847 //
848 //
849 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
850 // CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
851 // CHECK2-NEXT: entry:
852 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
853 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
854 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
855 // 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]])
856 // CHECK2-NEXT: ret void
857 //
858 //
859 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
860 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
861 // CHECK2-NEXT: entry:
862 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
863 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
864 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
865 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
866 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
867 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
868 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
869 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
870 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
871 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
872 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
873 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
874 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
875 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
876 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
877 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
878 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
879 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
880 // CHECK2-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
881 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
882 // 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)
883 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
884 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
885 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
886 // CHECK2: cond.true:
887 // CHECK2-NEXT: br label [[COND_END:%.*]]
888 // CHECK2: cond.false:
889 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
890 // CHECK2-NEXT: br label [[COND_END]]
891 // CHECK2: cond.end:
892 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
893 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
894 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
895 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
896 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
897 // CHECK2: omp.inner.for.cond:
898 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
899 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
900 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
901 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
902 // CHECK2: omp.inner.for.body:
903 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
904 // CHECK2-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
905 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
906 // CHECK2-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
907 // 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]])
908 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
909 // CHECK2: omp.inner.for.inc:
910 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
911 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
912 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
913 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
914 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
915 // CHECK2: omp.inner.for.end:
916 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
917 // CHECK2: omp.loop.exit:
918 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
919 // CHECK2-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
920 // CHECK2-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
921 // CHECK2-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
922 // CHECK2: .omp.final.then:
923 // CHECK2-NEXT: store i32 123, i32* [[I]], align 4
924 // CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
925 // CHECK2: .omp.final.done:
926 // CHECK2-NEXT: ret void
927 //
928 //
929 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
930 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
931 // CHECK2-NEXT: entry:
932 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
933 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
934 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
935 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
936 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
937 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
938 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
939 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
940 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
941 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
942 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
943 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
944 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
945 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
946 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
947 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
948 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
949 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
950 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
951 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
952 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
953 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
954 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
955 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
956 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
957 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
958 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
959 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
960 // CHECK2-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
961 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
962 // 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)
963 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
964 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
965 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
966 // CHECK2: cond.true:
967 // CHECK2-NEXT: br label [[COND_END:%.*]]
968 // CHECK2: cond.false:
969 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
970 // CHECK2-NEXT: br label [[COND_END]]
971 // CHECK2: cond.end:
972 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
973 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
974 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
975 // CHECK2-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
976 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
977 // CHECK2: omp.inner.for.cond:
978 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
979 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
980 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
981 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
982 // CHECK2: omp.inner.for.body:
983 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
984 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
985 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
986 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
987 // CHECK2-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
988 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
989 // CHECK2-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
990 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
991 // CHECK2-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
992 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
993 // CHECK2: omp.body.continue:
994 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
995 // CHECK2: omp.inner.for.inc:
996 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
997 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
998 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
999 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
1000 // CHECK2: omp.inner.for.end:
1001 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1002 // CHECK2: omp.loop.exit:
1003 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1004 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1005 // CHECK2-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1006 // CHECK2-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1007 // CHECK2: .omp.final.then:
1008 // CHECK2-NEXT: store i32 123, i32* [[I]], align 4
1009 // CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
1010 // CHECK2: .omp.final.done:
1011 // CHECK2-NEXT: ret void
1012 //
1013 //
1014 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33
1015 // CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1016 // CHECK2-NEXT: entry:
1017 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1018 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1019 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1020 // 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]])
1021 // CHECK2-NEXT: ret void
1022 //
1023 //
1024 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
1025 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1026 // CHECK2-NEXT: entry:
1027 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1028 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1029 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1030 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1031 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1032 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1033 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1034 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1035 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1036 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1037 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1038 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1039 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1040 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1041 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1042 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1043 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1044 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1045 // CHECK2-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1046 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1047 // 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)
1048 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1049 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1050 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1051 // CHECK2: cond.true:
1052 // CHECK2-NEXT: br label [[COND_END:%.*]]
1053 // CHECK2: cond.false:
1054 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1055 // CHECK2-NEXT: br label [[COND_END]]
1056 // CHECK2: cond.end:
1057 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1058 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1059 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1060 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1061 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1062 // CHECK2: omp.inner.for.cond:
1063 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1064 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1065 // CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1066 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1067 // CHECK2: omp.inner.for.body:
1068 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1069 // CHECK2-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
1070 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1071 // CHECK2-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
1072 // 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]])
1073 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1074 // CHECK2: omp.inner.for.inc:
1075 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1076 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1077 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1078 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1079 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
1080 // CHECK2: omp.inner.for.end:
1081 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1082 // CHECK2: omp.loop.exit:
1083 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1084 // CHECK2-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1085 // CHECK2-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
1086 // CHECK2-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1087 // CHECK2: .omp.final.then:
1088 // CHECK2-NEXT: store i32 123, i32* [[I]], align 4
1089 // CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
1090 // CHECK2: .omp.final.done:
1091 // CHECK2-NEXT: ret void
1092 //
1093 //
1094 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
1095 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1096 // CHECK2-NEXT: entry:
1097 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1098 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1099 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1100 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1101 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1102 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1103 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1104 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1105 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1106 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1107 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1108 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1109 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1110 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1111 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1112 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1113 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1114 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1115 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1116 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1117 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1118 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
1119 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1120 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
1121 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1122 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1123 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1124 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1125 // CHECK2-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1126 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1127 // 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)
1128 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1129 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1130 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1131 // CHECK2: cond.true:
1132 // CHECK2-NEXT: br label [[COND_END:%.*]]
1133 // CHECK2: cond.false:
1134 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1135 // CHECK2-NEXT: br label [[COND_END]]
1136 // CHECK2: cond.end:
1137 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1138 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1139 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1140 // CHECK2-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1141 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1142 // CHECK2: omp.inner.for.cond:
1143 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1144 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1145 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1146 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1147 // CHECK2: omp.inner.for.body:
1148 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1149 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1150 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1151 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1152 // CHECK2-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1153 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
1154 // CHECK2-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
1155 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
1156 // CHECK2-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1157 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1158 // CHECK2: omp.body.continue:
1159 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1160 // CHECK2: omp.inner.for.inc:
1161 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1162 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
1163 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1164 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
1165 // CHECK2: omp.inner.for.end:
1166 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1167 // CHECK2: omp.loop.exit:
1168 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1169 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1170 // CHECK2-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1171 // CHECK2-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1172 // CHECK2: .omp.final.then:
1173 // CHECK2-NEXT: store i32 123, i32* [[I]], align 4
1174 // CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
1175 // CHECK2: .omp.final.done:
1176 // CHECK2-NEXT: ret void
1177 //
1178 //
1179 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38
1180 // CHECK2-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1181 // CHECK2-NEXT: entry:
1182 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1183 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1184 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1185 // 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]])
1186 // CHECK2-NEXT: ret void
1187 //
1188 //
1189 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..6
1190 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1191 // CHECK2-NEXT: entry:
1192 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1193 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1194 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1195 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1196 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1197 // CHECK2-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1198 // CHECK2-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1199 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1200 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1201 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1202 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1203 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1204 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1205 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1206 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1207 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1208 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1209 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1210 // CHECK2-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1211 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1212 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
1213 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1214 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1215 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1216 // CHECK2: cond.true:
1217 // CHECK2-NEXT: br label [[COND_END:%.*]]
1218 // CHECK2: cond.false:
1219 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1220 // CHECK2-NEXT: br label [[COND_END]]
1221 // CHECK2: cond.end:
1222 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1223 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1224 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1225 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1226 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1227 // CHECK2: omp.inner.for.cond:
1228 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1229 // CHECK2-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 123
1230 // CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1231 // CHECK2: omp.inner.for.body:
1232 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1233 // CHECK2-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1234 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1235 // CHECK2-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1236 // 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 [[TMP8]], i64 [[TMP10]], %struct.SS* [[TMP0]])
1237 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1238 // CHECK2: omp.inner.for.inc:
1239 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1240 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1241 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1242 // CHECK2-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1243 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1244 // CHECK2-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1245 // CHECK2-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
1246 // CHECK2-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4
1247 // CHECK2-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1248 // CHECK2-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1249 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
1250 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4
1251 // CHECK2-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1252 // CHECK2-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP17]], 122
1253 // CHECK2-NEXT: br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
1254 // CHECK2: cond.true5:
1255 // CHECK2-NEXT: br label [[COND_END7:%.*]]
1256 // CHECK2: cond.false6:
1257 // CHECK2-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1258 // CHECK2-NEXT: br label [[COND_END7]]
1259 // CHECK2: cond.end7:
1260 // CHECK2-NEXT: [[COND8:%.*]] = phi i32 [ 122, [[COND_TRUE5]] ], [ [[TMP18]], [[COND_FALSE6]] ]
1261 // CHECK2-NEXT: store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4
1262 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1263 // CHECK2-NEXT: store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4
1264 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
1265 // CHECK2: omp.inner.for.end:
1266 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1267 // CHECK2: omp.loop.exit:
1268 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1269 // CHECK2-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1270 // CHECK2-NEXT: [[TMP21:%.*]] = icmp ne i32 [[TMP20]], 0
1271 // CHECK2-NEXT: br i1 [[TMP21]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1272 // CHECK2: .omp.final.then:
1273 // CHECK2-NEXT: store i32 123, i32* [[I]], align 4
1274 // CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
1275 // CHECK2: .omp.final.done:
1276 // CHECK2-NEXT: ret void
1277 //
1278 //
1279 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..7
1280 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1281 // CHECK2-NEXT: entry:
1282 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1283 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1284 // CHECK2-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1285 // CHECK2-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1286 // CHECK2-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1287 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1288 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
1289 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1290 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1291 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1292 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1293 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
1294 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1295 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1296 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1297 // CHECK2-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1298 // CHECK2-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1299 // CHECK2-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1300 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1301 // CHECK2-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1302 // CHECK2-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1303 // CHECK2-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
1304 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1305 // CHECK2-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
1306 // CHECK2-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1307 // CHECK2-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1308 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1309 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1310 // CHECK2-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1311 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1312 // 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)
1313 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1314 // CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1315 // CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1316 // CHECK2: cond.true:
1317 // CHECK2-NEXT: br label [[COND_END:%.*]]
1318 // CHECK2: cond.false:
1319 // CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1320 // CHECK2-NEXT: br label [[COND_END]]
1321 // CHECK2: cond.end:
1322 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1323 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1324 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1325 // CHECK2-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1326 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1327 // CHECK2: omp.inner.for.cond:
1328 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1329 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1330 // CHECK2-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1331 // CHECK2-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1332 // CHECK2: omp.inner.for.body:
1333 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1334 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1335 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1336 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1337 // CHECK2-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1338 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
1339 // CHECK2-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
1340 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
1341 // CHECK2-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1342 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1343 // CHECK2: omp.body.continue:
1344 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1345 // CHECK2: omp.inner.for.inc:
1346 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1347 // CHECK2-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
1348 // CHECK2-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1349 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
1350 // CHECK2: omp.inner.for.end:
1351 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1352 // CHECK2: omp.loop.exit:
1353 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1354 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1355 // CHECK2-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1356 // CHECK2-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1357 // CHECK2: .omp.final.then:
1358 // CHECK2-NEXT: store i32 123, i32* [[I]], align 4
1359 // CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
1360 // CHECK2: .omp.final.done:
1361 // CHECK2-NEXT: ret void
1362 //
1363 //
1364 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1365 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] {
1366 // CHECK2-NEXT: entry:
1367 // CHECK2-NEXT: call void @__tgt_register_requires(i64 1)
1368 // CHECK2-NEXT: ret void
1369 //
1370 //
1371 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
1372 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
1373 // CHECK3-NEXT: entry:
1374 // CHECK3-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
1375 // CHECK3-NEXT: [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
1376 // CHECK3-NEXT: ret i32 [[CALL]]
1377 //
1378 //
1379 // CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
1380 // CHECK3-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
1381 // CHECK3-NEXT: entry:
1382 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1383 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
1384 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
1385 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
1386 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1387 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
1388 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
1389 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
1390 // CHECK3-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
1391 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 4
1392 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 4
1393 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 4
1394 // CHECK3-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
1395 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1396 // CHECK3-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1397 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
1398 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1399 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
1400 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
1401 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1402 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
1403 // CHECK3-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
1404 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1405 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
1406 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1407 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1408 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
1409 // 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_l28.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)
1410 // CHECK3-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
1411 // CHECK3-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1412 // CHECK3: omp_offload.failed:
1413 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
1414 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
1415 // CHECK3: omp_offload.cont:
1416 // CHECK3-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1417 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
1418 // CHECK3-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
1419 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 4
1420 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
1421 // CHECK3-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
1422 // CHECK3-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 4
1423 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
1424 // CHECK3-NEXT: store i8* null, i8** [[TMP13]], align 4
1425 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
1426 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
1427 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
1428 // 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_l33.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)
1429 // CHECK3-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
1430 // CHECK3-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
1431 // CHECK3: omp_offload.failed7:
1432 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33(%struct.SS* [[THIS1]]) #[[ATTR2]]
1433 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT8]]
1434 // CHECK3: omp_offload.cont8:
1435 // CHECK3-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1436 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
1437 // CHECK3-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
1438 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
1439 // CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
1440 // CHECK3-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
1441 // CHECK3-NEXT: store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 4
1442 // CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i32 0, i32 0
1443 // CHECK3-NEXT: store i8* null, i8** [[TMP22]], align 4
1444 // CHECK3-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
1445 // CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
1446 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
1447 // 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_l38.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)
1448 // CHECK3-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
1449 // CHECK3-NEXT: br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
1450 // CHECK3: omp_offload.failed14:
1451 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38(%struct.SS* [[THIS1]]) #[[ATTR2]]
1452 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT15]]
1453 // CHECK3: omp_offload.cont15:
1454 // CHECK3-NEXT: [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1455 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A16]], i32 0, i32 0
1456 // CHECK3-NEXT: [[TMP27:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1457 // CHECK3-NEXT: ret i32 [[TMP27]]
1458 //
1459 //
1460 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
1461 // CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
1462 // CHECK3-NEXT: entry:
1463 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1464 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1465 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1466 // 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]])
1467 // CHECK3-NEXT: ret void
1468 //
1469 //
1470 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
1471 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1472 // CHECK3-NEXT: entry:
1473 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1474 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1475 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1476 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1477 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1478 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1479 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1480 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1481 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1482 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1483 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1484 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1485 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1486 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1487 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1488 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1489 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1490 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1491 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1492 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1493 // 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)
1494 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1495 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1496 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1497 // CHECK3: cond.true:
1498 // CHECK3-NEXT: br label [[COND_END:%.*]]
1499 // CHECK3: cond.false:
1500 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1501 // CHECK3-NEXT: br label [[COND_END]]
1502 // CHECK3: cond.end:
1503 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1504 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1505 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1506 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1507 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1508 // CHECK3: omp.inner.for.cond:
1509 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1510 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1511 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1512 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1513 // CHECK3: omp.inner.for.body:
1514 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1515 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1516 // 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]])
1517 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1518 // CHECK3: omp.inner.for.inc:
1519 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1520 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1521 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
1522 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1523 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
1524 // CHECK3: omp.inner.for.end:
1525 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1526 // CHECK3: omp.loop.exit:
1527 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1528 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1529 // CHECK3-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
1530 // CHECK3-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1531 // CHECK3: .omp.final.then:
1532 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
1533 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1534 // CHECK3: .omp.final.done:
1535 // CHECK3-NEXT: ret void
1536 //
1537 //
1538 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
1539 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1540 // CHECK3-NEXT: entry:
1541 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1542 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1543 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1544 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1545 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1546 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1547 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1548 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1549 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1550 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1551 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1552 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1553 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1554 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1555 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1556 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1557 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1558 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1559 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1560 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1561 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1562 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1563 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
1564 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
1565 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1566 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1567 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1568 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1569 // 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)
1570 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1571 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1572 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1573 // CHECK3: cond.true:
1574 // CHECK3-NEXT: br label [[COND_END:%.*]]
1575 // CHECK3: cond.false:
1576 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1577 // CHECK3-NEXT: br label [[COND_END]]
1578 // CHECK3: cond.end:
1579 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1580 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1581 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1582 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1583 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1584 // CHECK3: omp.inner.for.cond:
1585 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1586 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1587 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1588 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1589 // CHECK3: omp.inner.for.body:
1590 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1591 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1592 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1593 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1594 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1595 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
1596 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
1597 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1598 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1599 // CHECK3: omp.body.continue:
1600 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1601 // CHECK3: omp.inner.for.inc:
1602 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1603 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
1604 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
1605 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP8:![0-9]+]]
1606 // CHECK3: omp.inner.for.end:
1607 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1608 // CHECK3: omp.loop.exit:
1609 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1610 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1611 // CHECK3-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1612 // CHECK3-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1613 // CHECK3: .omp.final.then:
1614 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
1615 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1616 // CHECK3: .omp.final.done:
1617 // CHECK3-NEXT: ret void
1618 //
1619 //
1620 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33
1621 // CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1622 // CHECK3-NEXT: entry:
1623 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1624 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1625 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1626 // 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]])
1627 // CHECK3-NEXT: ret void
1628 //
1629 //
1630 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
1631 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1632 // CHECK3-NEXT: entry:
1633 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1634 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1635 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1636 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1637 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1638 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1639 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1640 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1641 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1642 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1643 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1644 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1645 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1646 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1647 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1648 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1649 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1650 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1651 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1652 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1653 // 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)
1654 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1655 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1656 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1657 // CHECK3: cond.true:
1658 // CHECK3-NEXT: br label [[COND_END:%.*]]
1659 // CHECK3: cond.false:
1660 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1661 // CHECK3-NEXT: br label [[COND_END]]
1662 // CHECK3: cond.end:
1663 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1664 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1665 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1666 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1667 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1668 // CHECK3: omp.inner.for.cond:
1669 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1670 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1671 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1672 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1673 // CHECK3: omp.inner.for.body:
1674 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1675 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1676 // 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]])
1677 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1678 // CHECK3: omp.inner.for.inc:
1679 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1680 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1681 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
1682 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1683 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
1684 // CHECK3: omp.inner.for.end:
1685 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1686 // CHECK3: omp.loop.exit:
1687 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1688 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1689 // CHECK3-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
1690 // CHECK3-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1691 // CHECK3: .omp.final.then:
1692 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
1693 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1694 // CHECK3: .omp.final.done:
1695 // CHECK3-NEXT: ret void
1696 //
1697 //
1698 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
1699 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1700 // CHECK3-NEXT: entry:
1701 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1702 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1703 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1704 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1705 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1706 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1707 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1708 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1709 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1710 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1711 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1712 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1713 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1714 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1715 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1716 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1717 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1718 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1719 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1720 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1721 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1722 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1723 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
1724 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
1725 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1726 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1727 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1728 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1729 // 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)
1730 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1731 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1732 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1733 // CHECK3: cond.true:
1734 // CHECK3-NEXT: br label [[COND_END:%.*]]
1735 // CHECK3: cond.false:
1736 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1737 // CHECK3-NEXT: br label [[COND_END]]
1738 // CHECK3: cond.end:
1739 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1740 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1741 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1742 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1743 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1744 // CHECK3: omp.inner.for.cond:
1745 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1746 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1747 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1748 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1749 // CHECK3: omp.inner.for.body:
1750 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1751 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1752 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1753 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1754 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1755 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
1756 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
1757 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1758 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1759 // CHECK3: omp.body.continue:
1760 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1761 // CHECK3: omp.inner.for.inc:
1762 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1763 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
1764 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
1765 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
1766 // CHECK3: omp.inner.for.end:
1767 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1768 // CHECK3: omp.loop.exit:
1769 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1770 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1771 // CHECK3-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1772 // CHECK3-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1773 // CHECK3: .omp.final.then:
1774 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
1775 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1776 // CHECK3: .omp.final.done:
1777 // CHECK3-NEXT: ret void
1778 //
1779 //
1780 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38
1781 // CHECK3-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1782 // CHECK3-NEXT: entry:
1783 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1784 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1785 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1786 // 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]])
1787 // CHECK3-NEXT: ret void
1788 //
1789 //
1790 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..6
1791 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1792 // CHECK3-NEXT: entry:
1793 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1794 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1795 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1796 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1797 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1798 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1799 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1800 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1801 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1802 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1803 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1804 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1805 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1806 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1807 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1808 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1809 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1810 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1811 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1812 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1813 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
1814 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1815 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1816 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1817 // CHECK3: cond.true:
1818 // CHECK3-NEXT: br label [[COND_END:%.*]]
1819 // CHECK3: cond.false:
1820 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1821 // CHECK3-NEXT: br label [[COND_END]]
1822 // CHECK3: cond.end:
1823 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1824 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1825 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1826 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1827 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1828 // CHECK3: omp.inner.for.cond:
1829 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1830 // CHECK3-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 123
1831 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1832 // CHECK3: omp.inner.for.body:
1833 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1834 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1835 // 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 [[TMP7]], i32 [[TMP8]], %struct.SS* [[TMP0]])
1836 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1837 // CHECK3: omp.inner.for.inc:
1838 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1839 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1840 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
1841 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1842 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1843 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1844 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1845 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4
1846 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1847 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1848 // CHECK3-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
1849 // CHECK3-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4
1850 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1851 // CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP15]], 122
1852 // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
1853 // CHECK3: cond.true5:
1854 // CHECK3-NEXT: br label [[COND_END7:%.*]]
1855 // CHECK3: cond.false6:
1856 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1857 // CHECK3-NEXT: br label [[COND_END7]]
1858 // CHECK3: cond.end7:
1859 // CHECK3-NEXT: [[COND8:%.*]] = phi i32 [ 122, [[COND_TRUE5]] ], [ [[TMP16]], [[COND_FALSE6]] ]
1860 // CHECK3-NEXT: store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4
1861 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1862 // CHECK3-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
1863 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
1864 // CHECK3: omp.inner.for.end:
1865 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1866 // CHECK3: omp.loop.exit:
1867 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1868 // CHECK3-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1869 // CHECK3-NEXT: [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
1870 // CHECK3-NEXT: br i1 [[TMP19]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1871 // CHECK3: .omp.final.then:
1872 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
1873 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1874 // CHECK3: .omp.final.done:
1875 // CHECK3-NEXT: ret void
1876 //
1877 //
1878 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
1879 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
1880 // CHECK3-NEXT: entry:
1881 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1882 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1883 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1884 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1885 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1886 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1887 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1888 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1889 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1890 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1891 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1892 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1893 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1894 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1895 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1896 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1897 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1898 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1899 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1900 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1901 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1902 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1903 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
1904 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
1905 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1906 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1907 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1908 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1909 // 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)
1910 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1911 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1912 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1913 // CHECK3: cond.true:
1914 // CHECK3-NEXT: br label [[COND_END:%.*]]
1915 // CHECK3: cond.false:
1916 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1917 // CHECK3-NEXT: br label [[COND_END]]
1918 // CHECK3: cond.end:
1919 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1920 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1921 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1922 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1923 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1924 // CHECK3: omp.inner.for.cond:
1925 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1926 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1927 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1928 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1929 // CHECK3: omp.inner.for.body:
1930 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1931 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1932 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1933 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1934 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1935 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
1936 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
1937 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1938 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1939 // CHECK3: omp.body.continue:
1940 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1941 // CHECK3: omp.inner.for.inc:
1942 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1943 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
1944 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
1945 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
1946 // CHECK3: omp.inner.for.end:
1947 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1948 // CHECK3: omp.loop.exit:
1949 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1950 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1951 // CHECK3-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1952 // CHECK3-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1953 // CHECK3: .omp.final.then:
1954 // CHECK3-NEXT: store i32 123, i32* [[I]], align 4
1955 // CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
1956 // CHECK3: .omp.final.done:
1957 // CHECK3-NEXT: ret void
1958 //
1959 //
1960 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1961 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
1962 // CHECK3-NEXT: entry:
1963 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
1964 // CHECK3-NEXT: ret void
1965 //
1966 //
1967 // CHECK4-LABEL: define {{[^@]+}}@_Z21teams_template_structv
1968 // CHECK4-SAME: () #[[ATTR0:[0-9]+]] {
1969 // CHECK4-NEXT: entry:
1970 // CHECK4-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
1971 // CHECK4-NEXT: [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
1972 // CHECK4-NEXT: ret i32 [[CALL]]
1973 //
1974 //
1975 // CHECK4-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
1976 // CHECK4-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
1977 // CHECK4-NEXT: entry:
1978 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1979 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
1980 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
1981 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
1982 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
1983 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
1984 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
1985 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
1986 // CHECK4-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
1987 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS10:%.*]] = alloca [1 x i8*], align 4
1988 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS11:%.*]] = alloca [1 x i8*], align 4
1989 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS12:%.*]] = alloca [1 x i8*], align 4
1990 // CHECK4-NEXT: [[_TMP13:%.*]] = alloca i32, align 4
1991 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1992 // CHECK4-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1993 // CHECK4-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
1994 // CHECK4-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1995 // CHECK4-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
1996 // CHECK4-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
1997 // CHECK4-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1998 // CHECK4-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
1999 // CHECK4-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
2000 // CHECK4-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2001 // CHECK4-NEXT: store i8* null, i8** [[TMP4]], align 4
2002 // CHECK4-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2003 // CHECK4-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2004 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 123)
2005 // 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_l28.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)
2006 // CHECK4-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
2007 // CHECK4-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2008 // CHECK4: omp_offload.failed:
2009 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
2010 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT]]
2011 // CHECK4: omp_offload.cont:
2012 // CHECK4-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2013 // CHECK4-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
2014 // CHECK4-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to %struct.SS**
2015 // CHECK4-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP10]], align 4
2016 // CHECK4-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
2017 // CHECK4-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [123 x i32]**
2018 // CHECK4-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP12]], align 4
2019 // CHECK4-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
2020 // CHECK4-NEXT: store i8* null, i8** [[TMP13]], align 4
2021 // CHECK4-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
2022 // CHECK4-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
2023 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
2024 // 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_l33.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)
2025 // CHECK4-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
2026 // CHECK4-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED7:%.*]], label [[OMP_OFFLOAD_CONT8:%.*]]
2027 // CHECK4: omp_offload.failed7:
2028 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33(%struct.SS* [[THIS1]]) #[[ATTR2]]
2029 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT8]]
2030 // CHECK4: omp_offload.cont8:
2031 // CHECK4-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2032 // CHECK4-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
2033 // CHECK4-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
2034 // CHECK4-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
2035 // CHECK4-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
2036 // CHECK4-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
2037 // CHECK4-NEXT: store [123 x i32]* [[A9]], [123 x i32]** [[TMP21]], align 4
2038 // CHECK4-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS12]], i32 0, i32 0
2039 // CHECK4-NEXT: store i8* null, i8** [[TMP22]], align 4
2040 // CHECK4-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS10]], i32 0, i32 0
2041 // CHECK4-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS11]], i32 0, i32 0
2042 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 123)
2043 // 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_l38.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)
2044 // CHECK4-NEXT: [[TMP26:%.*]] = icmp ne i32 [[TMP25]], 0
2045 // CHECK4-NEXT: br i1 [[TMP26]], label [[OMP_OFFLOAD_FAILED14:%.*]], label [[OMP_OFFLOAD_CONT15:%.*]]
2046 // CHECK4: omp_offload.failed14:
2047 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38(%struct.SS* [[THIS1]]) #[[ATTR2]]
2048 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT15]]
2049 // CHECK4: omp_offload.cont15:
2050 // CHECK4-NEXT: [[A16:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2051 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A16]], i32 0, i32 0
2052 // CHECK4-NEXT: [[TMP27:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2053 // CHECK4-NEXT: ret i32 [[TMP27]]
2054 //
2055 //
2056 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
2057 // CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
2058 // CHECK4-NEXT: entry:
2059 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2060 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2061 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2062 // 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]])
2063 // CHECK4-NEXT: ret void
2064 //
2065 //
2066 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
2067 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2068 // CHECK4-NEXT: entry:
2069 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2070 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2071 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2072 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2073 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
2074 // CHECK4-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2075 // CHECK4-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2076 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2077 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2078 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
2079 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2080 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2081 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2082 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2083 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2084 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
2085 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2086 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2087 // CHECK4-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2088 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2089 // 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)
2090 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2091 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2092 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2093 // CHECK4: cond.true:
2094 // CHECK4-NEXT: br label [[COND_END:%.*]]
2095 // CHECK4: cond.false:
2096 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2097 // CHECK4-NEXT: br label [[COND_END]]
2098 // CHECK4: cond.end:
2099 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2100 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2101 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2102 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2103 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2104 // CHECK4: omp.inner.for.cond:
2105 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2106 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2107 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2108 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2109 // CHECK4: omp.inner.for.body:
2110 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2111 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2112 // 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]])
2113 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2114 // CHECK4: omp.inner.for.inc:
2115 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2116 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2117 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
2118 // CHECK4-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2119 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
2120 // CHECK4: omp.inner.for.end:
2121 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2122 // CHECK4: omp.loop.exit:
2123 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2124 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2125 // CHECK4-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
2126 // CHECK4-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2127 // CHECK4: .omp.final.then:
2128 // CHECK4-NEXT: store i32 123, i32* [[I]], align 4
2129 // CHECK4-NEXT: br label [[DOTOMP_FINAL_DONE]]
2130 // CHECK4: .omp.final.done:
2131 // CHECK4-NEXT: ret void
2132 //
2133 //
2134 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
2135 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2136 // CHECK4-NEXT: entry:
2137 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2138 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2139 // CHECK4-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
2140 // CHECK4-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
2141 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2142 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2143 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
2144 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2145 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2146 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2147 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2148 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
2149 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2150 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2151 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2152 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2153 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2154 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2155 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2156 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2157 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2158 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2159 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
2160 // CHECK4-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
2161 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2162 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2163 // CHECK4-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2164 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2165 // 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)
2166 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2167 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
2168 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2169 // CHECK4: cond.true:
2170 // CHECK4-NEXT: br label [[COND_END:%.*]]
2171 // CHECK4: cond.false:
2172 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2173 // CHECK4-NEXT: br label [[COND_END]]
2174 // CHECK4: cond.end:
2175 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2176 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2177 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2178 // CHECK4-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
2179 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2180 // CHECK4: omp.inner.for.cond:
2181 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2182 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2183 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2184 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2185 // CHECK4: omp.inner.for.body:
2186 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2187 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2188 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2189 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2190 // CHECK4-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2191 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
2192 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
2193 // CHECK4-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2194 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2195 // CHECK4: omp.body.continue:
2196 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2197 // CHECK4: omp.inner.for.inc:
2198 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2199 // CHECK4-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
2200 // CHECK4-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2201 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP8:![0-9]+]]
2202 // CHECK4: omp.inner.for.end:
2203 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2204 // CHECK4: omp.loop.exit:
2205 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
2206 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2207 // CHECK4-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
2208 // CHECK4-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2209 // CHECK4: .omp.final.then:
2210 // CHECK4-NEXT: store i32 123, i32* [[I]], align 4
2211 // CHECK4-NEXT: br label [[DOTOMP_FINAL_DONE]]
2212 // CHECK4: .omp.final.done:
2213 // CHECK4-NEXT: ret void
2214 //
2215 //
2216 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33
2217 // CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2218 // CHECK4-NEXT: entry:
2219 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2220 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2221 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2222 // 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]])
2223 // CHECK4-NEXT: ret void
2224 //
2225 //
2226 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
2227 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2228 // CHECK4-NEXT: entry:
2229 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2230 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2231 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2232 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2233 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
2234 // CHECK4-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2235 // CHECK4-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2236 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2237 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2238 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
2239 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2240 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2241 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2242 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2243 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2244 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
2245 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2246 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2247 // CHECK4-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2248 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2249 // 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)
2250 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2251 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2252 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2253 // CHECK4: cond.true:
2254 // CHECK4-NEXT: br label [[COND_END:%.*]]
2255 // CHECK4: cond.false:
2256 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2257 // CHECK4-NEXT: br label [[COND_END]]
2258 // CHECK4: cond.end:
2259 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2260 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2261 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2262 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2263 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2264 // CHECK4: omp.inner.for.cond:
2265 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2266 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2267 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2268 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2269 // CHECK4: omp.inner.for.body:
2270 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2271 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2272 // 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]])
2273 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2274 // CHECK4: omp.inner.for.inc:
2275 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2276 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2277 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
2278 // CHECK4-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2279 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
2280 // CHECK4: omp.inner.for.end:
2281 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2282 // CHECK4: omp.loop.exit:
2283 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2284 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2285 // CHECK4-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
2286 // CHECK4-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2287 // CHECK4: .omp.final.then:
2288 // CHECK4-NEXT: store i32 123, i32* [[I]], align 4
2289 // CHECK4-NEXT: br label [[DOTOMP_FINAL_DONE]]
2290 // CHECK4: .omp.final.done:
2291 // CHECK4-NEXT: ret void
2292 //
2293 //
2294 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3
2295 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2296 // CHECK4-NEXT: entry:
2297 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2298 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2299 // CHECK4-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
2300 // CHECK4-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
2301 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2302 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2303 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
2304 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2305 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2306 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2307 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2308 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
2309 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2310 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2311 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2312 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2313 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2314 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2315 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2316 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2317 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2318 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2319 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
2320 // CHECK4-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
2321 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2322 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2323 // CHECK4-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2324 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2325 // 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)
2326 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2327 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
2328 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2329 // CHECK4: cond.true:
2330 // CHECK4-NEXT: br label [[COND_END:%.*]]
2331 // CHECK4: cond.false:
2332 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2333 // CHECK4-NEXT: br label [[COND_END]]
2334 // CHECK4: cond.end:
2335 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2336 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2337 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2338 // CHECK4-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
2339 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2340 // CHECK4: omp.inner.for.cond:
2341 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2342 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2343 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2344 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2345 // CHECK4: omp.inner.for.body:
2346 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2347 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2348 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2349 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2350 // CHECK4-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2351 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
2352 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
2353 // CHECK4-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2354 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2355 // CHECK4: omp.body.continue:
2356 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2357 // CHECK4: omp.inner.for.inc:
2358 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2359 // CHECK4-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
2360 // CHECK4-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2361 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
2362 // CHECK4: omp.inner.for.end:
2363 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2364 // CHECK4: omp.loop.exit:
2365 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
2366 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2367 // CHECK4-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
2368 // CHECK4-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2369 // CHECK4: .omp.final.then:
2370 // CHECK4-NEXT: store i32 123, i32* [[I]], align 4
2371 // CHECK4-NEXT: br label [[DOTOMP_FINAL_DONE]]
2372 // CHECK4: .omp.final.done:
2373 // CHECK4-NEXT: ret void
2374 //
2375 //
2376 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38
2377 // CHECK4-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2378 // CHECK4-NEXT: entry:
2379 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2380 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2381 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2382 // 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]])
2383 // CHECK4-NEXT: ret void
2384 //
2385 //
2386 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..6
2387 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2388 // CHECK4-NEXT: entry:
2389 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2390 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2391 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2392 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2393 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
2394 // CHECK4-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2395 // CHECK4-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2396 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2397 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2398 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
2399 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2400 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2401 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2402 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2403 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2404 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
2405 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2406 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2407 // CHECK4-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2408 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2409 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
2410 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2411 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2412 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2413 // CHECK4: cond.true:
2414 // CHECK4-NEXT: br label [[COND_END:%.*]]
2415 // CHECK4: cond.false:
2416 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2417 // CHECK4-NEXT: br label [[COND_END]]
2418 // CHECK4: cond.end:
2419 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2420 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2421 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2422 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2423 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2424 // CHECK4: omp.inner.for.cond:
2425 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2426 // CHECK4-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 123
2427 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2428 // CHECK4: omp.inner.for.body:
2429 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2430 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2431 // 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 [[TMP7]], i32 [[TMP8]], %struct.SS* [[TMP0]])
2432 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2433 // CHECK4: omp.inner.for.inc:
2434 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2435 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2436 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
2437 // CHECK4-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2438 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2439 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2440 // CHECK4-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
2441 // CHECK4-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4
2442 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2443 // CHECK4-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2444 // CHECK4-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
2445 // CHECK4-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4
2446 // CHECK4-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2447 // CHECK4-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP15]], 122
2448 // CHECK4-NEXT: br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
2449 // CHECK4: cond.true5:
2450 // CHECK4-NEXT: br label [[COND_END7:%.*]]
2451 // CHECK4: cond.false6:
2452 // CHECK4-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2453 // CHECK4-NEXT: br label [[COND_END7]]
2454 // CHECK4: cond.end7:
2455 // CHECK4-NEXT: [[COND8:%.*]] = phi i32 [ 122, [[COND_TRUE5]] ], [ [[TMP16]], [[COND_FALSE6]] ]
2456 // CHECK4-NEXT: store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4
2457 // CHECK4-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2458 // CHECK4-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
2459 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
2460 // CHECK4: omp.inner.for.end:
2461 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2462 // CHECK4: omp.loop.exit:
2463 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2464 // CHECK4-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2465 // CHECK4-NEXT: [[TMP19:%.*]] = icmp ne i32 [[TMP18]], 0
2466 // CHECK4-NEXT: br i1 [[TMP19]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2467 // CHECK4: .omp.final.then:
2468 // CHECK4-NEXT: store i32 123, i32* [[I]], align 4
2469 // CHECK4-NEXT: br label [[DOTOMP_FINAL_DONE]]
2470 // CHECK4: .omp.final.done:
2471 // CHECK4-NEXT: ret void
2472 //
2473 //
2474 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..7
2475 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32 [[DOTPREVIOUS_LB_:%.*]], i32 [[DOTPREVIOUS_UB_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2476 // CHECK4-NEXT: entry:
2477 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2478 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2479 // CHECK4-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
2480 // CHECK4-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
2481 // CHECK4-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2482 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2483 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
2484 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2485 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2486 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2487 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2488 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
2489 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2490 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2491 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2492 // CHECK4-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2493 // CHECK4-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2494 // CHECK4-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2495 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2496 // CHECK4-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2497 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
2498 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
2499 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
2500 // CHECK4-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
2501 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2502 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2503 // CHECK4-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2504 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2505 // 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)
2506 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2507 // CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
2508 // CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2509 // CHECK4: cond.true:
2510 // CHECK4-NEXT: br label [[COND_END:%.*]]
2511 // CHECK4: cond.false:
2512 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2513 // CHECK4-NEXT: br label [[COND_END]]
2514 // CHECK4: cond.end:
2515 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2516 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2517 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2518 // CHECK4-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
2519 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2520 // CHECK4: omp.inner.for.cond:
2521 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2522 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2523 // CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2524 // CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2525 // CHECK4: omp.inner.for.body:
2526 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2527 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2528 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2529 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2530 // CHECK4-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2531 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
2532 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
2533 // CHECK4-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2534 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2535 // CHECK4: omp.body.continue:
2536 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2537 // CHECK4: omp.inner.for.inc:
2538 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2539 // CHECK4-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
2540 // CHECK4-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2541 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
2542 // CHECK4: omp.inner.for.end:
2543 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2544 // CHECK4: omp.loop.exit:
2545 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
2546 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2547 // CHECK4-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
2548 // CHECK4-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2549 // CHECK4: .omp.final.then:
2550 // CHECK4-NEXT: store i32 123, i32* [[I]], align 4
2551 // CHECK4-NEXT: br label [[DOTOMP_FINAL_DONE]]
2552 // CHECK4: .omp.final.done:
2553 // CHECK4-NEXT: ret void
2554 //
2555 //
2556 // CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2557 // CHECK4-SAME: () #[[ATTR3:[0-9]+]] {
2558 // CHECK4-NEXT: entry:
2559 // CHECK4-NEXT: call void @__tgt_register_requires(i64 1)
2560 // CHECK4-NEXT: ret void
2561 //
2562 //
2563 // CHECK5-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2564 // CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
2565 // CHECK5-NEXT: entry:
2566 // CHECK5-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2567 // CHECK5-NEXT: [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
2568 // CHECK5-NEXT: ret i32 [[CALL]]
2569 //
2570 //
2571 // CHECK5-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2572 // CHECK5-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2573 // CHECK5-NEXT: entry:
2574 // CHECK5-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2575 // CHECK5-NEXT: [[TMP:%.*]] = alloca i32, align 4
2576 // CHECK5-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2577 // CHECK5-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2578 // CHECK5-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2579 // CHECK5-NEXT: [[I:%.*]] = alloca i32, align 4
2580 // CHECK5-NEXT: [[_TMP3:%.*]] = alloca i32, align 4
2581 // CHECK5-NEXT: [[DOTOMP_LB4:%.*]] = alloca i32, align 4
2582 // CHECK5-NEXT: [[DOTOMP_UB5:%.*]] = alloca i32, align 4
2583 // CHECK5-NEXT: [[DOTOMP_IV6:%.*]] = alloca i32, align 4
2584 // CHECK5-NEXT: [[I7:%.*]] = alloca i32, align 4
2585 // CHECK5-NEXT: [[_TMP20:%.*]] = alloca i32, align 4
2586 // CHECK5-NEXT: [[DOTOMP_LB21:%.*]] = alloca i32, align 4
2587 // CHECK5-NEXT: [[DOTOMP_UB22:%.*]] = alloca i32, align 4
2588 // CHECK5-NEXT: [[DOTOMP_IV23:%.*]] = alloca i32, align 4
2589 // CHECK5-NEXT: [[I24:%.*]] = alloca i32, align 4
2590 // CHECK5-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2591 // CHECK5-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2592 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2593 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2594 // CHECK5-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2595 // CHECK5-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
2596 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2597 // CHECK5: omp.inner.for.cond:
2598 // CHECK5-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2599 // CHECK5-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
2600 // CHECK5-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
2601 // CHECK5-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2602 // CHECK5: omp.inner.for.body:
2603 // CHECK5-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2604 // CHECK5-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
2605 // CHECK5-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2606 // CHECK5-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
2607 // CHECK5-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
2608 // CHECK5-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !2
2609 // CHECK5-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
2610 // CHECK5-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
2611 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
2612 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2613 // CHECK5: omp.body.continue:
2614 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2615 // CHECK5: omp.inner.for.inc:
2616 // CHECK5-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2617 // CHECK5-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
2618 // CHECK5-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2619 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
2620 // CHECK5: omp.inner.for.end:
2621 // CHECK5-NEXT: store i32 123, i32* [[I]], align 4
2622 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB4]], align 4
2623 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_UB5]], align 4
2624 // CHECK5-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
2625 // CHECK5-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV6]], align 4
2626 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND8:%.*]]
2627 // CHECK5: omp.inner.for.cond8:
2628 // CHECK5-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
2629 // CHECK5-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !6
2630 // CHECK5-NEXT: [[CMP9:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2631 // CHECK5-NEXT: br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END19:%.*]]
2632 // CHECK5: omp.inner.for.body10:
2633 // CHECK5-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
2634 // CHECK5-NEXT: [[MUL11:%.*]] = mul nsw i32 [[TMP9]], 1
2635 // CHECK5-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
2636 // CHECK5-NEXT: store i32 [[ADD12]], i32* [[I7]], align 4, !llvm.access.group !6
2637 // CHECK5-NEXT: [[A13:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2638 // CHECK5-NEXT: [[TMP10:%.*]] = load i32, i32* [[I7]], align 4, !llvm.access.group !6
2639 // CHECK5-NEXT: [[IDXPROM14:%.*]] = sext i32 [[TMP10]] to i64
2640 // CHECK5-NEXT: [[ARRAYIDX15:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A13]], i64 0, i64 [[IDXPROM14]]
2641 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX15]], align 4, !llvm.access.group !6
2642 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE16:%.*]]
2643 // CHECK5: omp.body.continue16:
2644 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC17:%.*]]
2645 // CHECK5: omp.inner.for.inc17:
2646 // CHECK5-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
2647 // CHECK5-NEXT: [[ADD18:%.*]] = add nsw i32 [[TMP11]], 1
2648 // CHECK5-NEXT: store i32 [[ADD18]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
2649 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP7:![0-9]+]]
2650 // CHECK5: omp.inner.for.end19:
2651 // CHECK5-NEXT: store i32 123, i32* [[I7]], align 4
2652 // CHECK5-NEXT: store i32 0, i32* [[DOTOMP_LB21]], align 4
2653 // CHECK5-NEXT: store i32 122, i32* [[DOTOMP_UB22]], align 4
2654 // CHECK5-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB21]], align 4
2655 // CHECK5-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV23]], align 4
2656 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND25:%.*]]
2657 // CHECK5: omp.inner.for.cond25:
2658 // CHECK5-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
2659 // CHECK5-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB22]], align 4, !llvm.access.group !9
2660 // CHECK5-NEXT: [[CMP26:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
2661 // CHECK5-NEXT: br i1 [[CMP26]], label [[OMP_INNER_FOR_BODY27:%.*]], label [[OMP_INNER_FOR_END36:%.*]]
2662 // CHECK5: omp.inner.for.body27:
2663 // CHECK5-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
2664 // CHECK5-NEXT: [[MUL28:%.*]] = mul nsw i32 [[TMP15]], 1
2665 // CHECK5-NEXT: [[ADD29:%.*]] = add nsw i32 0, [[MUL28]]
2666 // CHECK5-NEXT: store i32 [[ADD29]], i32* [[I24]], align 4, !llvm.access.group !9
2667 // CHECK5-NEXT: [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2668 // CHECK5-NEXT: [[TMP16:%.*]] = load i32, i32* [[I24]], align 4, !llvm.access.group !9
2669 // CHECK5-NEXT: [[IDXPROM31:%.*]] = sext i32 [[TMP16]] to i64
2670 // CHECK5-NEXT: [[ARRAYIDX32:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 [[IDXPROM31]]
2671 // CHECK5-NEXT: store i32 0, i32* [[ARRAYIDX32]], align 4, !llvm.access.group !9
2672 // CHECK5-NEXT: br label [[OMP_BODY_CONTINUE33:%.*]]
2673 // CHECK5: omp.body.continue33:
2674 // CHECK5-NEXT: br label [[OMP_INNER_FOR_INC34:%.*]]
2675 // CHECK5: omp.inner.for.inc34:
2676 // CHECK5-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
2677 // CHECK5-NEXT: [[ADD35:%.*]] = add nsw i32 [[TMP17]], 1
2678 // CHECK5-NEXT: store i32 [[ADD35]], i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
2679 // CHECK5-NEXT: br label [[OMP_INNER_FOR_COND25]], !llvm.loop [[LOOP10:![0-9]+]]
2680 // CHECK5: omp.inner.for.end36:
2681 // CHECK5-NEXT: store i32 123, i32* [[I24]], align 4
2682 // CHECK5-NEXT: [[A37:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2683 // CHECK5-NEXT: [[ARRAYIDX38:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A37]], i64 0, i64 0
2684 // CHECK5-NEXT: [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX38]], align 4
2685 // CHECK5-NEXT: ret i32 [[TMP18]]
2686 //
2687 //
2688 // CHECK6-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2689 // CHECK6-SAME: () #[[ATTR0:[0-9]+]] {
2690 // CHECK6-NEXT: entry:
2691 // CHECK6-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2692 // CHECK6-NEXT: [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
2693 // CHECK6-NEXT: ret i32 [[CALL]]
2694 //
2695 //
2696 // CHECK6-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2697 // CHECK6-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2698 // CHECK6-NEXT: entry:
2699 // CHECK6-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2700 // CHECK6-NEXT: [[TMP:%.*]] = alloca i32, align 4
2701 // CHECK6-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2702 // CHECK6-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2703 // CHECK6-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2704 // CHECK6-NEXT: [[I:%.*]] = alloca i32, align 4
2705 // CHECK6-NEXT: [[_TMP3:%.*]] = alloca i32, align 4
2706 // CHECK6-NEXT: [[DOTOMP_LB4:%.*]] = alloca i32, align 4
2707 // CHECK6-NEXT: [[DOTOMP_UB5:%.*]] = alloca i32, align 4
2708 // CHECK6-NEXT: [[DOTOMP_IV6:%.*]] = alloca i32, align 4
2709 // CHECK6-NEXT: [[I7:%.*]] = alloca i32, align 4
2710 // CHECK6-NEXT: [[_TMP20:%.*]] = alloca i32, align 4
2711 // CHECK6-NEXT: [[DOTOMP_LB21:%.*]] = alloca i32, align 4
2712 // CHECK6-NEXT: [[DOTOMP_UB22:%.*]] = alloca i32, align 4
2713 // CHECK6-NEXT: [[DOTOMP_IV23:%.*]] = alloca i32, align 4
2714 // CHECK6-NEXT: [[I24:%.*]] = alloca i32, align 4
2715 // CHECK6-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2716 // CHECK6-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2717 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2718 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2719 // CHECK6-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2720 // CHECK6-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
2721 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2722 // CHECK6: omp.inner.for.cond:
2723 // CHECK6-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2724 // CHECK6-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
2725 // CHECK6-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
2726 // CHECK6-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2727 // CHECK6: omp.inner.for.body:
2728 // CHECK6-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2729 // CHECK6-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
2730 // CHECK6-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2731 // CHECK6-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
2732 // CHECK6-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
2733 // CHECK6-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !2
2734 // CHECK6-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
2735 // CHECK6-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
2736 // CHECK6-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
2737 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2738 // CHECK6: omp.body.continue:
2739 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2740 // CHECK6: omp.inner.for.inc:
2741 // CHECK6-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2742 // CHECK6-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
2743 // CHECK6-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
2744 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
2745 // CHECK6: omp.inner.for.end:
2746 // CHECK6-NEXT: store i32 123, i32* [[I]], align 4
2747 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB4]], align 4
2748 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_UB5]], align 4
2749 // CHECK6-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
2750 // CHECK6-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV6]], align 4
2751 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND8:%.*]]
2752 // CHECK6: omp.inner.for.cond8:
2753 // CHECK6-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
2754 // CHECK6-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !6
2755 // CHECK6-NEXT: [[CMP9:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2756 // CHECK6-NEXT: br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END19:%.*]]
2757 // CHECK6: omp.inner.for.body10:
2758 // CHECK6-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
2759 // CHECK6-NEXT: [[MUL11:%.*]] = mul nsw i32 [[TMP9]], 1
2760 // CHECK6-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
2761 // CHECK6-NEXT: store i32 [[ADD12]], i32* [[I7]], align 4, !llvm.access.group !6
2762 // CHECK6-NEXT: [[A13:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2763 // CHECK6-NEXT: [[TMP10:%.*]] = load i32, i32* [[I7]], align 4, !llvm.access.group !6
2764 // CHECK6-NEXT: [[IDXPROM14:%.*]] = sext i32 [[TMP10]] to i64
2765 // CHECK6-NEXT: [[ARRAYIDX15:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A13]], i64 0, i64 [[IDXPROM14]]
2766 // CHECK6-NEXT: store i32 0, i32* [[ARRAYIDX15]], align 4, !llvm.access.group !6
2767 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE16:%.*]]
2768 // CHECK6: omp.body.continue16:
2769 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC17:%.*]]
2770 // CHECK6: omp.inner.for.inc17:
2771 // CHECK6-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
2772 // CHECK6-NEXT: [[ADD18:%.*]] = add nsw i32 [[TMP11]], 1
2773 // CHECK6-NEXT: store i32 [[ADD18]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !6
2774 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP7:![0-9]+]]
2775 // CHECK6: omp.inner.for.end19:
2776 // CHECK6-NEXT: store i32 123, i32* [[I7]], align 4
2777 // CHECK6-NEXT: store i32 0, i32* [[DOTOMP_LB21]], align 4
2778 // CHECK6-NEXT: store i32 122, i32* [[DOTOMP_UB22]], align 4
2779 // CHECK6-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB21]], align 4
2780 // CHECK6-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV23]], align 4
2781 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND25:%.*]]
2782 // CHECK6: omp.inner.for.cond25:
2783 // CHECK6-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
2784 // CHECK6-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB22]], align 4, !llvm.access.group !9
2785 // CHECK6-NEXT: [[CMP26:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
2786 // CHECK6-NEXT: br i1 [[CMP26]], label [[OMP_INNER_FOR_BODY27:%.*]], label [[OMP_INNER_FOR_END36:%.*]]
2787 // CHECK6: omp.inner.for.body27:
2788 // CHECK6-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
2789 // CHECK6-NEXT: [[MUL28:%.*]] = mul nsw i32 [[TMP15]], 1
2790 // CHECK6-NEXT: [[ADD29:%.*]] = add nsw i32 0, [[MUL28]]
2791 // CHECK6-NEXT: store i32 [[ADD29]], i32* [[I24]], align 4, !llvm.access.group !9
2792 // CHECK6-NEXT: [[A30:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2793 // CHECK6-NEXT: [[TMP16:%.*]] = load i32, i32* [[I24]], align 4, !llvm.access.group !9
2794 // CHECK6-NEXT: [[IDXPROM31:%.*]] = sext i32 [[TMP16]] to i64
2795 // CHECK6-NEXT: [[ARRAYIDX32:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A30]], i64 0, i64 [[IDXPROM31]]
2796 // CHECK6-NEXT: store i32 0, i32* [[ARRAYIDX32]], align 4, !llvm.access.group !9
2797 // CHECK6-NEXT: br label [[OMP_BODY_CONTINUE33:%.*]]
2798 // CHECK6: omp.body.continue33:
2799 // CHECK6-NEXT: br label [[OMP_INNER_FOR_INC34:%.*]]
2800 // CHECK6: omp.inner.for.inc34:
2801 // CHECK6-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
2802 // CHECK6-NEXT: [[ADD35:%.*]] = add nsw i32 [[TMP17]], 1
2803 // CHECK6-NEXT: store i32 [[ADD35]], i32* [[DOTOMP_IV23]], align 4, !llvm.access.group !9
2804 // CHECK6-NEXT: br label [[OMP_INNER_FOR_COND25]], !llvm.loop [[LOOP10:![0-9]+]]
2805 // CHECK6: omp.inner.for.end36:
2806 // CHECK6-NEXT: store i32 123, i32* [[I24]], align 4
2807 // CHECK6-NEXT: [[A37:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2808 // CHECK6-NEXT: [[ARRAYIDX38:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A37]], i64 0, i64 0
2809 // CHECK6-NEXT: [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX38]], align 4
2810 // CHECK6-NEXT: ret i32 [[TMP18]]
2811 //
2812 //
2813 // CHECK7-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2814 // CHECK7-SAME: () #[[ATTR0:[0-9]+]] {
2815 // CHECK7-NEXT: entry:
2816 // CHECK7-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2817 // CHECK7-NEXT: [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
2818 // CHECK7-NEXT: ret i32 [[CALL]]
2819 //
2820 //
2821 // CHECK7-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2822 // CHECK7-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2823 // CHECK7-NEXT: entry:
2824 // CHECK7-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2825 // CHECK7-NEXT: [[TMP:%.*]] = alloca i32, align 4
2826 // CHECK7-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2827 // CHECK7-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2828 // CHECK7-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2829 // CHECK7-NEXT: [[I:%.*]] = alloca i32, align 4
2830 // CHECK7-NEXT: [[_TMP3:%.*]] = alloca i32, align 4
2831 // CHECK7-NEXT: [[DOTOMP_LB4:%.*]] = alloca i32, align 4
2832 // CHECK7-NEXT: [[DOTOMP_UB5:%.*]] = alloca i32, align 4
2833 // CHECK7-NEXT: [[DOTOMP_IV6:%.*]] = alloca i32, align 4
2834 // CHECK7-NEXT: [[I7:%.*]] = alloca i32, align 4
2835 // CHECK7-NEXT: [[_TMP19:%.*]] = alloca i32, align 4
2836 // CHECK7-NEXT: [[DOTOMP_LB20:%.*]] = alloca i32, align 4
2837 // CHECK7-NEXT: [[DOTOMP_UB21:%.*]] = alloca i32, align 4
2838 // CHECK7-NEXT: [[DOTOMP_IV22:%.*]] = alloca i32, align 4
2839 // CHECK7-NEXT: [[I23:%.*]] = alloca i32, align 4
2840 // CHECK7-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2841 // CHECK7-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2842 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2843 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2844 // CHECK7-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2845 // CHECK7-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
2846 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2847 // CHECK7: omp.inner.for.cond:
2848 // CHECK7-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
2849 // CHECK7-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
2850 // CHECK7-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
2851 // CHECK7-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2852 // CHECK7: omp.inner.for.body:
2853 // CHECK7-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
2854 // CHECK7-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
2855 // CHECK7-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2856 // CHECK7-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !3
2857 // CHECK7-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
2858 // CHECK7-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !3
2859 // CHECK7-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP4]]
2860 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
2861 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2862 // CHECK7: omp.body.continue:
2863 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2864 // CHECK7: omp.inner.for.inc:
2865 // CHECK7-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
2866 // CHECK7-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
2867 // CHECK7-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
2868 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
2869 // CHECK7: omp.inner.for.end:
2870 // CHECK7-NEXT: store i32 123, i32* [[I]], align 4
2871 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB4]], align 4
2872 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_UB5]], align 4
2873 // CHECK7-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
2874 // CHECK7-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV6]], align 4
2875 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND8:%.*]]
2876 // CHECK7: omp.inner.for.cond8:
2877 // CHECK7-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
2878 // CHECK7-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !7
2879 // CHECK7-NEXT: [[CMP9:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2880 // CHECK7-NEXT: br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
2881 // CHECK7: omp.inner.for.body10:
2882 // CHECK7-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
2883 // CHECK7-NEXT: [[MUL11:%.*]] = mul nsw i32 [[TMP9]], 1
2884 // CHECK7-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
2885 // CHECK7-NEXT: store i32 [[ADD12]], i32* [[I7]], align 4, !llvm.access.group !7
2886 // CHECK7-NEXT: [[A13:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2887 // CHECK7-NEXT: [[TMP10:%.*]] = load i32, i32* [[I7]], align 4, !llvm.access.group !7
2888 // CHECK7-NEXT: [[ARRAYIDX14:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A13]], i32 0, i32 [[TMP10]]
2889 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX14]], align 4, !llvm.access.group !7
2890 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE15:%.*]]
2891 // CHECK7: omp.body.continue15:
2892 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC16:%.*]]
2893 // CHECK7: omp.inner.for.inc16:
2894 // CHECK7-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
2895 // CHECK7-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP11]], 1
2896 // CHECK7-NEXT: store i32 [[ADD17]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
2897 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP8:![0-9]+]]
2898 // CHECK7: omp.inner.for.end18:
2899 // CHECK7-NEXT: store i32 123, i32* [[I7]], align 4
2900 // CHECK7-NEXT: store i32 0, i32* [[DOTOMP_LB20]], align 4
2901 // CHECK7-NEXT: store i32 122, i32* [[DOTOMP_UB21]], align 4
2902 // CHECK7-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB20]], align 4
2903 // CHECK7-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV22]], align 4
2904 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND24:%.*]]
2905 // CHECK7: omp.inner.for.cond24:
2906 // CHECK7-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
2907 // CHECK7-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB21]], align 4, !llvm.access.group !10
2908 // CHECK7-NEXT: [[CMP25:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
2909 // CHECK7-NEXT: br i1 [[CMP25]], label [[OMP_INNER_FOR_BODY26:%.*]], label [[OMP_INNER_FOR_END34:%.*]]
2910 // CHECK7: omp.inner.for.body26:
2911 // CHECK7-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
2912 // CHECK7-NEXT: [[MUL27:%.*]] = mul nsw i32 [[TMP15]], 1
2913 // CHECK7-NEXT: [[ADD28:%.*]] = add nsw i32 0, [[MUL27]]
2914 // CHECK7-NEXT: store i32 [[ADD28]], i32* [[I23]], align 4, !llvm.access.group !10
2915 // CHECK7-NEXT: [[A29:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2916 // CHECK7-NEXT: [[TMP16:%.*]] = load i32, i32* [[I23]], align 4, !llvm.access.group !10
2917 // CHECK7-NEXT: [[ARRAYIDX30:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A29]], i32 0, i32 [[TMP16]]
2918 // CHECK7-NEXT: store i32 0, i32* [[ARRAYIDX30]], align 4, !llvm.access.group !10
2919 // CHECK7-NEXT: br label [[OMP_BODY_CONTINUE31:%.*]]
2920 // CHECK7: omp.body.continue31:
2921 // CHECK7-NEXT: br label [[OMP_INNER_FOR_INC32:%.*]]
2922 // CHECK7: omp.inner.for.inc32:
2923 // CHECK7-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
2924 // CHECK7-NEXT: [[ADD33:%.*]] = add nsw i32 [[TMP17]], 1
2925 // CHECK7-NEXT: store i32 [[ADD33]], i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
2926 // CHECK7-NEXT: br label [[OMP_INNER_FOR_COND24]], !llvm.loop [[LOOP11:![0-9]+]]
2927 // CHECK7: omp.inner.for.end34:
2928 // CHECK7-NEXT: store i32 123, i32* [[I23]], align 4
2929 // CHECK7-NEXT: [[A35:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2930 // CHECK7-NEXT: [[ARRAYIDX36:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A35]], i32 0, i32 0
2931 // CHECK7-NEXT: [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX36]], align 4
2932 // CHECK7-NEXT: ret i32 [[TMP18]]
2933 //
2934 //
2935 // CHECK8-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2936 // CHECK8-SAME: () #[[ATTR0:[0-9]+]] {
2937 // CHECK8-NEXT: entry:
2938 // CHECK8-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2939 // CHECK8-NEXT: [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
2940 // CHECK8-NEXT: ret i32 [[CALL]]
2941 //
2942 //
2943 // CHECK8-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2944 // CHECK8-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2945 // CHECK8-NEXT: entry:
2946 // CHECK8-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2947 // CHECK8-NEXT: [[TMP:%.*]] = alloca i32, align 4
2948 // CHECK8-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2949 // CHECK8-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2950 // CHECK8-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2951 // CHECK8-NEXT: [[I:%.*]] = alloca i32, align 4
2952 // CHECK8-NEXT: [[_TMP3:%.*]] = alloca i32, align 4
2953 // CHECK8-NEXT: [[DOTOMP_LB4:%.*]] = alloca i32, align 4
2954 // CHECK8-NEXT: [[DOTOMP_UB5:%.*]] = alloca i32, align 4
2955 // CHECK8-NEXT: [[DOTOMP_IV6:%.*]] = alloca i32, align 4
2956 // CHECK8-NEXT: [[I7:%.*]] = alloca i32, align 4
2957 // CHECK8-NEXT: [[_TMP19:%.*]] = alloca i32, align 4
2958 // CHECK8-NEXT: [[DOTOMP_LB20:%.*]] = alloca i32, align 4
2959 // CHECK8-NEXT: [[DOTOMP_UB21:%.*]] = alloca i32, align 4
2960 // CHECK8-NEXT: [[DOTOMP_IV22:%.*]] = alloca i32, align 4
2961 // CHECK8-NEXT: [[I23:%.*]] = alloca i32, align 4
2962 // CHECK8-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2963 // CHECK8-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2964 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2965 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2966 // CHECK8-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2967 // CHECK8-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
2968 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2969 // CHECK8: omp.inner.for.cond:
2970 // CHECK8-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
2971 // CHECK8-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
2972 // CHECK8-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
2973 // CHECK8-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2974 // CHECK8: omp.inner.for.body:
2975 // CHECK8-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
2976 // CHECK8-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
2977 // CHECK8-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2978 // CHECK8-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !3
2979 // CHECK8-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
2980 // CHECK8-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !3
2981 // CHECK8-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP4]]
2982 // CHECK8-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
2983 // CHECK8-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2984 // CHECK8: omp.body.continue:
2985 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2986 // CHECK8: omp.inner.for.inc:
2987 // CHECK8-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
2988 // CHECK8-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP5]], 1
2989 // CHECK8-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
2990 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
2991 // CHECK8: omp.inner.for.end:
2992 // CHECK8-NEXT: store i32 123, i32* [[I]], align 4
2993 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_LB4]], align 4
2994 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_UB5]], align 4
2995 // CHECK8-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB4]], align 4
2996 // CHECK8-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV6]], align 4
2997 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND8:%.*]]
2998 // CHECK8: omp.inner.for.cond8:
2999 // CHECK8-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
3000 // CHECK8-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB5]], align 4, !llvm.access.group !7
3001 // CHECK8-NEXT: [[CMP9:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3002 // CHECK8-NEXT: br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
3003 // CHECK8: omp.inner.for.body10:
3004 // CHECK8-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
3005 // CHECK8-NEXT: [[MUL11:%.*]] = mul nsw i32 [[TMP9]], 1
3006 // CHECK8-NEXT: [[ADD12:%.*]] = add nsw i32 0, [[MUL11]]
3007 // CHECK8-NEXT: store i32 [[ADD12]], i32* [[I7]], align 4, !llvm.access.group !7
3008 // CHECK8-NEXT: [[A13:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3009 // CHECK8-NEXT: [[TMP10:%.*]] = load i32, i32* [[I7]], align 4, !llvm.access.group !7
3010 // CHECK8-NEXT: [[ARRAYIDX14:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A13]], i32 0, i32 [[TMP10]]
3011 // CHECK8-NEXT: store i32 0, i32* [[ARRAYIDX14]], align 4, !llvm.access.group !7
3012 // CHECK8-NEXT: br label [[OMP_BODY_CONTINUE15:%.*]]
3013 // CHECK8: omp.body.continue15:
3014 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC16:%.*]]
3015 // CHECK8: omp.inner.for.inc16:
3016 // CHECK8-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
3017 // CHECK8-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP11]], 1
3018 // CHECK8-NEXT: store i32 [[ADD17]], i32* [[DOTOMP_IV6]], align 4, !llvm.access.group !7
3019 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP8:![0-9]+]]
3020 // CHECK8: omp.inner.for.end18:
3021 // CHECK8-NEXT: store i32 123, i32* [[I7]], align 4
3022 // CHECK8-NEXT: store i32 0, i32* [[DOTOMP_LB20]], align 4
3023 // CHECK8-NEXT: store i32 122, i32* [[DOTOMP_UB21]], align 4
3024 // CHECK8-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB20]], align 4
3025 // CHECK8-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV22]], align 4
3026 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND24:%.*]]
3027 // CHECK8: omp.inner.for.cond24:
3028 // CHECK8-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
3029 // CHECK8-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB21]], align 4, !llvm.access.group !10
3030 // CHECK8-NEXT: [[CMP25:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
3031 // CHECK8-NEXT: br i1 [[CMP25]], label [[OMP_INNER_FOR_BODY26:%.*]], label [[OMP_INNER_FOR_END34:%.*]]
3032 // CHECK8: omp.inner.for.body26:
3033 // CHECK8-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
3034 // CHECK8-NEXT: [[MUL27:%.*]] = mul nsw i32 [[TMP15]], 1
3035 // CHECK8-NEXT: [[ADD28:%.*]] = add nsw i32 0, [[MUL27]]
3036 // CHECK8-NEXT: store i32 [[ADD28]], i32* [[I23]], align 4, !llvm.access.group !10
3037 // CHECK8-NEXT: [[A29:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3038 // CHECK8-NEXT: [[TMP16:%.*]] = load i32, i32* [[I23]], align 4, !llvm.access.group !10
3039 // CHECK8-NEXT: [[ARRAYIDX30:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A29]], i32 0, i32 [[TMP16]]
3040 // CHECK8-NEXT: store i32 0, i32* [[ARRAYIDX30]], align 4, !llvm.access.group !10
3041 // CHECK8-NEXT: br label [[OMP_BODY_CONTINUE31:%.*]]
3042 // CHECK8: omp.body.continue31:
3043 // CHECK8-NEXT: br label [[OMP_INNER_FOR_INC32:%.*]]
3044 // CHECK8: omp.inner.for.inc32:
3045 // CHECK8-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
3046 // CHECK8-NEXT: [[ADD33:%.*]] = add nsw i32 [[TMP17]], 1
3047 // CHECK8-NEXT: store i32 [[ADD33]], i32* [[DOTOMP_IV22]], align 4, !llvm.access.group !10
3048 // CHECK8-NEXT: br label [[OMP_INNER_FOR_COND24]], !llvm.loop [[LOOP11:![0-9]+]]
3049 // CHECK8: omp.inner.for.end34:
3050 // CHECK8-NEXT: store i32 123, i32* [[I23]], align 4
3051 // CHECK8-NEXT: [[A35:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
3052 // CHECK8-NEXT: [[ARRAYIDX36:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A35]], i32 0, i32 0
3053 // CHECK8-NEXT: [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX36]], align 4
3054 // CHECK8-NEXT: ret i32 [[TMP18]]
3055 //
3056 //
3057 // CHECK9-LABEL: define {{[^@]+}}@main
3058 // CHECK9-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
3059 // CHECK9-NEXT: entry:
3060 // CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
3061 // CHECK9-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3062 // CHECK9-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
3063 // CHECK9-NEXT: [[N:%.*]] = alloca i32, align 4
3064 // CHECK9-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
3065 // CHECK9-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
3066 // CHECK9-NEXT: [[M:%.*]] = alloca i32, align 4
3067 // CHECK9-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
3068 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
3069 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
3070 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
3071 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
3072 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
3073 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3074 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3075 // CHECK9-NEXT: [[N_CASTED3:%.*]] = alloca i64, align 8
3076 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
3077 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
3078 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
3079 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
3080 // CHECK9-NEXT: [[_TMP9:%.*]] = alloca i32, align 4
3081 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
3082 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
3083 // CHECK9-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
3084 // CHECK9-NEXT: [[N_CASTED19:%.*]] = alloca i64, align 8
3085 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [4 x i8*], align 8
3086 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS22:%.*]] = alloca [4 x i8*], align 8
3087 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [4 x i8*], align 8
3088 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES24:%.*]] = alloca [4 x i64], align 8
3089 // CHECK9-NEXT: [[_TMP25:%.*]] = alloca i32, align 4
3090 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
3091 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
3092 // CHECK9-NEXT: store i32 0, i32* [[RETVAL]], align 4
3093 // CHECK9-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3094 // CHECK9-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
3095 // CHECK9-NEXT: store i32 100, i32* [[N]], align 4
3096 // CHECK9-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
3097 // CHECK9-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
3098 // CHECK9-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
3099 // CHECK9-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
3100 // CHECK9-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
3101 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
3102 // CHECK9-NEXT: store i32 10, i32* [[M]], align 4
3103 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
3104 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
3105 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
3106 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
3107 // CHECK9-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
3108 // CHECK9-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3109 // CHECK9-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
3110 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP7]], align 8
3111 // CHECK9-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3112 // CHECK9-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
3113 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP9]], align 8
3114 // CHECK9-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3115 // CHECK9-NEXT: store i64 4, i64* [[TMP10]], align 8
3116 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3117 // CHECK9-NEXT: store i8* null, i8** [[TMP11]], align 8
3118 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3119 // CHECK9-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
3120 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
3121 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3122 // CHECK9-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
3123 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
3124 // CHECK9-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
3125 // CHECK9-NEXT: store i64 8, i64* [[TMP16]], align 8
3126 // CHECK9-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3127 // CHECK9-NEXT: store i8* null, i8** [[TMP17]], align 8
3128 // CHECK9-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3129 // CHECK9-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
3130 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 8
3131 // CHECK9-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3132 // CHECK9-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
3133 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 8
3134 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
3135 // CHECK9-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 8
3136 // CHECK9-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3137 // CHECK9-NEXT: store i8* null, i8** [[TMP23]], align 8
3138 // CHECK9-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3139 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3140 // CHECK9-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3141 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
3142 // CHECK9-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
3143 // CHECK9-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3144 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
3145 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3146 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3147 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3148 // CHECK9-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3149 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
3150 // CHECK9-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
3151 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
3152 // CHECK9-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108.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)
3153 // CHECK9-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
3154 // CHECK9-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3155 // CHECK9: omp_offload.failed:
3156 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
3157 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
3158 // CHECK9: omp_offload.cont:
3159 // CHECK9-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
3160 // CHECK9-NEXT: [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
3161 // CHECK9-NEXT: store i32 [[TMP33]], i32* [[CONV4]], align 4
3162 // CHECK9-NEXT: [[TMP34:%.*]] = load i64, i64* [[N_CASTED3]], align 8
3163 // CHECK9-NEXT: [[TMP35:%.*]] = mul nuw i64 [[TMP1]], 4
3164 // CHECK9-NEXT: [[TMP36:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
3165 // CHECK9-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
3166 // CHECK9-NEXT: store i64 [[TMP34]], i64* [[TMP37]], align 8
3167 // CHECK9-NEXT: [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
3168 // CHECK9-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
3169 // CHECK9-NEXT: store i64 [[TMP34]], i64* [[TMP39]], align 8
3170 // CHECK9-NEXT: [[TMP40:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
3171 // CHECK9-NEXT: store i64 4, i64* [[TMP40]], align 8
3172 // CHECK9-NEXT: [[TMP41:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
3173 // CHECK9-NEXT: store i8* null, i8** [[TMP41]], align 8
3174 // CHECK9-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
3175 // CHECK9-NEXT: [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
3176 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP43]], align 8
3177 // CHECK9-NEXT: [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
3178 // CHECK9-NEXT: [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i64*
3179 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP45]], align 8
3180 // CHECK9-NEXT: [[TMP46:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 1
3181 // CHECK9-NEXT: store i64 8, i64* [[TMP46]], align 8
3182 // CHECK9-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
3183 // CHECK9-NEXT: store i8* null, i8** [[TMP47]], align 8
3184 // CHECK9-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
3185 // CHECK9-NEXT: [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32**
3186 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP49]], align 8
3187 // CHECK9-NEXT: [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
3188 // CHECK9-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i32**
3189 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP51]], align 8
3190 // CHECK9-NEXT: [[TMP52:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
3191 // CHECK9-NEXT: store i64 [[TMP35]], i64* [[TMP52]], align 8
3192 // CHECK9-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
3193 // CHECK9-NEXT: store i8* null, i8** [[TMP53]], align 8
3194 // CHECK9-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
3195 // CHECK9-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
3196 // CHECK9-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
3197 // CHECK9-NEXT: [[TMP57:%.*]] = load i32, i32* [[N]], align 4
3198 // CHECK9-NEXT: store i32 [[TMP57]], i32* [[DOTCAPTURE_EXPR_10]], align 4
3199 // CHECK9-NEXT: [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
3200 // CHECK9-NEXT: [[SUB12:%.*]] = sub nsw i32 [[TMP58]], 0
3201 // CHECK9-NEXT: [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
3202 // CHECK9-NEXT: [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
3203 // CHECK9-NEXT: store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
3204 // CHECK9-NEXT: [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
3205 // CHECK9-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP59]], 1
3206 // CHECK9-NEXT: [[TMP60:%.*]] = zext i32 [[ADD15]] to i64
3207 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP60]])
3208 // CHECK9-NEXT: [[TMP61:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113.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)
3209 // CHECK9-NEXT: [[TMP62:%.*]] = icmp ne i32 [[TMP61]], 0
3210 // CHECK9-NEXT: br i1 [[TMP62]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
3211 // CHECK9: omp_offload.failed16:
3212 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113(i64 [[TMP34]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
3213 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT17]]
3214 // CHECK9: omp_offload.cont17:
3215 // CHECK9-NEXT: [[TMP63:%.*]] = load i32, i32* [[M]], align 4
3216 // CHECK9-NEXT: [[CONV18:%.*]] = bitcast i64* [[M_CASTED]] to i32*
3217 // CHECK9-NEXT: store i32 [[TMP63]], i32* [[CONV18]], align 4
3218 // CHECK9-NEXT: [[TMP64:%.*]] = load i64, i64* [[M_CASTED]], align 8
3219 // CHECK9-NEXT: [[TMP65:%.*]] = load i32, i32* [[N]], align 4
3220 // CHECK9-NEXT: [[CONV20:%.*]] = bitcast i64* [[N_CASTED19]] to i32*
3221 // CHECK9-NEXT: store i32 [[TMP65]], i32* [[CONV20]], align 4
3222 // CHECK9-NEXT: [[TMP66:%.*]] = load i64, i64* [[N_CASTED19]], align 8
3223 // CHECK9-NEXT: [[TMP67:%.*]] = mul nuw i64 [[TMP1]], 4
3224 // CHECK9-NEXT: [[TMP68:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
3225 // CHECK9-NEXT: [[TMP69:%.*]] = bitcast i8** [[TMP68]] to i64*
3226 // CHECK9-NEXT: store i64 [[TMP64]], i64* [[TMP69]], align 8
3227 // CHECK9-NEXT: [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
3228 // CHECK9-NEXT: [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i64*
3229 // CHECK9-NEXT: store i64 [[TMP64]], i64* [[TMP71]], align 8
3230 // CHECK9-NEXT: [[TMP72:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
3231 // CHECK9-NEXT: store i64 4, i64* [[TMP72]], align 8
3232 // CHECK9-NEXT: [[TMP73:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
3233 // CHECK9-NEXT: store i8* null, i8** [[TMP73]], align 8
3234 // CHECK9-NEXT: [[TMP74:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
3235 // CHECK9-NEXT: [[TMP75:%.*]] = bitcast i8** [[TMP74]] to i64*
3236 // CHECK9-NEXT: store i64 [[TMP66]], i64* [[TMP75]], align 8
3237 // CHECK9-NEXT: [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
3238 // CHECK9-NEXT: [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i64*
3239 // CHECK9-NEXT: store i64 [[TMP66]], i64* [[TMP77]], align 8
3240 // CHECK9-NEXT: [[TMP78:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 1
3241 // CHECK9-NEXT: store i64 4, i64* [[TMP78]], align 8
3242 // CHECK9-NEXT: [[TMP79:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
3243 // CHECK9-NEXT: store i8* null, i8** [[TMP79]], align 8
3244 // CHECK9-NEXT: [[TMP80:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 2
3245 // CHECK9-NEXT: [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
3246 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP81]], align 8
3247 // CHECK9-NEXT: [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 2
3248 // CHECK9-NEXT: [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
3249 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP83]], align 8
3250 // CHECK9-NEXT: [[TMP84:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 2
3251 // CHECK9-NEXT: store i64 8, i64* [[TMP84]], align 8
3252 // CHECK9-NEXT: [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 2
3253 // CHECK9-NEXT: store i8* null, i8** [[TMP85]], align 8
3254 // CHECK9-NEXT: [[TMP86:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 3
3255 // CHECK9-NEXT: [[TMP87:%.*]] = bitcast i8** [[TMP86]] to i32**
3256 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP87]], align 8
3257 // CHECK9-NEXT: [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 3
3258 // CHECK9-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
3259 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP89]], align 8
3260 // CHECK9-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 3
3261 // CHECK9-NEXT: store i64 [[TMP67]], i64* [[TMP90]], align 8
3262 // CHECK9-NEXT: [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 3
3263 // CHECK9-NEXT: store i8* null, i8** [[TMP91]], align 8
3264 // CHECK9-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
3265 // CHECK9-NEXT: [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
3266 // CHECK9-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
3267 // CHECK9-NEXT: [[TMP95:%.*]] = load i32, i32* [[N]], align 4
3268 // CHECK9-NEXT: store i32 [[TMP95]], i32* [[DOTCAPTURE_EXPR_26]], align 4
3269 // CHECK9-NEXT: [[TMP96:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_26]], align 4
3270 // CHECK9-NEXT: [[SUB28:%.*]] = sub nsw i32 [[TMP96]], 0
3271 // CHECK9-NEXT: [[DIV29:%.*]] = sdiv i32 [[SUB28]], 1
3272 // CHECK9-NEXT: [[SUB30:%.*]] = sub nsw i32 [[DIV29]], 1
3273 // CHECK9-NEXT: store i32 [[SUB30]], i32* [[DOTCAPTURE_EXPR_27]], align 4
3274 // CHECK9-NEXT: [[TMP97:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
3275 // CHECK9-NEXT: [[ADD31:%.*]] = add nsw i32 [[TMP97]], 1
3276 // CHECK9-NEXT: [[TMP98:%.*]] = zext i32 [[ADD31]] to i64
3277 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP98]])
3278 // CHECK9-NEXT: [[TMP99:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118.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)
3279 // CHECK9-NEXT: [[TMP100:%.*]] = icmp ne i32 [[TMP99]], 0
3280 // CHECK9-NEXT: br i1 [[TMP100]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
3281 // CHECK9: omp_offload.failed32:
3282 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118(i64 [[TMP64]], i64 [[TMP66]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
3283 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT33]]
3284 // CHECK9: omp_offload.cont33:
3285 // CHECK9-NEXT: [[TMP101:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3286 // CHECK9-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP101]])
3287 // CHECK9-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
3288 // CHECK9-NEXT: [[TMP102:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
3289 // CHECK9-NEXT: call void @llvm.stackrestore(i8* [[TMP102]])
3290 // CHECK9-NEXT: [[TMP103:%.*]] = load i32, i32* [[RETVAL]], align 4
3291 // CHECK9-NEXT: ret i32 [[TMP103]]
3292 //
3293 //
3294 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108
3295 // CHECK9-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
3296 // CHECK9-NEXT: entry:
3297 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
3298 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3299 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3300 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
3301 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3302 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3303 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
3304 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3305 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3306 // CHECK9-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]])
3307 // CHECK9-NEXT: ret void
3308 //
3309 //
3310 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
3311 // CHECK9-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]] {
3312 // CHECK9-NEXT: entry:
3313 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3314 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3315 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
3316 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3317 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3318 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3319 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
3320 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3321 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3322 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
3323 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3324 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3325 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3326 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3327 // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4
3328 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3329 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3330 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
3331 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3332 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3333 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
3334 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3335 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3336 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3337 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3338 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3339 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3340 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3341 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3342 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3343 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
3344 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3345 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3346 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3347 // CHECK9: omp.precond.then:
3348 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3349 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3350 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
3351 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3352 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3353 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3354 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3355 // CHECK9-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)
3356 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3357 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3358 // CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
3359 // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3360 // CHECK9: cond.true:
3361 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3362 // CHECK9-NEXT: br label [[COND_END:%.*]]
3363 // CHECK9: cond.false:
3364 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3365 // CHECK9-NEXT: br label [[COND_END]]
3366 // CHECK9: cond.end:
3367 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3368 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3369 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3370 // CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
3371 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3372 // CHECK9: omp.inner.for.cond:
3373 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3374 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3375 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
3376 // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3377 // CHECK9: omp.inner.for.body:
3378 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3379 // CHECK9-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
3380 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3381 // CHECK9-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
3382 // CHECK9-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]])
3383 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3384 // CHECK9: omp.inner.for.inc:
3385 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3386 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3387 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
3388 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3389 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP8:![0-9]+]]
3390 // CHECK9: omp.inner.for.end:
3391 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3392 // CHECK9: omp.loop.exit:
3393 // CHECK9-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3394 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
3395 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
3396 // CHECK9-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3397 // CHECK9-NEXT: [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
3398 // CHECK9-NEXT: br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3399 // CHECK9: .omp.final.then:
3400 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3401 // CHECK9-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
3402 // CHECK9-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
3403 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
3404 // CHECK9-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
3405 // CHECK9-NEXT: store i32 [[ADD8]], i32* [[I3]], align 4
3406 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
3407 // CHECK9: .omp.final.done:
3408 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
3409 // CHECK9: omp.precond.end:
3410 // CHECK9-NEXT: ret void
3411 //
3412 //
3413 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
3414 // CHECK9-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]] {
3415 // CHECK9-NEXT: entry:
3416 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3417 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3418 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3419 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3420 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
3421 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3422 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3423 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3424 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
3425 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3426 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3427 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
3428 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3429 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3430 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3431 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3432 // CHECK9-NEXT: [[I4:%.*]] = alloca i32, align 4
3433 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3434 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3435 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3436 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3437 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
3438 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3439 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3440 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
3441 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3442 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3443 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3444 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3445 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3446 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3447 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3448 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3449 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3450 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
3451 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3452 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3453 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3454 // CHECK9: omp.precond.then:
3455 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3456 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3457 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
3458 // CHECK9-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3459 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
3460 // CHECK9-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3461 // CHECK9-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
3462 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3463 // CHECK9-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
3464 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3465 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3466 // CHECK9-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3467 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3468 // CHECK9-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)
3469 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3470 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3471 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
3472 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3473 // CHECK9: cond.true:
3474 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3475 // CHECK9-NEXT: br label [[COND_END:%.*]]
3476 // CHECK9: cond.false:
3477 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3478 // CHECK9-NEXT: br label [[COND_END]]
3479 // CHECK9: cond.end:
3480 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
3481 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3482 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3483 // CHECK9-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
3484 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3485 // CHECK9: omp.inner.for.cond:
3486 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3487 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3488 // CHECK9-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
3489 // CHECK9-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3490 // CHECK9: omp.inner.for.body:
3491 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3492 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
3493 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3494 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
3495 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
3496 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
3497 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
3498 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3499 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3500 // CHECK9: omp.body.continue:
3501 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3502 // CHECK9: omp.inner.for.inc:
3503 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3504 // CHECK9-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
3505 // CHECK9-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
3506 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
3507 // CHECK9: omp.inner.for.end:
3508 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3509 // CHECK9: omp.loop.exit:
3510 // CHECK9-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3511 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
3512 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
3513 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3514 // CHECK9-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
3515 // CHECK9-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3516 // CHECK9: .omp.final.then:
3517 // CHECK9-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3518 // CHECK9-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
3519 // CHECK9-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
3520 // CHECK9-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
3521 // CHECK9-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
3522 // CHECK9-NEXT: store i32 [[ADD11]], i32* [[I4]], align 4
3523 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
3524 // CHECK9: .omp.final.done:
3525 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
3526 // CHECK9: omp.precond.end:
3527 // CHECK9-NEXT: ret void
3528 //
3529 //
3530 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113
3531 // CHECK9-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
3532 // CHECK9-NEXT: entry:
3533 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
3534 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3535 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3536 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
3537 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3538 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3539 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
3540 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3541 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3542 // CHECK9-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]])
3543 // CHECK9-NEXT: ret void
3544 //
3545 //
3546 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
3547 // CHECK9-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]] {
3548 // CHECK9-NEXT: entry:
3549 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3550 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3551 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
3552 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3553 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3554 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3555 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
3556 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3557 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3558 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
3559 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3560 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3561 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3562 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3563 // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4
3564 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3565 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3566 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
3567 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3568 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3569 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
3570 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3571 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3572 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3573 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3574 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3575 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3576 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3577 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3578 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3579 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
3580 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3581 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3582 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3583 // CHECK9: omp.precond.then:
3584 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3585 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3586 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
3587 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3588 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3589 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3590 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3591 // CHECK9-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)
3592 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3593 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3594 // CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
3595 // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3596 // CHECK9: cond.true:
3597 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3598 // CHECK9-NEXT: br label [[COND_END:%.*]]
3599 // CHECK9: cond.false:
3600 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3601 // CHECK9-NEXT: br label [[COND_END]]
3602 // CHECK9: cond.end:
3603 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3604 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3605 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3606 // CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
3607 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3608 // CHECK9: omp.inner.for.cond:
3609 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3610 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3611 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
3612 // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3613 // CHECK9: omp.inner.for.body:
3614 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3615 // CHECK9-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
3616 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3617 // CHECK9-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
3618 // CHECK9-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]])
3619 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3620 // CHECK9: omp.inner.for.inc:
3621 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3622 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3623 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
3624 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3625 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
3626 // CHECK9: omp.inner.for.end:
3627 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3628 // CHECK9: omp.loop.exit:
3629 // CHECK9-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3630 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
3631 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
3632 // CHECK9-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3633 // CHECK9-NEXT: [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
3634 // CHECK9-NEXT: br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3635 // CHECK9: .omp.final.then:
3636 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3637 // CHECK9-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
3638 // CHECK9-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
3639 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
3640 // CHECK9-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
3641 // CHECK9-NEXT: store i32 [[ADD8]], i32* [[I3]], align 4
3642 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
3643 // CHECK9: .omp.final.done:
3644 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
3645 // CHECK9: omp.precond.end:
3646 // CHECK9-NEXT: ret void
3647 //
3648 //
3649 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
3650 // CHECK9-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]] {
3651 // CHECK9-NEXT: entry:
3652 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3653 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3654 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3655 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3656 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
3657 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3658 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3659 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3660 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
3661 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3662 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3663 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
3664 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3665 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3666 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3667 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3668 // CHECK9-NEXT: [[I4:%.*]] = alloca i32, align 4
3669 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3670 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3671 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3672 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3673 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
3674 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3675 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3676 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
3677 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3678 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3679 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3680 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3681 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3682 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3683 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3684 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3685 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3686 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
3687 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3688 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3689 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3690 // CHECK9: omp.precond.then:
3691 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3692 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3693 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
3694 // CHECK9-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3695 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
3696 // CHECK9-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3697 // CHECK9-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
3698 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3699 // CHECK9-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
3700 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3701 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3702 // CHECK9-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3703 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3704 // CHECK9-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)
3705 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3706 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3707 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
3708 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3709 // CHECK9: cond.true:
3710 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3711 // CHECK9-NEXT: br label [[COND_END:%.*]]
3712 // CHECK9: cond.false:
3713 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3714 // CHECK9-NEXT: br label [[COND_END]]
3715 // CHECK9: cond.end:
3716 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
3717 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3718 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3719 // CHECK9-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
3720 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3721 // CHECK9: omp.inner.for.cond:
3722 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3723 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3724 // CHECK9-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
3725 // CHECK9-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3726 // CHECK9: omp.inner.for.body:
3727 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3728 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
3729 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3730 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
3731 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
3732 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
3733 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
3734 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3735 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3736 // CHECK9: omp.body.continue:
3737 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3738 // CHECK9: omp.inner.for.inc:
3739 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3740 // CHECK9-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
3741 // CHECK9-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
3742 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
3743 // CHECK9: omp.inner.for.end:
3744 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3745 // CHECK9: omp.loop.exit:
3746 // CHECK9-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3747 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
3748 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
3749 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3750 // CHECK9-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
3751 // CHECK9-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3752 // CHECK9: .omp.final.then:
3753 // CHECK9-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3754 // CHECK9-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
3755 // CHECK9-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
3756 // CHECK9-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
3757 // CHECK9-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
3758 // CHECK9-NEXT: store i32 [[ADD11]], i32* [[I4]], align 4
3759 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
3760 // CHECK9: .omp.final.done:
3761 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
3762 // CHECK9: omp.precond.end:
3763 // CHECK9-NEXT: ret void
3764 //
3765 //
3766 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118
3767 // CHECK9-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
3768 // CHECK9-NEXT: entry:
3769 // CHECK9-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
3770 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
3771 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3772 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3773 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3774 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
3775 // CHECK9-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
3776 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
3777 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3778 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3779 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
3780 // CHECK9-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
3781 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3782 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3783 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
3784 // CHECK9-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
3785 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3786 // CHECK9-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
3787 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
3788 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
3789 // CHECK9-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]])
3790 // CHECK9-NEXT: ret void
3791 //
3792 //
3793 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
3794 // CHECK9-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]] {
3795 // CHECK9-NEXT: entry:
3796 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3797 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3798 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
3799 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3800 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3801 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3802 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3803 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
3804 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3805 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
3806 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
3807 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3808 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3809 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3810 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3811 // CHECK9-NEXT: [[I4:%.*]] = alloca i32, align 4
3812 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
3813 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3814 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3815 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
3816 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3817 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3818 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3819 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
3820 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3821 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3822 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
3823 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3824 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3825 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3826 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3827 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3828 // CHECK9-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
3829 // CHECK9-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
3830 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
3831 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3832 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3833 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3834 // CHECK9: omp.precond.then:
3835 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3836 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3837 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
3838 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3839 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3840 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[CONV]], align 8
3841 // CHECK9-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3842 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
3843 // CHECK9-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]])
3844 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3845 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3846 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
3847 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3848 // CHECK9: cond.true:
3849 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3850 // CHECK9-NEXT: br label [[COND_END:%.*]]
3851 // CHECK9: cond.false:
3852 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3853 // CHECK9-NEXT: br label [[COND_END]]
3854 // CHECK9: cond.end:
3855 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
3856 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3857 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3858 // CHECK9-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
3859 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3860 // CHECK9: omp.inner.for.cond:
3861 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3862 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3863 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
3864 // CHECK9-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
3865 // CHECK9-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3866 // CHECK9: omp.inner.for.body:
3867 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3868 // CHECK9-NEXT: [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
3869 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3870 // CHECK9-NEXT: [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
3871 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[CONV]], align 8
3872 // CHECK9-NEXT: [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
3873 // CHECK9-NEXT: store i32 [[TMP21]], i32* [[CONV7]], align 4
3874 // CHECK9-NEXT: [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
3875 // CHECK9-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]])
3876 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3877 // CHECK9: omp.inner.for.inc:
3878 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3879 // CHECK9-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3880 // CHECK9-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
3881 // CHECK9-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
3882 // CHECK9-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3883 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3884 // CHECK9-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
3885 // CHECK9-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4
3886 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3887 // CHECK9-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3888 // CHECK9-NEXT: [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
3889 // CHECK9-NEXT: store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4
3890 // CHECK9-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3891 // CHECK9-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3892 // CHECK9-NEXT: [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
3893 // CHECK9-NEXT: br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
3894 // CHECK9: cond.true12:
3895 // CHECK9-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3896 // CHECK9-NEXT: br label [[COND_END14:%.*]]
3897 // CHECK9: cond.false13:
3898 // CHECK9-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3899 // CHECK9-NEXT: br label [[COND_END14]]
3900 // CHECK9: cond.end14:
3901 // CHECK9-NEXT: [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
3902 // CHECK9-NEXT: store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4
3903 // CHECK9-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3904 // CHECK9-NEXT: store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4
3905 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
3906 // CHECK9: omp.inner.for.end:
3907 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3908 // CHECK9: omp.loop.exit:
3909 // CHECK9-NEXT: [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3910 // CHECK9-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
3911 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
3912 // CHECK9-NEXT: [[TMP36:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3913 // CHECK9-NEXT: [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
3914 // CHECK9-NEXT: br i1 [[TMP37]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3915 // CHECK9: .omp.final.then:
3916 // CHECK9-NEXT: [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3917 // CHECK9-NEXT: [[SUB16:%.*]] = sub nsw i32 [[TMP38]], 0
3918 // CHECK9-NEXT: [[DIV17:%.*]] = sdiv i32 [[SUB16]], 1
3919 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV17]], 1
3920 // CHECK9-NEXT: [[ADD18:%.*]] = add nsw i32 0, [[MUL]]
3921 // CHECK9-NEXT: store i32 [[ADD18]], i32* [[I4]], align 4
3922 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
3923 // CHECK9: .omp.final.done:
3924 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
3925 // CHECK9: omp.precond.end:
3926 // CHECK9-NEXT: ret void
3927 //
3928 //
3929 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..6
3930 // CHECK9-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]] {
3931 // CHECK9-NEXT: entry:
3932 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3933 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3934 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3935 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3936 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
3937 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3938 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3939 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3940 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3941 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
3942 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3943 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
3944 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
3945 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3946 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3947 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3948 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3949 // CHECK9-NEXT: [[I6:%.*]] = alloca i32, align 4
3950 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3951 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3952 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3953 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3954 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
3955 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3956 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3957 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3958 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
3959 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3960 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3961 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
3962 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3963 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3964 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3965 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3966 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3967 // CHECK9-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
3968 // CHECK9-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
3969 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
3970 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3971 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3972 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3973 // CHECK9: omp.precond.then:
3974 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3975 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3976 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
3977 // CHECK9-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3978 // CHECK9-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
3979 // CHECK9-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3980 // CHECK9-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
3981 // CHECK9-NEXT: store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
3982 // CHECK9-NEXT: store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
3983 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3984 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3985 // CHECK9-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3986 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3987 // CHECK9-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)
3988 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3989 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3990 // CHECK9-NEXT: [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
3991 // CHECK9-NEXT: br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3992 // CHECK9: cond.true:
3993 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3994 // CHECK9-NEXT: br label [[COND_END:%.*]]
3995 // CHECK9: cond.false:
3996 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3997 // CHECK9-NEXT: br label [[COND_END]]
3998 // CHECK9: cond.end:
3999 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
4000 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4001 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4002 // CHECK9-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
4003 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4004 // CHECK9: omp.inner.for.cond:
4005 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4006 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4007 // CHECK9-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
4008 // CHECK9-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4009 // CHECK9: omp.inner.for.body:
4010 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4011 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
4012 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4013 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I6]], align 4
4014 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[I6]], align 4
4015 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
4016 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
4017 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4018 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4019 // CHECK9: omp.body.continue:
4020 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4021 // CHECK9: omp.inner.for.inc:
4022 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4023 // CHECK9-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
4024 // CHECK9-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
4025 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
4026 // CHECK9: omp.inner.for.end:
4027 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4028 // CHECK9: omp.loop.exit:
4029 // CHECK9-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4030 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
4031 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
4032 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4033 // CHECK9-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
4034 // CHECK9-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4035 // CHECK9: .omp.final.then:
4036 // CHECK9-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4037 // CHECK9-NEXT: [[SUB10:%.*]] = sub nsw i32 [[TMP25]], 0
4038 // CHECK9-NEXT: [[DIV11:%.*]] = sdiv i32 [[SUB10]], 1
4039 // CHECK9-NEXT: [[MUL12:%.*]] = mul nsw i32 [[DIV11]], 1
4040 // CHECK9-NEXT: [[ADD13:%.*]] = add nsw i32 0, [[MUL12]]
4041 // CHECK9-NEXT: store i32 [[ADD13]], i32* [[I6]], align 4
4042 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
4043 // CHECK9: .omp.final.done:
4044 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
4045 // CHECK9: omp.precond.end:
4046 // CHECK9-NEXT: ret void
4047 //
4048 //
4049 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
4050 // CHECK9-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
4051 // CHECK9-NEXT: entry:
4052 // CHECK9-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
4053 // CHECK9-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
4054 // CHECK9-NEXT: [[M:%.*]] = alloca i32, align 4
4055 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
4056 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
4057 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
4058 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
4059 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
4060 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
4061 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
4062 // CHECK9-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
4063 // CHECK9-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
4064 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
4065 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
4066 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
4067 // CHECK9-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
4068 // CHECK9-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4069 // CHECK9-NEXT: store i32 10, i32* [[M]], align 4
4070 // CHECK9-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4071 // CHECK9-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
4072 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
4073 // CHECK9-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4074 // CHECK9-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
4075 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
4076 // CHECK9-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4077 // CHECK9-NEXT: store i8* null, i8** [[TMP4]], align 8
4078 // CHECK9-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4079 // CHECK9-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4080 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
4081 // CHECK9-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86.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.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4082 // CHECK9-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
4083 // CHECK9-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4084 // CHECK9: omp_offload.failed:
4085 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86([10 x i32]* [[A]]) #[[ATTR3]]
4086 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
4087 // CHECK9: omp_offload.cont:
4088 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
4089 // CHECK9-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
4090 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 8
4091 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
4092 // CHECK9-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
4093 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 8
4094 // CHECK9-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
4095 // CHECK9-NEXT: store i8* null, i8** [[TMP13]], align 8
4096 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
4097 // CHECK9-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
4098 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
4099 // CHECK9-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.13, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.14, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4100 // CHECK9-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
4101 // CHECK9-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
4102 // CHECK9: omp_offload.failed5:
4103 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91([10 x i32]* [[A]]) #[[ATTR3]]
4104 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT6]]
4105 // CHECK9: omp_offload.cont6:
4106 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[M]], align 4
4107 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
4108 // CHECK9-NEXT: store i32 [[TMP18]], i32* [[CONV]], align 4
4109 // CHECK9-NEXT: [[TMP19:%.*]] = load i64, i64* [[M_CASTED]], align 8
4110 // CHECK9-NEXT: [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
4111 // CHECK9-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
4112 // CHECK9-NEXT: store i64 [[TMP19]], i64* [[TMP21]], align 8
4113 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
4114 // CHECK9-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
4115 // CHECK9-NEXT: store i64 [[TMP19]], i64* [[TMP23]], align 8
4116 // CHECK9-NEXT: [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
4117 // CHECK9-NEXT: store i8* null, i8** [[TMP24]], align 8
4118 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
4119 // CHECK9-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
4120 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 8
4121 // CHECK9-NEXT: [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
4122 // CHECK9-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
4123 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 8
4124 // CHECK9-NEXT: [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
4125 // CHECK9-NEXT: store i8* null, i8** [[TMP29]], align 8
4126 // CHECK9-NEXT: [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
4127 // CHECK9-NEXT: [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
4128 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
4129 // CHECK9-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.17, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.18, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
4130 // CHECK9-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
4131 // CHECK9-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
4132 // CHECK9: omp_offload.failed11:
4133 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96(i64 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
4134 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT12]]
4135 // CHECK9: omp_offload.cont12:
4136 // CHECK9-NEXT: ret i32 0
4137 //
4138 //
4139 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86
4140 // CHECK9-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4141 // CHECK9-NEXT: entry:
4142 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
4143 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
4144 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
4145 // CHECK9-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..8 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
4146 // CHECK9-NEXT: ret void
4147 //
4148 //
4149 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..8
4150 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4151 // CHECK9-NEXT: entry:
4152 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4153 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4154 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
4155 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4156 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
4157 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4158 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4159 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4160 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4161 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
4162 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4163 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4164 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
4165 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
4166 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4167 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
4168 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4169 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4170 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4171 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4172 // CHECK9-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)
4173 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4174 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
4175 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4176 // CHECK9: cond.true:
4177 // CHECK9-NEXT: br label [[COND_END:%.*]]
4178 // CHECK9: cond.false:
4179 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4180 // CHECK9-NEXT: br label [[COND_END]]
4181 // CHECK9: cond.end:
4182 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4183 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4184 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4185 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4186 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4187 // CHECK9: omp.inner.for.cond:
4188 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4189 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4190 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4191 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4192 // CHECK9: omp.inner.for.body:
4193 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4194 // CHECK9-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
4195 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4196 // CHECK9-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
4197 // CHECK9-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..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
4198 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4199 // CHECK9: omp.inner.for.inc:
4200 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4201 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4202 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4203 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4204 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
4205 // CHECK9: omp.inner.for.end:
4206 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4207 // CHECK9: omp.loop.exit:
4208 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4209 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4210 // CHECK9-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
4211 // CHECK9-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4212 // CHECK9: .omp.final.then:
4213 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
4214 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
4215 // CHECK9: .omp.final.done:
4216 // CHECK9-NEXT: ret void
4217 //
4218 //
4219 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..9
4220 // CHECK9-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]] {
4221 // CHECK9-NEXT: entry:
4222 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4223 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4224 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4225 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4226 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
4227 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4228 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
4229 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4230 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4231 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4232 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4233 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
4234 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4235 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4236 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4237 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4238 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
4239 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
4240 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4241 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
4242 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4243 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
4244 // CHECK9-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4245 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
4246 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4247 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4248 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4249 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4250 // CHECK9-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4251 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4252 // CHECK9-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)
4253 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4254 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
4255 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4256 // CHECK9: cond.true:
4257 // CHECK9-NEXT: br label [[COND_END:%.*]]
4258 // CHECK9: cond.false:
4259 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4260 // CHECK9-NEXT: br label [[COND_END]]
4261 // CHECK9: cond.end:
4262 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
4263 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4264 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4265 // CHECK9-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
4266 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4267 // CHECK9: omp.inner.for.cond:
4268 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4269 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4270 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
4271 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4272 // CHECK9: omp.inner.for.body:
4273 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4274 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
4275 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4276 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4277 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
4278 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
4279 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
4280 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4281 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4282 // CHECK9: omp.body.continue:
4283 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4284 // CHECK9: omp.inner.for.inc:
4285 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4286 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
4287 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
4288 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
4289 // CHECK9: omp.inner.for.end:
4290 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4291 // CHECK9: omp.loop.exit:
4292 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
4293 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4294 // CHECK9-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4295 // CHECK9-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4296 // CHECK9: .omp.final.then:
4297 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
4298 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
4299 // CHECK9: .omp.final.done:
4300 // CHECK9-NEXT: ret void
4301 //
4302 //
4303 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91
4304 // CHECK9-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4305 // CHECK9-NEXT: entry:
4306 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
4307 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
4308 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
4309 // CHECK9-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..11 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
4310 // CHECK9-NEXT: ret void
4311 //
4312 //
4313 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..11
4314 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4315 // CHECK9-NEXT: entry:
4316 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4317 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4318 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
4319 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4320 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
4321 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4322 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4323 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4324 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4325 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
4326 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4327 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4328 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
4329 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
4330 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4331 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
4332 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4333 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4334 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4335 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4336 // CHECK9-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)
4337 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4338 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
4339 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4340 // CHECK9: cond.true:
4341 // CHECK9-NEXT: br label [[COND_END:%.*]]
4342 // CHECK9: cond.false:
4343 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4344 // CHECK9-NEXT: br label [[COND_END]]
4345 // CHECK9: cond.end:
4346 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4347 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4348 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4349 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4350 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4351 // CHECK9: omp.inner.for.cond:
4352 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4353 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4354 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4355 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4356 // CHECK9: omp.inner.for.body:
4357 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4358 // CHECK9-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
4359 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4360 // CHECK9-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
4361 // CHECK9-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..12 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
4362 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4363 // CHECK9: omp.inner.for.inc:
4364 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4365 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4366 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4367 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4368 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
4369 // CHECK9: omp.inner.for.end:
4370 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4371 // CHECK9: omp.loop.exit:
4372 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4373 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4374 // CHECK9-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
4375 // CHECK9-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4376 // CHECK9: .omp.final.then:
4377 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
4378 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
4379 // CHECK9: .omp.final.done:
4380 // CHECK9-NEXT: ret void
4381 //
4382 //
4383 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..12
4384 // CHECK9-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]] {
4385 // CHECK9-NEXT: entry:
4386 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4387 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4388 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4389 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4390 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
4391 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4392 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
4393 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4394 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4395 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4396 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4397 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
4398 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4399 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4400 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4401 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4402 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
4403 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
4404 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4405 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
4406 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4407 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
4408 // CHECK9-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4409 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
4410 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4411 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4412 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4413 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4414 // CHECK9-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4415 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4416 // CHECK9-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)
4417 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4418 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
4419 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4420 // CHECK9: cond.true:
4421 // CHECK9-NEXT: br label [[COND_END:%.*]]
4422 // CHECK9: cond.false:
4423 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4424 // CHECK9-NEXT: br label [[COND_END]]
4425 // CHECK9: cond.end:
4426 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
4427 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4428 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4429 // CHECK9-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
4430 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4431 // CHECK9: omp.inner.for.cond:
4432 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4433 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4434 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
4435 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4436 // CHECK9: omp.inner.for.body:
4437 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4438 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
4439 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4440 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4441 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
4442 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
4443 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
4444 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4445 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4446 // CHECK9: omp.body.continue:
4447 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4448 // CHECK9: omp.inner.for.inc:
4449 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4450 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
4451 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
4452 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
4453 // CHECK9: omp.inner.for.end:
4454 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4455 // CHECK9: omp.loop.exit:
4456 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
4457 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4458 // CHECK9-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4459 // CHECK9-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4460 // CHECK9: .omp.final.then:
4461 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
4462 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
4463 // CHECK9: .omp.final.done:
4464 // CHECK9-NEXT: ret void
4465 //
4466 //
4467 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96
4468 // CHECK9-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4469 // CHECK9-NEXT: entry:
4470 // CHECK9-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
4471 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
4472 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4473 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
4474 // CHECK9-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
4475 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
4476 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
4477 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
4478 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
4479 // CHECK9-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
4480 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4481 // CHECK9-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
4482 // CHECK9-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
4483 // CHECK9-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
4484 // CHECK9-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..15 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
4485 // CHECK9-NEXT: ret void
4486 //
4487 //
4488 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..15
4489 // CHECK9-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
4490 // CHECK9-NEXT: entry:
4491 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4492 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4493 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
4494 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
4495 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4496 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
4497 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4498 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4499 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4500 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4501 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
4502 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
4503 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4504 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4505 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
4506 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
4507 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
4508 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
4509 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4510 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
4511 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4512 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4513 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
4514 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4515 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4516 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP1]])
4517 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4518 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
4519 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4520 // CHECK9: cond.true:
4521 // CHECK9-NEXT: br label [[COND_END:%.*]]
4522 // CHECK9: cond.false:
4523 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4524 // CHECK9-NEXT: br label [[COND_END]]
4525 // CHECK9: cond.end:
4526 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4527 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4528 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4529 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4530 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4531 // CHECK9: omp.inner.for.cond:
4532 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4533 // CHECK9-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP7]], 10
4534 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4535 // CHECK9: omp.inner.for.body:
4536 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4537 // CHECK9-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
4538 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4539 // CHECK9-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
4540 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8
4541 // CHECK9-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
4542 // CHECK9-NEXT: store i32 [[TMP12]], i32* [[CONV2]], align 4
4543 // CHECK9-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
4544 // CHECK9-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..16 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
4545 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4546 // CHECK9: omp.inner.for.inc:
4547 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4548 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4549 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
4550 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4551 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4552 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4553 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
4554 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_LB]], align 4
4555 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4556 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4557 // CHECK9-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
4558 // CHECK9-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_COMB_UB]], align 4
4559 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4560 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP20]], 9
4561 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
4562 // CHECK9: cond.true6:
4563 // CHECK9-NEXT: br label [[COND_END8:%.*]]
4564 // CHECK9: cond.false7:
4565 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4566 // CHECK9-NEXT: br label [[COND_END8]]
4567 // CHECK9: cond.end8:
4568 // CHECK9-NEXT: [[COND9:%.*]] = phi i32 [ 9, [[COND_TRUE6]] ], [ [[TMP21]], [[COND_FALSE7]] ]
4569 // CHECK9-NEXT: store i32 [[COND9]], i32* [[DOTOMP_COMB_UB]], align 4
4570 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4571 // CHECK9-NEXT: store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
4572 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
4573 // CHECK9: omp.inner.for.end:
4574 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4575 // CHECK9: omp.loop.exit:
4576 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4577 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4578 // CHECK9-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
4579 // CHECK9-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4580 // CHECK9: .omp.final.then:
4581 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
4582 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
4583 // CHECK9: .omp.final.done:
4584 // CHECK9-NEXT: ret void
4585 //
4586 //
4587 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..16
4588 // CHECK9-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]] {
4589 // CHECK9-NEXT: entry:
4590 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4591 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4592 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4593 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4594 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
4595 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
4596 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4597 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
4598 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4599 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4600 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4601 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4602 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
4603 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4604 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4605 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4606 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4607 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
4608 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
4609 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
4610 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
4611 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4612 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
4613 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4614 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4615 // CHECK9-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4616 // CHECK9-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
4617 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
4618 // CHECK9-NEXT: store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
4619 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4620 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4621 // CHECK9-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4622 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4623 // CHECK9-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)
4624 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4625 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
4626 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4627 // CHECK9: cond.true:
4628 // CHECK9-NEXT: br label [[COND_END:%.*]]
4629 // CHECK9: cond.false:
4630 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4631 // CHECK9-NEXT: br label [[COND_END]]
4632 // CHECK9: cond.end:
4633 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
4634 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4635 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4636 // CHECK9-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
4637 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4638 // CHECK9: omp.inner.for.cond:
4639 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4640 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4641 // CHECK9-NEXT: [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
4642 // CHECK9-NEXT: br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4643 // CHECK9: omp.inner.for.body:
4644 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4645 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
4646 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4647 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4648 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
4649 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
4650 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
4651 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4652 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4653 // CHECK9: omp.body.continue:
4654 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4655 // CHECK9: omp.inner.for.inc:
4656 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4657 // CHECK9-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
4658 // CHECK9-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4
4659 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
4660 // CHECK9: omp.inner.for.end:
4661 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4662 // CHECK9: omp.loop.exit:
4663 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
4664 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4665 // CHECK9-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4666 // CHECK9-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4667 // CHECK9: .omp.final.then:
4668 // CHECK9-NEXT: store i32 10, i32* [[I]], align 4
4669 // CHECK9-NEXT: br label [[DOTOMP_FINAL_DONE]]
4670 // CHECK9: .omp.final.done:
4671 // CHECK9-NEXT: ret void
4672 //
4673 //
4674 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4675 // CHECK9-SAME: () #[[ATTR5:[0-9]+]] {
4676 // CHECK9-NEXT: entry:
4677 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
4678 // CHECK9-NEXT: ret void
4679 //
4680 //
4681 // CHECK10-LABEL: define {{[^@]+}}@main
4682 // CHECK10-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
4683 // CHECK10-NEXT: entry:
4684 // CHECK10-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
4685 // CHECK10-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
4686 // CHECK10-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
4687 // CHECK10-NEXT: [[N:%.*]] = alloca i32, align 4
4688 // CHECK10-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
4689 // CHECK10-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
4690 // CHECK10-NEXT: [[M:%.*]] = alloca i32, align 4
4691 // CHECK10-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
4692 // CHECK10-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
4693 // CHECK10-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
4694 // CHECK10-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
4695 // CHECK10-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
4696 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
4697 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4698 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4699 // CHECK10-NEXT: [[N_CASTED3:%.*]] = alloca i64, align 8
4700 // CHECK10-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
4701 // CHECK10-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
4702 // CHECK10-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
4703 // CHECK10-NEXT: [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
4704 // CHECK10-NEXT: [[_TMP9:%.*]] = alloca i32, align 4
4705 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
4706 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
4707 // CHECK10-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
4708 // CHECK10-NEXT: [[N_CASTED19:%.*]] = alloca i64, align 8
4709 // CHECK10-NEXT: [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [4 x i8*], align 8
4710 // CHECK10-NEXT: [[DOTOFFLOAD_PTRS22:%.*]] = alloca [4 x i8*], align 8
4711 // CHECK10-NEXT: [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [4 x i8*], align 8
4712 // CHECK10-NEXT: [[DOTOFFLOAD_SIZES24:%.*]] = alloca [4 x i64], align 8
4713 // CHECK10-NEXT: [[_TMP25:%.*]] = alloca i32, align 4
4714 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
4715 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
4716 // CHECK10-NEXT: store i32 0, i32* [[RETVAL]], align 4
4717 // CHECK10-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4718 // CHECK10-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
4719 // CHECK10-NEXT: store i32 100, i32* [[N]], align 4
4720 // CHECK10-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
4721 // CHECK10-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
4722 // CHECK10-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
4723 // CHECK10-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
4724 // CHECK10-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
4725 // CHECK10-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
4726 // CHECK10-NEXT: store i32 10, i32* [[M]], align 4
4727 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
4728 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
4729 // CHECK10-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
4730 // CHECK10-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
4731 // CHECK10-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
4732 // CHECK10-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4733 // CHECK10-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
4734 // CHECK10-NEXT: store i64 [[TMP4]], i64* [[TMP7]], align 8
4735 // CHECK10-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4736 // CHECK10-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
4737 // CHECK10-NEXT: store i64 [[TMP4]], i64* [[TMP9]], align 8
4738 // CHECK10-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
4739 // CHECK10-NEXT: store i64 4, i64* [[TMP10]], align 8
4740 // CHECK10-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4741 // CHECK10-NEXT: store i8* null, i8** [[TMP11]], align 8
4742 // CHECK10-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4743 // CHECK10-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
4744 // CHECK10-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
4745 // CHECK10-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4746 // CHECK10-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
4747 // CHECK10-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
4748 // CHECK10-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
4749 // CHECK10-NEXT: store i64 8, i64* [[TMP16]], align 8
4750 // CHECK10-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
4751 // CHECK10-NEXT: store i8* null, i8** [[TMP17]], align 8
4752 // CHECK10-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4753 // CHECK10-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
4754 // CHECK10-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 8
4755 // CHECK10-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4756 // CHECK10-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
4757 // CHECK10-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 8
4758 // CHECK10-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
4759 // CHECK10-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 8
4760 // CHECK10-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
4761 // CHECK10-NEXT: store i8* null, i8** [[TMP23]], align 8
4762 // CHECK10-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4763 // CHECK10-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4764 // CHECK10-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
4765 // CHECK10-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
4766 // CHECK10-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
4767 // CHECK10-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4768 // CHECK10-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
4769 // CHECK10-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4770 // CHECK10-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
4771 // CHECK10-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4772 // CHECK10-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4773 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
4774 // CHECK10-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
4775 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
4776 // CHECK10-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108.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)
4777 // CHECK10-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
4778 // CHECK10-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4779 // CHECK10: omp_offload.failed:
4780 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
4781 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT]]
4782 // CHECK10: omp_offload.cont:
4783 // CHECK10-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
4784 // CHECK10-NEXT: [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
4785 // CHECK10-NEXT: store i32 [[TMP33]], i32* [[CONV4]], align 4
4786 // CHECK10-NEXT: [[TMP34:%.*]] = load i64, i64* [[N_CASTED3]], align 8
4787 // CHECK10-NEXT: [[TMP35:%.*]] = mul nuw i64 [[TMP1]], 4
4788 // CHECK10-NEXT: [[TMP36:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
4789 // CHECK10-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to i64*
4790 // CHECK10-NEXT: store i64 [[TMP34]], i64* [[TMP37]], align 8
4791 // CHECK10-NEXT: [[TMP38:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
4792 // CHECK10-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
4793 // CHECK10-NEXT: store i64 [[TMP34]], i64* [[TMP39]], align 8
4794 // CHECK10-NEXT: [[TMP40:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
4795 // CHECK10-NEXT: store i64 4, i64* [[TMP40]], align 8
4796 // CHECK10-NEXT: [[TMP41:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
4797 // CHECK10-NEXT: store i8* null, i8** [[TMP41]], align 8
4798 // CHECK10-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
4799 // CHECK10-NEXT: [[TMP43:%.*]] = bitcast i8** [[TMP42]] to i64*
4800 // CHECK10-NEXT: store i64 [[TMP1]], i64* [[TMP43]], align 8
4801 // CHECK10-NEXT: [[TMP44:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
4802 // CHECK10-NEXT: [[TMP45:%.*]] = bitcast i8** [[TMP44]] to i64*
4803 // CHECK10-NEXT: store i64 [[TMP1]], i64* [[TMP45]], align 8
4804 // CHECK10-NEXT: [[TMP46:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 1
4805 // CHECK10-NEXT: store i64 8, i64* [[TMP46]], align 8
4806 // CHECK10-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
4807 // CHECK10-NEXT: store i8* null, i8** [[TMP47]], align 8
4808 // CHECK10-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
4809 // CHECK10-NEXT: [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32**
4810 // CHECK10-NEXT: store i32* [[VLA]], i32** [[TMP49]], align 8
4811 // CHECK10-NEXT: [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
4812 // CHECK10-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i32**
4813 // CHECK10-NEXT: store i32* [[VLA]], i32** [[TMP51]], align 8
4814 // CHECK10-NEXT: [[TMP52:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
4815 // CHECK10-NEXT: store i64 [[TMP35]], i64* [[TMP52]], align 8
4816 // CHECK10-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
4817 // CHECK10-NEXT: store i8* null, i8** [[TMP53]], align 8
4818 // CHECK10-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
4819 // CHECK10-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
4820 // CHECK10-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
4821 // CHECK10-NEXT: [[TMP57:%.*]] = load i32, i32* [[N]], align 4
4822 // CHECK10-NEXT: store i32 [[TMP57]], i32* [[DOTCAPTURE_EXPR_10]], align 4
4823 // CHECK10-NEXT: [[TMP58:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
4824 // CHECK10-NEXT: [[SUB12:%.*]] = sub nsw i32 [[TMP58]], 0
4825 // CHECK10-NEXT: [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
4826 // CHECK10-NEXT: [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
4827 // CHECK10-NEXT: store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
4828 // CHECK10-NEXT: [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
4829 // CHECK10-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP59]], 1
4830 // CHECK10-NEXT: [[TMP60:%.*]] = zext i32 [[ADD15]] to i64
4831 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP60]])
4832 // CHECK10-NEXT: [[TMP61:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113.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)
4833 // CHECK10-NEXT: [[TMP62:%.*]] = icmp ne i32 [[TMP61]], 0
4834 // CHECK10-NEXT: br i1 [[TMP62]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
4835 // CHECK10: omp_offload.failed16:
4836 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113(i64 [[TMP34]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
4837 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT17]]
4838 // CHECK10: omp_offload.cont17:
4839 // CHECK10-NEXT: [[TMP63:%.*]] = load i32, i32* [[M]], align 4
4840 // CHECK10-NEXT: [[CONV18:%.*]] = bitcast i64* [[M_CASTED]] to i32*
4841 // CHECK10-NEXT: store i32 [[TMP63]], i32* [[CONV18]], align 4
4842 // CHECK10-NEXT: [[TMP64:%.*]] = load i64, i64* [[M_CASTED]], align 8
4843 // CHECK10-NEXT: [[TMP65:%.*]] = load i32, i32* [[N]], align 4
4844 // CHECK10-NEXT: [[CONV20:%.*]] = bitcast i64* [[N_CASTED19]] to i32*
4845 // CHECK10-NEXT: store i32 [[TMP65]], i32* [[CONV20]], align 4
4846 // CHECK10-NEXT: [[TMP66:%.*]] = load i64, i64* [[N_CASTED19]], align 8
4847 // CHECK10-NEXT: [[TMP67:%.*]] = mul nuw i64 [[TMP1]], 4
4848 // CHECK10-NEXT: [[TMP68:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
4849 // CHECK10-NEXT: [[TMP69:%.*]] = bitcast i8** [[TMP68]] to i64*
4850 // CHECK10-NEXT: store i64 [[TMP64]], i64* [[TMP69]], align 8
4851 // CHECK10-NEXT: [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
4852 // CHECK10-NEXT: [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i64*
4853 // CHECK10-NEXT: store i64 [[TMP64]], i64* [[TMP71]], align 8
4854 // CHECK10-NEXT: [[TMP72:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
4855 // CHECK10-NEXT: store i64 4, i64* [[TMP72]], align 8
4856 // CHECK10-NEXT: [[TMP73:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
4857 // CHECK10-NEXT: store i8* null, i8** [[TMP73]], align 8
4858 // CHECK10-NEXT: [[TMP74:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
4859 // CHECK10-NEXT: [[TMP75:%.*]] = bitcast i8** [[TMP74]] to i64*
4860 // CHECK10-NEXT: store i64 [[TMP66]], i64* [[TMP75]], align 8
4861 // CHECK10-NEXT: [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
4862 // CHECK10-NEXT: [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i64*
4863 // CHECK10-NEXT: store i64 [[TMP66]], i64* [[TMP77]], align 8
4864 // CHECK10-NEXT: [[TMP78:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 1
4865 // CHECK10-NEXT: store i64 4, i64* [[TMP78]], align 8
4866 // CHECK10-NEXT: [[TMP79:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
4867 // CHECK10-NEXT: store i8* null, i8** [[TMP79]], align 8
4868 // CHECK10-NEXT: [[TMP80:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 2
4869 // CHECK10-NEXT: [[TMP81:%.*]] = bitcast i8** [[TMP80]] to i64*
4870 // CHECK10-NEXT: store i64 [[TMP1]], i64* [[TMP81]], align 8
4871 // CHECK10-NEXT: [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 2
4872 // CHECK10-NEXT: [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i64*
4873 // CHECK10-NEXT: store i64 [[TMP1]], i64* [[TMP83]], align 8
4874 // CHECK10-NEXT: [[TMP84:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 2
4875 // CHECK10-NEXT: store i64 8, i64* [[TMP84]], align 8
4876 // CHECK10-NEXT: [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 2
4877 // CHECK10-NEXT: store i8* null, i8** [[TMP85]], align 8
4878 // CHECK10-NEXT: [[TMP86:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 3
4879 // CHECK10-NEXT: [[TMP87:%.*]] = bitcast i8** [[TMP86]] to i32**
4880 // CHECK10-NEXT: store i32* [[VLA]], i32** [[TMP87]], align 8
4881 // CHECK10-NEXT: [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 3
4882 // CHECK10-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
4883 // CHECK10-NEXT: store i32* [[VLA]], i32** [[TMP89]], align 8
4884 // CHECK10-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 3
4885 // CHECK10-NEXT: store i64 [[TMP67]], i64* [[TMP90]], align 8
4886 // CHECK10-NEXT: [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 3
4887 // CHECK10-NEXT: store i8* null, i8** [[TMP91]], align 8
4888 // CHECK10-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
4889 // CHECK10-NEXT: [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
4890 // CHECK10-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
4891 // CHECK10-NEXT: [[TMP95:%.*]] = load i32, i32* [[N]], align 4
4892 // CHECK10-NEXT: store i32 [[TMP95]], i32* [[DOTCAPTURE_EXPR_26]], align 4
4893 // CHECK10-NEXT: [[TMP96:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_26]], align 4
4894 // CHECK10-NEXT: [[SUB28:%.*]] = sub nsw i32 [[TMP96]], 0
4895 // CHECK10-NEXT: [[DIV29:%.*]] = sdiv i32 [[SUB28]], 1
4896 // CHECK10-NEXT: [[SUB30:%.*]] = sub nsw i32 [[DIV29]], 1
4897 // CHECK10-NEXT: store i32 [[SUB30]], i32* [[DOTCAPTURE_EXPR_27]], align 4
4898 // CHECK10-NEXT: [[TMP97:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
4899 // CHECK10-NEXT: [[ADD31:%.*]] = add nsw i32 [[TMP97]], 1
4900 // CHECK10-NEXT: [[TMP98:%.*]] = zext i32 [[ADD31]] to i64
4901 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP98]])
4902 // CHECK10-NEXT: [[TMP99:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118.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)
4903 // CHECK10-NEXT: [[TMP100:%.*]] = icmp ne i32 [[TMP99]], 0
4904 // CHECK10-NEXT: br i1 [[TMP100]], label [[OMP_OFFLOAD_FAILED32:%.*]], label [[OMP_OFFLOAD_CONT33:%.*]]
4905 // CHECK10: omp_offload.failed32:
4906 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118(i64 [[TMP64]], i64 [[TMP66]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
4907 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT33]]
4908 // CHECK10: omp_offload.cont33:
4909 // CHECK10-NEXT: [[TMP101:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4910 // CHECK10-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP101]])
4911 // CHECK10-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
4912 // CHECK10-NEXT: [[TMP102:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
4913 // CHECK10-NEXT: call void @llvm.stackrestore(i8* [[TMP102]])
4914 // CHECK10-NEXT: [[TMP103:%.*]] = load i32, i32* [[RETVAL]], align 4
4915 // CHECK10-NEXT: ret i32 [[TMP103]]
4916 //
4917 //
4918 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108
4919 // CHECK10-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
4920 // CHECK10-NEXT: entry:
4921 // CHECK10-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
4922 // CHECK10-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
4923 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
4924 // CHECK10-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
4925 // CHECK10-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
4926 // CHECK10-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
4927 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
4928 // CHECK10-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
4929 // CHECK10-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
4930 // CHECK10-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]])
4931 // CHECK10-NEXT: ret void
4932 //
4933 //
4934 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined.
4935 // CHECK10-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]] {
4936 // CHECK10-NEXT: entry:
4937 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4938 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4939 // CHECK10-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
4940 // CHECK10-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
4941 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
4942 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4943 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
4944 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4945 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4946 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
4947 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4948 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4949 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4950 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4951 // CHECK10-NEXT: [[I3:%.*]] = alloca i32, align 4
4952 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4953 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4954 // CHECK10-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
4955 // CHECK10-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
4956 // CHECK10-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
4957 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
4958 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
4959 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
4960 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
4961 // CHECK10-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
4962 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4963 // CHECK10-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
4964 // CHECK10-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4965 // CHECK10-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
4966 // CHECK10-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4967 // CHECK10-NEXT: store i32 0, i32* [[I]], align 4
4968 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4969 // CHECK10-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
4970 // CHECK10-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4971 // CHECK10: omp.precond.then:
4972 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4973 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4974 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
4975 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4976 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4977 // CHECK10-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4978 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
4979 // CHECK10-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)
4980 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4981 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4982 // CHECK10-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
4983 // CHECK10-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4984 // CHECK10: cond.true:
4985 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4986 // CHECK10-NEXT: br label [[COND_END:%.*]]
4987 // CHECK10: cond.false:
4988 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4989 // CHECK10-NEXT: br label [[COND_END]]
4990 // CHECK10: cond.end:
4991 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
4992 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4993 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4994 // CHECK10-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
4995 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4996 // CHECK10: omp.inner.for.cond:
4997 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4998 // CHECK10-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4999 // CHECK10-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
5000 // CHECK10-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5001 // CHECK10: omp.inner.for.body:
5002 // CHECK10-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5003 // CHECK10-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
5004 // CHECK10-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5005 // CHECK10-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
5006 // CHECK10-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]])
5007 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5008 // CHECK10: omp.inner.for.inc:
5009 // CHECK10-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5010 // CHECK10-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5011 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
5012 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5013 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP8:![0-9]+]]
5014 // CHECK10: omp.inner.for.end:
5015 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5016 // CHECK10: omp.loop.exit:
5017 // CHECK10-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5018 // CHECK10-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
5019 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
5020 // CHECK10-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5021 // CHECK10-NEXT: [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
5022 // CHECK10-NEXT: br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5023 // CHECK10: .omp.final.then:
5024 // CHECK10-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5025 // CHECK10-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
5026 // CHECK10-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
5027 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
5028 // CHECK10-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
5029 // CHECK10-NEXT: store i32 [[ADD8]], i32* [[I3]], align 4
5030 // CHECK10-NEXT: br label [[DOTOMP_FINAL_DONE]]
5031 // CHECK10: .omp.final.done:
5032 // CHECK10-NEXT: br label [[OMP_PRECOND_END]]
5033 // CHECK10: omp.precond.end:
5034 // CHECK10-NEXT: ret void
5035 //
5036 //
5037 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
5038 // CHECK10-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]] {
5039 // CHECK10-NEXT: entry:
5040 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5041 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5042 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5043 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5044 // CHECK10-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
5045 // CHECK10-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
5046 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
5047 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5048 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
5049 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
5050 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
5051 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
5052 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5053 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5054 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5055 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5056 // CHECK10-NEXT: [[I4:%.*]] = alloca i32, align 4
5057 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5058 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5059 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5060 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5061 // CHECK10-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
5062 // CHECK10-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
5063 // CHECK10-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
5064 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
5065 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
5066 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
5067 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
5068 // CHECK10-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
5069 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5070 // CHECK10-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
5071 // CHECK10-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
5072 // CHECK10-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
5073 // CHECK10-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
5074 // CHECK10-NEXT: store i32 0, i32* [[I]], align 4
5075 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5076 // CHECK10-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
5077 // CHECK10-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
5078 // CHECK10: omp.precond.then:
5079 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5080 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5081 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
5082 // CHECK10-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5083 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
5084 // CHECK10-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5085 // CHECK10-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
5086 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5087 // CHECK10-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
5088 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5089 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5090 // CHECK10-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5091 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
5092 // CHECK10-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)
5093 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5094 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5095 // CHECK10-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
5096 // CHECK10-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5097 // CHECK10: cond.true:
5098 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5099 // CHECK10-NEXT: br label [[COND_END:%.*]]
5100 // CHECK10: cond.false:
5101 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5102 // CHECK10-NEXT: br label [[COND_END]]
5103 // CHECK10: cond.end:
5104 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
5105 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5106 // CHECK10-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5107 // CHECK10-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
5108 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5109 // CHECK10: omp.inner.for.cond:
5110 // CHECK10-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5111 // CHECK10-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5112 // CHECK10-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
5113 // CHECK10-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5114 // CHECK10: omp.inner.for.body:
5115 // CHECK10-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5116 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
5117 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5118 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
5119 // CHECK10-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
5120 // CHECK10-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
5121 // CHECK10-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
5122 // CHECK10-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
5123 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5124 // CHECK10: omp.body.continue:
5125 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5126 // CHECK10: omp.inner.for.inc:
5127 // CHECK10-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5128 // CHECK10-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
5129 // CHECK10-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
5130 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
5131 // CHECK10: omp.inner.for.end:
5132 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5133 // CHECK10: omp.loop.exit:
5134 // CHECK10-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5135 // CHECK10-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
5136 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
5137 // CHECK10-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5138 // CHECK10-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
5139 // CHECK10-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5140 // CHECK10: .omp.final.then:
5141 // CHECK10-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5142 // CHECK10-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
5143 // CHECK10-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
5144 // CHECK10-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
5145 // CHECK10-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
5146 // CHECK10-NEXT: store i32 [[ADD11]], i32* [[I4]], align 4
5147 // CHECK10-NEXT: br label [[DOTOMP_FINAL_DONE]]
5148 // CHECK10: .omp.final.done:
5149 // CHECK10-NEXT: br label [[OMP_PRECOND_END]]
5150 // CHECK10: omp.precond.end:
5151 // CHECK10-NEXT: ret void
5152 //
5153 //
5154 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113
5155 // CHECK10-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
5156 // CHECK10-NEXT: entry:
5157 // CHECK10-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
5158 // CHECK10-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
5159 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
5160 // CHECK10-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
5161 // CHECK10-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
5162 // CHECK10-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
5163 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
5164 // CHECK10-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
5165 // CHECK10-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
5166 // CHECK10-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]])
5167 // CHECK10-NEXT: ret void
5168 //
5169 //
5170 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
5171 // CHECK10-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]] {
5172 // CHECK10-NEXT: entry:
5173 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5174 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5175 // CHECK10-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
5176 // CHECK10-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
5177 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
5178 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5179 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
5180 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
5181 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
5182 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
5183 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5184 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5185 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5186 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5187 // CHECK10-NEXT: [[I3:%.*]] = alloca i32, align 4
5188 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5189 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5190 // CHECK10-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
5191 // CHECK10-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
5192 // CHECK10-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
5193 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
5194 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
5195 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
5196 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
5197 // CHECK10-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
5198 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5199 // CHECK10-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
5200 // CHECK10-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
5201 // CHECK10-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
5202 // CHECK10-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
5203 // CHECK10-NEXT: store i32 0, i32* [[I]], align 4
5204 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5205 // CHECK10-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
5206 // CHECK10-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
5207 // CHECK10: omp.precond.then:
5208 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5209 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5210 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
5211 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5212 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5213 // CHECK10-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5214 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
5215 // CHECK10-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)
5216 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5217 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5218 // CHECK10-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
5219 // CHECK10-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5220 // CHECK10: cond.true:
5221 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5222 // CHECK10-NEXT: br label [[COND_END:%.*]]
5223 // CHECK10: cond.false:
5224 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5225 // CHECK10-NEXT: br label [[COND_END]]
5226 // CHECK10: cond.end:
5227 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
5228 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5229 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5230 // CHECK10-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
5231 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5232 // CHECK10: omp.inner.for.cond:
5233 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5234 // CHECK10-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5235 // CHECK10-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
5236 // CHECK10-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5237 // CHECK10: omp.inner.for.body:
5238 // CHECK10-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5239 // CHECK10-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
5240 // CHECK10-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5241 // CHECK10-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
5242 // CHECK10-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]])
5243 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5244 // CHECK10: omp.inner.for.inc:
5245 // CHECK10-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5246 // CHECK10-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5247 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
5248 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5249 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
5250 // CHECK10: omp.inner.for.end:
5251 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5252 // CHECK10: omp.loop.exit:
5253 // CHECK10-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5254 // CHECK10-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
5255 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
5256 // CHECK10-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5257 // CHECK10-NEXT: [[TMP25:%.*]] = icmp ne i32 [[TMP24]], 0
5258 // CHECK10-NEXT: br i1 [[TMP25]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5259 // CHECK10: .omp.final.then:
5260 // CHECK10-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5261 // CHECK10-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP26]], 0
5262 // CHECK10-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
5263 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
5264 // CHECK10-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
5265 // CHECK10-NEXT: store i32 [[ADD8]], i32* [[I3]], align 4
5266 // CHECK10-NEXT: br label [[DOTOMP_FINAL_DONE]]
5267 // CHECK10: .omp.final.done:
5268 // CHECK10-NEXT: br label [[OMP_PRECOND_END]]
5269 // CHECK10: omp.precond.end:
5270 // CHECK10-NEXT: ret void
5271 //
5272 //
5273 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
5274 // CHECK10-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]] {
5275 // CHECK10-NEXT: entry:
5276 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5277 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5278 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5279 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5280 // CHECK10-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
5281 // CHECK10-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
5282 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
5283 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5284 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
5285 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
5286 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
5287 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
5288 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5289 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5290 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5291 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5292 // CHECK10-NEXT: [[I4:%.*]] = alloca i32, align 4
5293 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5294 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5295 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5296 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5297 // CHECK10-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
5298 // CHECK10-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
5299 // CHECK10-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
5300 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
5301 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
5302 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
5303 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
5304 // CHECK10-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
5305 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5306 // CHECK10-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
5307 // CHECK10-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
5308 // CHECK10-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
5309 // CHECK10-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
5310 // CHECK10-NEXT: store i32 0, i32* [[I]], align 4
5311 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5312 // CHECK10-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
5313 // CHECK10-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
5314 // CHECK10: omp.precond.then:
5315 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5316 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5317 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
5318 // CHECK10-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5319 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
5320 // CHECK10-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5321 // CHECK10-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
5322 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5323 // CHECK10-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
5324 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5325 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5326 // CHECK10-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5327 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
5328 // CHECK10-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)
5329 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5330 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5331 // CHECK10-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
5332 // CHECK10-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5333 // CHECK10: cond.true:
5334 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5335 // CHECK10-NEXT: br label [[COND_END:%.*]]
5336 // CHECK10: cond.false:
5337 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5338 // CHECK10-NEXT: br label [[COND_END]]
5339 // CHECK10: cond.end:
5340 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
5341 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5342 // CHECK10-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5343 // CHECK10-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
5344 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5345 // CHECK10: omp.inner.for.cond:
5346 // CHECK10-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5347 // CHECK10-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5348 // CHECK10-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
5349 // CHECK10-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5350 // CHECK10: omp.inner.for.body:
5351 // CHECK10-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5352 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
5353 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5354 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
5355 // CHECK10-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
5356 // CHECK10-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
5357 // CHECK10-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
5358 // CHECK10-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
5359 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5360 // CHECK10: omp.body.continue:
5361 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5362 // CHECK10: omp.inner.for.inc:
5363 // CHECK10-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5364 // CHECK10-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
5365 // CHECK10-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
5366 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
5367 // CHECK10: omp.inner.for.end:
5368 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5369 // CHECK10: omp.loop.exit:
5370 // CHECK10-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5371 // CHECK10-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
5372 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
5373 // CHECK10-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5374 // CHECK10-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
5375 // CHECK10-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5376 // CHECK10: .omp.final.then:
5377 // CHECK10-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5378 // CHECK10-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
5379 // CHECK10-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
5380 // CHECK10-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
5381 // CHECK10-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
5382 // CHECK10-NEXT: store i32 [[ADD11]], i32* [[I4]], align 4
5383 // CHECK10-NEXT: br label [[DOTOMP_FINAL_DONE]]
5384 // CHECK10: .omp.final.done:
5385 // CHECK10-NEXT: br label [[OMP_PRECOND_END]]
5386 // CHECK10: omp.precond.end:
5387 // CHECK10-NEXT: ret void
5388 //
5389 //
5390 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118
5391 // CHECK10-SAME: (i64 [[M:%.*]], i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
5392 // CHECK10-NEXT: entry:
5393 // CHECK10-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
5394 // CHECK10-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
5395 // CHECK10-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
5396 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
5397 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
5398 // CHECK10-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
5399 // CHECK10-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
5400 // CHECK10-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
5401 // CHECK10-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
5402 // CHECK10-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
5403 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
5404 // CHECK10-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
5405 // CHECK10-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
5406 // CHECK10-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
5407 // CHECK10-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
5408 // CHECK10-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
5409 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
5410 // CHECK10-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
5411 // CHECK10-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
5412 // CHECK10-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
5413 // CHECK10-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]])
5414 // CHECK10-NEXT: ret void
5415 //
5416 //
5417 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..5
5418 // CHECK10-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]] {
5419 // CHECK10-NEXT: entry:
5420 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5421 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5422 // CHECK10-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
5423 // CHECK10-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
5424 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
5425 // CHECK10-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
5426 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5427 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
5428 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
5429 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
5430 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
5431 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5432 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5433 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5434 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5435 // CHECK10-NEXT: [[I4:%.*]] = alloca i32, align 4
5436 // CHECK10-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
5437 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5438 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5439 // CHECK10-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
5440 // CHECK10-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
5441 // CHECK10-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
5442 // CHECK10-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
5443 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
5444 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
5445 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
5446 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
5447 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
5448 // CHECK10-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
5449 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5450 // CHECK10-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
5451 // CHECK10-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
5452 // CHECK10-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
5453 // CHECK10-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
5454 // CHECK10-NEXT: store i32 0, i32* [[I]], align 4
5455 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5456 // CHECK10-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
5457 // CHECK10-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
5458 // CHECK10: omp.precond.then:
5459 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5460 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5461 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
5462 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5463 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5464 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[CONV]], align 8
5465 // CHECK10-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5466 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
5467 // CHECK10-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]])
5468 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5469 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5470 // CHECK10-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
5471 // CHECK10-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5472 // CHECK10: cond.true:
5473 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5474 // CHECK10-NEXT: br label [[COND_END:%.*]]
5475 // CHECK10: cond.false:
5476 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5477 // CHECK10-NEXT: br label [[COND_END]]
5478 // CHECK10: cond.end:
5479 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
5480 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5481 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5482 // CHECK10-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
5483 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5484 // CHECK10: omp.inner.for.cond:
5485 // CHECK10-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5486 // CHECK10-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5487 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
5488 // CHECK10-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
5489 // CHECK10-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5490 // CHECK10: omp.inner.for.body:
5491 // CHECK10-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5492 // CHECK10-NEXT: [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
5493 // CHECK10-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5494 // CHECK10-NEXT: [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
5495 // CHECK10-NEXT: [[TMP21:%.*]] = load i32, i32* [[CONV]], align 8
5496 // CHECK10-NEXT: [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
5497 // CHECK10-NEXT: store i32 [[TMP21]], i32* [[CONV7]], align 4
5498 // CHECK10-NEXT: [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
5499 // CHECK10-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]])
5500 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5501 // CHECK10: omp.inner.for.inc:
5502 // CHECK10-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5503 // CHECK10-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5504 // CHECK10-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
5505 // CHECK10-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
5506 // CHECK10-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5507 // CHECK10-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5508 // CHECK10-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
5509 // CHECK10-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4
5510 // CHECK10-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5511 // CHECK10-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5512 // CHECK10-NEXT: [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
5513 // CHECK10-NEXT: store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4
5514 // CHECK10-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5515 // CHECK10-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5516 // CHECK10-NEXT: [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
5517 // CHECK10-NEXT: br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
5518 // CHECK10: cond.true12:
5519 // CHECK10-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5520 // CHECK10-NEXT: br label [[COND_END14:%.*]]
5521 // CHECK10: cond.false13:
5522 // CHECK10-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5523 // CHECK10-NEXT: br label [[COND_END14]]
5524 // CHECK10: cond.end14:
5525 // CHECK10-NEXT: [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
5526 // CHECK10-NEXT: store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4
5527 // CHECK10-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5528 // CHECK10-NEXT: store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4
5529 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
5530 // CHECK10: omp.inner.for.end:
5531 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5532 // CHECK10: omp.loop.exit:
5533 // CHECK10-NEXT: [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5534 // CHECK10-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
5535 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
5536 // CHECK10-NEXT: [[TMP36:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5537 // CHECK10-NEXT: [[TMP37:%.*]] = icmp ne i32 [[TMP36]], 0
5538 // CHECK10-NEXT: br i1 [[TMP37]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5539 // CHECK10: .omp.final.then:
5540 // CHECK10-NEXT: [[TMP38:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5541 // CHECK10-NEXT: [[SUB16:%.*]] = sub nsw i32 [[TMP38]], 0
5542 // CHECK10-NEXT: [[DIV17:%.*]] = sdiv i32 [[SUB16]], 1
5543 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV17]], 1
5544 // CHECK10-NEXT: [[ADD18:%.*]] = add nsw i32 0, [[MUL]]
5545 // CHECK10-NEXT: store i32 [[ADD18]], i32* [[I4]], align 4
5546 // CHECK10-NEXT: br label [[DOTOMP_FINAL_DONE]]
5547 // CHECK10: .omp.final.done:
5548 // CHECK10-NEXT: br label [[OMP_PRECOND_END]]
5549 // CHECK10: omp.precond.end:
5550 // CHECK10-NEXT: ret void
5551 //
5552 //
5553 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..6
5554 // CHECK10-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]] {
5555 // CHECK10-NEXT: entry:
5556 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5557 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5558 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5559 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5560 // CHECK10-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
5561 // CHECK10-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
5562 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
5563 // CHECK10-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
5564 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5565 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
5566 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
5567 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
5568 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
5569 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5570 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5571 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5572 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5573 // CHECK10-NEXT: [[I6:%.*]] = alloca i32, align 4
5574 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5575 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5576 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5577 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5578 // CHECK10-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
5579 // CHECK10-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
5580 // CHECK10-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
5581 // CHECK10-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
5582 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
5583 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
5584 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
5585 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
5586 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
5587 // CHECK10-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
5588 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5589 // CHECK10-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
5590 // CHECK10-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
5591 // CHECK10-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
5592 // CHECK10-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
5593 // CHECK10-NEXT: store i32 0, i32* [[I]], align 4
5594 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5595 // CHECK10-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
5596 // CHECK10-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
5597 // CHECK10: omp.precond.then:
5598 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5599 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5600 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
5601 // CHECK10-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5602 // CHECK10-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
5603 // CHECK10-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5604 // CHECK10-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
5605 // CHECK10-NEXT: store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
5606 // CHECK10-NEXT: store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
5607 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5608 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5609 // CHECK10-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5610 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
5611 // CHECK10-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)
5612 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5613 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5614 // CHECK10-NEXT: [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
5615 // CHECK10-NEXT: br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5616 // CHECK10: cond.true:
5617 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
5618 // CHECK10-NEXT: br label [[COND_END:%.*]]
5619 // CHECK10: cond.false:
5620 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5621 // CHECK10-NEXT: br label [[COND_END]]
5622 // CHECK10: cond.end:
5623 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
5624 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5625 // CHECK10-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5626 // CHECK10-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
5627 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5628 // CHECK10: omp.inner.for.cond:
5629 // CHECK10-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5630 // CHECK10-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5631 // CHECK10-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
5632 // CHECK10-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5633 // CHECK10: omp.inner.for.body:
5634 // CHECK10-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5635 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
5636 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5637 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I6]], align 4
5638 // CHECK10-NEXT: [[TMP19:%.*]] = load i32, i32* [[I6]], align 4
5639 // CHECK10-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
5640 // CHECK10-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
5641 // CHECK10-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
5642 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5643 // CHECK10: omp.body.continue:
5644 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5645 // CHECK10: omp.inner.for.inc:
5646 // CHECK10-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5647 // CHECK10-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
5648 // CHECK10-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
5649 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
5650 // CHECK10: omp.inner.for.end:
5651 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5652 // CHECK10: omp.loop.exit:
5653 // CHECK10-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5654 // CHECK10-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
5655 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
5656 // CHECK10-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5657 // CHECK10-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
5658 // CHECK10-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5659 // CHECK10: .omp.final.then:
5660 // CHECK10-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
5661 // CHECK10-NEXT: [[SUB10:%.*]] = sub nsw i32 [[TMP25]], 0
5662 // CHECK10-NEXT: [[DIV11:%.*]] = sdiv i32 [[SUB10]], 1
5663 // CHECK10-NEXT: [[MUL12:%.*]] = mul nsw i32 [[DIV11]], 1
5664 // CHECK10-NEXT: [[ADD13:%.*]] = add nsw i32 0, [[MUL12]]
5665 // CHECK10-NEXT: store i32 [[ADD13]], i32* [[I6]], align 4
5666 // CHECK10-NEXT: br label [[DOTOMP_FINAL_DONE]]
5667 // CHECK10: .omp.final.done:
5668 // CHECK10-NEXT: br label [[OMP_PRECOND_END]]
5669 // CHECK10: omp.precond.end:
5670 // CHECK10-NEXT: ret void
5671 //
5672 //
5673 // CHECK10-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
5674 // CHECK10-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
5675 // CHECK10-NEXT: entry:
5676 // CHECK10-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
5677 // CHECK10-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
5678 // CHECK10-NEXT: [[M:%.*]] = alloca i32, align 4
5679 // CHECK10-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
5680 // CHECK10-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
5681 // CHECK10-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
5682 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
5683 // CHECK10-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
5684 // CHECK10-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
5685 // CHECK10-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
5686 // CHECK10-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
5687 // CHECK10-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
5688 // CHECK10-NEXT: [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 8
5689 // CHECK10-NEXT: [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 8
5690 // CHECK10-NEXT: [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 8
5691 // CHECK10-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
5692 // CHECK10-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
5693 // CHECK10-NEXT: store i32 10, i32* [[M]], align 4
5694 // CHECK10-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5695 // CHECK10-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
5696 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
5697 // CHECK10-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5698 // CHECK10-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
5699 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
5700 // CHECK10-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
5701 // CHECK10-NEXT: store i8* null, i8** [[TMP4]], align 8
5702 // CHECK10-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5703 // CHECK10-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5704 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
5705 // CHECK10-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86.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.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5706 // CHECK10-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
5707 // CHECK10-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5708 // CHECK10: omp_offload.failed:
5709 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86([10 x i32]* [[A]]) #[[ATTR3]]
5710 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT]]
5711 // CHECK10: omp_offload.cont:
5712 // CHECK10-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
5713 // CHECK10-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
5714 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 8
5715 // CHECK10-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
5716 // CHECK10-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
5717 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 8
5718 // CHECK10-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
5719 // CHECK10-NEXT: store i8* null, i8** [[TMP13]], align 8
5720 // CHECK10-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
5721 // CHECK10-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
5722 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
5723 // CHECK10-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.13, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.14, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5724 // CHECK10-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
5725 // CHECK10-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
5726 // CHECK10: omp_offload.failed5:
5727 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91([10 x i32]* [[A]]) #[[ATTR3]]
5728 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT6]]
5729 // CHECK10: omp_offload.cont6:
5730 // CHECK10-NEXT: [[TMP18:%.*]] = load i32, i32* [[M]], align 4
5731 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
5732 // CHECK10-NEXT: store i32 [[TMP18]], i32* [[CONV]], align 4
5733 // CHECK10-NEXT: [[TMP19:%.*]] = load i64, i64* [[M_CASTED]], align 8
5734 // CHECK10-NEXT: [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
5735 // CHECK10-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i64*
5736 // CHECK10-NEXT: store i64 [[TMP19]], i64* [[TMP21]], align 8
5737 // CHECK10-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
5738 // CHECK10-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i64*
5739 // CHECK10-NEXT: store i64 [[TMP19]], i64* [[TMP23]], align 8
5740 // CHECK10-NEXT: [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 0
5741 // CHECK10-NEXT: store i8* null, i8** [[TMP24]], align 8
5742 // CHECK10-NEXT: [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
5743 // CHECK10-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
5744 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 8
5745 // CHECK10-NEXT: [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
5746 // CHECK10-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
5747 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 8
5748 // CHECK10-NEXT: [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i64 0, i64 1
5749 // CHECK10-NEXT: store i8* null, i8** [[TMP29]], align 8
5750 // CHECK10-NEXT: [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
5751 // CHECK10-NEXT: [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
5752 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
5753 // CHECK10-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.17, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.18, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
5754 // CHECK10-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
5755 // CHECK10-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
5756 // CHECK10: omp_offload.failed11:
5757 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96(i64 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
5758 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT12]]
5759 // CHECK10: omp_offload.cont12:
5760 // CHECK10-NEXT: ret i32 0
5761 //
5762 //
5763 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86
5764 // CHECK10-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
5765 // CHECK10-NEXT: entry:
5766 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
5767 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
5768 // CHECK10-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
5769 // CHECK10-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..8 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
5770 // CHECK10-NEXT: ret void
5771 //
5772 //
5773 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..8
5774 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
5775 // CHECK10-NEXT: entry:
5776 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5777 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5778 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
5779 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5780 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
5781 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5782 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5783 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5784 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5785 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
5786 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5787 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5788 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
5789 // CHECK10-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
5790 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5791 // CHECK10-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
5792 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5793 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5794 // CHECK10-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5795 // CHECK10-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5796 // CHECK10-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)
5797 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5798 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
5799 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5800 // CHECK10: cond.true:
5801 // CHECK10-NEXT: br label [[COND_END:%.*]]
5802 // CHECK10: cond.false:
5803 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5804 // CHECK10-NEXT: br label [[COND_END]]
5805 // CHECK10: cond.end:
5806 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5807 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5808 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5809 // CHECK10-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
5810 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5811 // CHECK10: omp.inner.for.cond:
5812 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5813 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5814 // CHECK10-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5815 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5816 // CHECK10: omp.inner.for.body:
5817 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5818 // CHECK10-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
5819 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5820 // CHECK10-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
5821 // CHECK10-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..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
5822 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5823 // CHECK10: omp.inner.for.inc:
5824 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5825 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5826 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5827 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5828 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
5829 // CHECK10: omp.inner.for.end:
5830 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5831 // CHECK10: omp.loop.exit:
5832 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
5833 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5834 // CHECK10-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
5835 // CHECK10-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5836 // CHECK10: .omp.final.then:
5837 // CHECK10-NEXT: store i32 10, i32* [[I]], align 4
5838 // CHECK10-NEXT: br label [[DOTOMP_FINAL_DONE]]
5839 // CHECK10: .omp.final.done:
5840 // CHECK10-NEXT: ret void
5841 //
5842 //
5843 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..9
5844 // CHECK10-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]] {
5845 // CHECK10-NEXT: entry:
5846 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5847 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5848 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5849 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5850 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
5851 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5852 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
5853 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
5854 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
5855 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5856 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5857 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
5858 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5859 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5860 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5861 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5862 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
5863 // CHECK10-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
5864 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
5865 // CHECK10-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
5866 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5867 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
5868 // CHECK10-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5869 // CHECK10-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
5870 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5871 // CHECK10-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5872 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5873 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5874 // CHECK10-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5875 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
5876 // CHECK10-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)
5877 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5878 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
5879 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5880 // CHECK10: cond.true:
5881 // CHECK10-NEXT: br label [[COND_END:%.*]]
5882 // CHECK10: cond.false:
5883 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5884 // CHECK10-NEXT: br label [[COND_END]]
5885 // CHECK10: cond.end:
5886 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
5887 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5888 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5889 // CHECK10-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
5890 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5891 // CHECK10: omp.inner.for.cond:
5892 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5893 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5894 // CHECK10-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
5895 // CHECK10-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5896 // CHECK10: omp.inner.for.body:
5897 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5898 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
5899 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5900 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4
5901 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
5902 // CHECK10-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
5903 // CHECK10-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
5904 // CHECK10-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
5905 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
5906 // CHECK10: omp.body.continue:
5907 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5908 // CHECK10: omp.inner.for.inc:
5909 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5910 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
5911 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
5912 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
5913 // CHECK10: omp.inner.for.end:
5914 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5915 // CHECK10: omp.loop.exit:
5916 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
5917 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5918 // CHECK10-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
5919 // CHECK10-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5920 // CHECK10: .omp.final.then:
5921 // CHECK10-NEXT: store i32 10, i32* [[I]], align 4
5922 // CHECK10-NEXT: br label [[DOTOMP_FINAL_DONE]]
5923 // CHECK10: .omp.final.done:
5924 // CHECK10-NEXT: ret void
5925 //
5926 //
5927 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91
5928 // CHECK10-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
5929 // CHECK10-NEXT: entry:
5930 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
5931 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
5932 // CHECK10-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
5933 // CHECK10-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..11 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
5934 // CHECK10-NEXT: ret void
5935 //
5936 //
5937 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..11
5938 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
5939 // CHECK10-NEXT: entry:
5940 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5941 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5942 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
5943 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
5944 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
5945 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5946 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5947 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5948 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5949 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
5950 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5951 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5952 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
5953 // CHECK10-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
5954 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5955 // CHECK10-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
5956 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5957 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5958 // CHECK10-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5959 // CHECK10-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5960 // CHECK10-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)
5961 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5962 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
5963 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5964 // CHECK10: cond.true:
5965 // CHECK10-NEXT: br label [[COND_END:%.*]]
5966 // CHECK10: cond.false:
5967 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5968 // CHECK10-NEXT: br label [[COND_END]]
5969 // CHECK10: cond.end:
5970 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5971 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5972 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5973 // CHECK10-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
5974 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
5975 // CHECK10: omp.inner.for.cond:
5976 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5977 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5978 // CHECK10-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5979 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5980 // CHECK10: omp.inner.for.body:
5981 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5982 // CHECK10-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
5983 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5984 // CHECK10-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
5985 // CHECK10-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..12 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
5986 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
5987 // CHECK10: omp.inner.for.inc:
5988 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5989 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5990 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5991 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5992 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
5993 // CHECK10: omp.inner.for.end:
5994 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
5995 // CHECK10: omp.loop.exit:
5996 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
5997 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5998 // CHECK10-NEXT: [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
5999 // CHECK10-NEXT: br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6000 // CHECK10: .omp.final.then:
6001 // CHECK10-NEXT: store i32 10, i32* [[I]], align 4
6002 // CHECK10-NEXT: br label [[DOTOMP_FINAL_DONE]]
6003 // CHECK10: .omp.final.done:
6004 // CHECK10-NEXT: ret void
6005 //
6006 //
6007 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..12
6008 // CHECK10-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]] {
6009 // CHECK10-NEXT: entry:
6010 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6011 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6012 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6013 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6014 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
6015 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6016 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
6017 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6018 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6019 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6020 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6021 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
6022 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6023 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6024 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6025 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6026 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
6027 // CHECK10-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
6028 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6029 // CHECK10-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
6030 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6031 // CHECK10-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
6032 // CHECK10-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6033 // CHECK10-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
6034 // CHECK10-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
6035 // CHECK10-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
6036 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6037 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6038 // CHECK10-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6039 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
6040 // CHECK10-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)
6041 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6042 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
6043 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6044 // CHECK10: cond.true:
6045 // CHECK10-NEXT: br label [[COND_END:%.*]]
6046 // CHECK10: cond.false:
6047 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6048 // CHECK10-NEXT: br label [[COND_END]]
6049 // CHECK10: cond.end:
6050 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
6051 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6052 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6053 // CHECK10-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
6054 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6055 // CHECK10: omp.inner.for.cond:
6056 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6057 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6058 // CHECK10-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
6059 // CHECK10-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6060 // CHECK10: omp.inner.for.body:
6061 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6062 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
6063 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6064 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4
6065 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
6066 // CHECK10-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
6067 // CHECK10-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
6068 // CHECK10-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
6069 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6070 // CHECK10: omp.body.continue:
6071 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6072 // CHECK10: omp.inner.for.inc:
6073 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6074 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
6075 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
6076 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
6077 // CHECK10: omp.inner.for.end:
6078 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6079 // CHECK10: omp.loop.exit:
6080 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
6081 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6082 // CHECK10-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
6083 // CHECK10-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6084 // CHECK10: .omp.final.then:
6085 // CHECK10-NEXT: store i32 10, i32* [[I]], align 4
6086 // CHECK10-NEXT: br label [[DOTOMP_FINAL_DONE]]
6087 // CHECK10: .omp.final.done:
6088 // CHECK10-NEXT: ret void
6089 //
6090 //
6091 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96
6092 // CHECK10-SAME: (i64 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
6093 // CHECK10-NEXT: entry:
6094 // CHECK10-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
6095 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
6096 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
6097 // CHECK10-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
6098 // CHECK10-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
6099 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
6100 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
6101 // CHECK10-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
6102 // CHECK10-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
6103 // CHECK10-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
6104 // CHECK10-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6105 // CHECK10-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
6106 // CHECK10-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
6107 // CHECK10-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
6108 // CHECK10-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..15 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
6109 // CHECK10-NEXT: ret void
6110 //
6111 //
6112 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..15
6113 // CHECK10-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i64 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
6114 // CHECK10-NEXT: entry:
6115 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6116 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6117 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
6118 // CHECK10-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
6119 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6120 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
6121 // CHECK10-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6122 // CHECK10-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6123 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6124 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6125 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
6126 // CHECK10-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
6127 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6128 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6129 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
6130 // CHECK10-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
6131 // CHECK10-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
6132 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
6133 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6134 // CHECK10-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
6135 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6136 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6137 // CHECK10-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 8
6138 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6139 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
6140 // CHECK10-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP1]])
6141 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6142 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
6143 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6144 // CHECK10: cond.true:
6145 // CHECK10-NEXT: br label [[COND_END:%.*]]
6146 // CHECK10: cond.false:
6147 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6148 // CHECK10-NEXT: br label [[COND_END]]
6149 // CHECK10: cond.end:
6150 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6151 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6152 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6153 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6154 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6155 // CHECK10: omp.inner.for.cond:
6156 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6157 // CHECK10-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP7]], 10
6158 // CHECK10-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6159 // CHECK10: omp.inner.for.body:
6160 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6161 // CHECK10-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
6162 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6163 // CHECK10-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
6164 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[CONV]], align 8
6165 // CHECK10-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
6166 // CHECK10-NEXT: store i32 [[TMP12]], i32* [[CONV2]], align 4
6167 // CHECK10-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
6168 // CHECK10-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..16 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
6169 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6170 // CHECK10: omp.inner.for.inc:
6171 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6172 // CHECK10-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6173 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
6174 // CHECK10-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6175 // CHECK10-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6176 // CHECK10-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6177 // CHECK10-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
6178 // CHECK10-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_LB]], align 4
6179 // CHECK10-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6180 // CHECK10-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6181 // CHECK10-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
6182 // CHECK10-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_COMB_UB]], align 4
6183 // CHECK10-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6184 // CHECK10-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP20]], 9
6185 // CHECK10-NEXT: br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
6186 // CHECK10: cond.true6:
6187 // CHECK10-NEXT: br label [[COND_END8:%.*]]
6188 // CHECK10: cond.false7:
6189 // CHECK10-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6190 // CHECK10-NEXT: br label [[COND_END8]]
6191 // CHECK10: cond.end8:
6192 // CHECK10-NEXT: [[COND9:%.*]] = phi i32 [ 9, [[COND_TRUE6]] ], [ [[TMP21]], [[COND_FALSE7]] ]
6193 // CHECK10-NEXT: store i32 [[COND9]], i32* [[DOTOMP_COMB_UB]], align 4
6194 // CHECK10-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6195 // CHECK10-NEXT: store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
6196 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
6197 // CHECK10: omp.inner.for.end:
6198 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6199 // CHECK10: omp.loop.exit:
6200 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6201 // CHECK10-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6202 // CHECK10-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
6203 // CHECK10-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6204 // CHECK10: .omp.final.then:
6205 // CHECK10-NEXT: store i32 10, i32* [[I]], align 4
6206 // CHECK10-NEXT: br label [[DOTOMP_FINAL_DONE]]
6207 // CHECK10: .omp.final.done:
6208 // CHECK10-NEXT: ret void
6209 //
6210 //
6211 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..16
6212 // CHECK10-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]] {
6213 // CHECK10-NEXT: entry:
6214 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6215 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6216 // CHECK10-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6217 // CHECK10-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6218 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
6219 // CHECK10-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
6220 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6221 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
6222 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6223 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6224 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6225 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6226 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
6227 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6228 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6229 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6230 // CHECK10-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6231 // CHECK10-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
6232 // CHECK10-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
6233 // CHECK10-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
6234 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
6235 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6236 // CHECK10-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
6237 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6238 // CHECK10-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
6239 // CHECK10-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6240 // CHECK10-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
6241 // CHECK10-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
6242 // CHECK10-NEXT: store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
6243 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6244 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6245 // CHECK10-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6246 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
6247 // CHECK10-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)
6248 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6249 // CHECK10-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
6250 // CHECK10-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6251 // CHECK10: cond.true:
6252 // CHECK10-NEXT: br label [[COND_END:%.*]]
6253 // CHECK10: cond.false:
6254 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6255 // CHECK10-NEXT: br label [[COND_END]]
6256 // CHECK10: cond.end:
6257 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
6258 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6259 // CHECK10-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6260 // CHECK10-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
6261 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6262 // CHECK10: omp.inner.for.cond:
6263 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6264 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6265 // CHECK10-NEXT: [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
6266 // CHECK10-NEXT: br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6267 // CHECK10: omp.inner.for.body:
6268 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6269 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
6270 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6271 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I]], align 4
6272 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
6273 // CHECK10-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
6274 // CHECK10-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
6275 // CHECK10-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
6276 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6277 // CHECK10: omp.body.continue:
6278 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6279 // CHECK10: omp.inner.for.inc:
6280 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6281 // CHECK10-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
6282 // CHECK10-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4
6283 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
6284 // CHECK10: omp.inner.for.end:
6285 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6286 // CHECK10: omp.loop.exit:
6287 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
6288 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6289 // CHECK10-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
6290 // CHECK10-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6291 // CHECK10: .omp.final.then:
6292 // CHECK10-NEXT: store i32 10, i32* [[I]], align 4
6293 // CHECK10-NEXT: br label [[DOTOMP_FINAL_DONE]]
6294 // CHECK10: .omp.final.done:
6295 // CHECK10-NEXT: ret void
6296 //
6297 //
6298 // CHECK10-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
6299 // CHECK10-SAME: () #[[ATTR5:[0-9]+]] {
6300 // CHECK10-NEXT: entry:
6301 // CHECK10-NEXT: call void @__tgt_register_requires(i64 1)
6302 // CHECK10-NEXT: ret void
6303 //
6304 //
6305 // CHECK11-LABEL: define {{[^@]+}}@main
6306 // CHECK11-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
6307 // CHECK11-NEXT: entry:
6308 // CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
6309 // CHECK11-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
6310 // CHECK11-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
6311 // CHECK11-NEXT: [[N:%.*]] = alloca i32, align 4
6312 // CHECK11-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
6313 // CHECK11-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
6314 // CHECK11-NEXT: [[M:%.*]] = alloca i32, align 4
6315 // CHECK11-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
6316 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
6317 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
6318 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
6319 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
6320 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
6321 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
6322 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
6323 // CHECK11-NEXT: [[N_CASTED3:%.*]] = alloca i32, align 4
6324 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
6325 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
6326 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
6327 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
6328 // CHECK11-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
6329 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
6330 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
6331 // CHECK11-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
6332 // CHECK11-NEXT: [[N_CASTED17:%.*]] = alloca i32, align 4
6333 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS18:%.*]] = alloca [4 x i8*], align 4
6334 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS19:%.*]] = alloca [4 x i8*], align 4
6335 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS20:%.*]] = alloca [4 x i8*], align 4
6336 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES21:%.*]] = alloca [4 x i64], align 4
6337 // CHECK11-NEXT: [[_TMP22:%.*]] = alloca i32, align 4
6338 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_23:%.*]] = alloca i32, align 4
6339 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
6340 // CHECK11-NEXT: store i32 0, i32* [[RETVAL]], align 4
6341 // CHECK11-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
6342 // CHECK11-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
6343 // CHECK11-NEXT: store i32 100, i32* [[N]], align 4
6344 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
6345 // CHECK11-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
6346 // CHECK11-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
6347 // CHECK11-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
6348 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
6349 // CHECK11-NEXT: store i32 10, i32* [[M]], align 4
6350 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
6351 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
6352 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
6353 // CHECK11-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
6354 // CHECK11-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
6355 // CHECK11-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6356 // CHECK11-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
6357 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP7]], align 4
6358 // CHECK11-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6359 // CHECK11-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
6360 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP9]], align 4
6361 // CHECK11-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
6362 // CHECK11-NEXT: store i64 4, i64* [[TMP10]], align 4
6363 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
6364 // CHECK11-NEXT: store i8* null, i8** [[TMP11]], align 4
6365 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
6366 // CHECK11-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
6367 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
6368 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
6369 // CHECK11-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
6370 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
6371 // CHECK11-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
6372 // CHECK11-NEXT: store i64 4, i64* [[TMP16]], align 4
6373 // CHECK11-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
6374 // CHECK11-NEXT: store i8* null, i8** [[TMP17]], align 4
6375 // CHECK11-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
6376 // CHECK11-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
6377 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 4
6378 // CHECK11-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
6379 // CHECK11-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
6380 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 4
6381 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
6382 // CHECK11-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 4
6383 // CHECK11-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
6384 // CHECK11-NEXT: store i8* null, i8** [[TMP23]], align 4
6385 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6386 // CHECK11-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6387 // CHECK11-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
6388 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
6389 // CHECK11-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
6390 // CHECK11-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6391 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
6392 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
6393 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
6394 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
6395 // CHECK11-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6396 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
6397 // CHECK11-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
6398 // CHECK11-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
6399 // CHECK11-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108.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)
6400 // CHECK11-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
6401 // CHECK11-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6402 // CHECK11: omp_offload.failed:
6403 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
6404 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
6405 // CHECK11: omp_offload.cont:
6406 // CHECK11-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
6407 // CHECK11-NEXT: store i32 [[TMP33]], i32* [[N_CASTED3]], align 4
6408 // CHECK11-NEXT: [[TMP34:%.*]] = load i32, i32* [[N_CASTED3]], align 4
6409 // CHECK11-NEXT: [[TMP35:%.*]] = mul nuw i32 [[TMP0]], 4
6410 // CHECK11-NEXT: [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
6411 // CHECK11-NEXT: [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
6412 // CHECK11-NEXT: [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
6413 // CHECK11-NEXT: store i32 [[TMP34]], i32* [[TMP38]], align 4
6414 // CHECK11-NEXT: [[TMP39:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
6415 // CHECK11-NEXT: [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
6416 // CHECK11-NEXT: store i32 [[TMP34]], i32* [[TMP40]], align 4
6417 // CHECK11-NEXT: [[TMP41:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
6418 // CHECK11-NEXT: store i64 4, i64* [[TMP41]], align 4
6419 // CHECK11-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
6420 // CHECK11-NEXT: store i8* null, i8** [[TMP42]], align 4
6421 // CHECK11-NEXT: [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
6422 // CHECK11-NEXT: [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i32*
6423 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP44]], align 4
6424 // CHECK11-NEXT: [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
6425 // CHECK11-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
6426 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP46]], align 4
6427 // CHECK11-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 1
6428 // CHECK11-NEXT: store i64 4, i64* [[TMP47]], align 4
6429 // CHECK11-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
6430 // CHECK11-NEXT: store i8* null, i8** [[TMP48]], align 4
6431 // CHECK11-NEXT: [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
6432 // CHECK11-NEXT: [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32**
6433 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP50]], align 4
6434 // CHECK11-NEXT: [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
6435 // CHECK11-NEXT: [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32**
6436 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP52]], align 4
6437 // CHECK11-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
6438 // CHECK11-NEXT: store i64 [[TMP36]], i64* [[TMP53]], align 4
6439 // CHECK11-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
6440 // CHECK11-NEXT: store i8* null, i8** [[TMP54]], align 4
6441 // CHECK11-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
6442 // CHECK11-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
6443 // CHECK11-NEXT: [[TMP57:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
6444 // CHECK11-NEXT: [[TMP58:%.*]] = load i32, i32* [[N]], align 4
6445 // CHECK11-NEXT: store i32 [[TMP58]], i32* [[DOTCAPTURE_EXPR_9]], align 4
6446 // CHECK11-NEXT: [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
6447 // CHECK11-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP59]], 0
6448 // CHECK11-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
6449 // CHECK11-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
6450 // CHECK11-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
6451 // CHECK11-NEXT: [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
6452 // CHECK11-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP60]], 1
6453 // CHECK11-NEXT: [[TMP61:%.*]] = zext i32 [[ADD14]] to i64
6454 // CHECK11-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP61]])
6455 // CHECK11-NEXT: [[TMP62:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113.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)
6456 // CHECK11-NEXT: [[TMP63:%.*]] = icmp ne i32 [[TMP62]], 0
6457 // CHECK11-NEXT: br i1 [[TMP63]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
6458 // CHECK11: omp_offload.failed15:
6459 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113(i32 [[TMP34]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
6460 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT16]]
6461 // CHECK11: omp_offload.cont16:
6462 // CHECK11-NEXT: [[TMP64:%.*]] = load i32, i32* [[M]], align 4
6463 // CHECK11-NEXT: store i32 [[TMP64]], i32* [[M_CASTED]], align 4
6464 // CHECK11-NEXT: [[TMP65:%.*]] = load i32, i32* [[M_CASTED]], align 4
6465 // CHECK11-NEXT: [[TMP66:%.*]] = load i32, i32* [[N]], align 4
6466 // CHECK11-NEXT: store i32 [[TMP66]], i32* [[N_CASTED17]], align 4
6467 // CHECK11-NEXT: [[TMP67:%.*]] = load i32, i32* [[N_CASTED17]], align 4
6468 // CHECK11-NEXT: [[TMP68:%.*]] = mul nuw i32 [[TMP0]], 4
6469 // CHECK11-NEXT: [[TMP69:%.*]] = sext i32 [[TMP68]] to i64
6470 // CHECK11-NEXT: [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
6471 // CHECK11-NEXT: [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i32*
6472 // CHECK11-NEXT: store i32 [[TMP65]], i32* [[TMP71]], align 4
6473 // CHECK11-NEXT: [[TMP72:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
6474 // CHECK11-NEXT: [[TMP73:%.*]] = bitcast i8** [[TMP72]] to i32*
6475 // CHECK11-NEXT: store i32 [[TMP65]], i32* [[TMP73]], align 4
6476 // CHECK11-NEXT: [[TMP74:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
6477 // CHECK11-NEXT: store i64 4, i64* [[TMP74]], align 4
6478 // CHECK11-NEXT: [[TMP75:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 0
6479 // CHECK11-NEXT: store i8* null, i8** [[TMP75]], align 4
6480 // CHECK11-NEXT: [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 1
6481 // CHECK11-NEXT: [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i32*
6482 // CHECK11-NEXT: store i32 [[TMP67]], i32* [[TMP77]], align 4
6483 // CHECK11-NEXT: [[TMP78:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 1
6484 // CHECK11-NEXT: [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
6485 // CHECK11-NEXT: store i32 [[TMP67]], i32* [[TMP79]], align 4
6486 // CHECK11-NEXT: [[TMP80:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 1
6487 // CHECK11-NEXT: store i64 4, i64* [[TMP80]], align 4
6488 // CHECK11-NEXT: [[TMP81:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 1
6489 // CHECK11-NEXT: store i8* null, i8** [[TMP81]], align 4
6490 // CHECK11-NEXT: [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 2
6491 // CHECK11-NEXT: [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i32*
6492 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP83]], align 4
6493 // CHECK11-NEXT: [[TMP84:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 2
6494 // CHECK11-NEXT: [[TMP85:%.*]] = bitcast i8** [[TMP84]] to i32*
6495 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP85]], align 4
6496 // CHECK11-NEXT: [[TMP86:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 2
6497 // CHECK11-NEXT: store i64 4, i64* [[TMP86]], align 4
6498 // CHECK11-NEXT: [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 2
6499 // CHECK11-NEXT: store i8* null, i8** [[TMP87]], align 4
6500 // CHECK11-NEXT: [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 3
6501 // CHECK11-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
6502 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP89]], align 4
6503 // CHECK11-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 3
6504 // CHECK11-NEXT: [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i32**
6505 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP91]], align 4
6506 // CHECK11-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 3
6507 // CHECK11-NEXT: store i64 [[TMP69]], i64* [[TMP92]], align 4
6508 // CHECK11-NEXT: [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 3
6509 // CHECK11-NEXT: store i8* null, i8** [[TMP93]], align 4
6510 // CHECK11-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
6511 // CHECK11-NEXT: [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
6512 // CHECK11-NEXT: [[TMP96:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
6513 // CHECK11-NEXT: [[TMP97:%.*]] = load i32, i32* [[N]], align 4
6514 // CHECK11-NEXT: store i32 [[TMP97]], i32* [[DOTCAPTURE_EXPR_23]], align 4
6515 // CHECK11-NEXT: [[TMP98:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_23]], align 4
6516 // CHECK11-NEXT: [[SUB25:%.*]] = sub nsw i32 [[TMP98]], 0
6517 // CHECK11-NEXT: [[DIV26:%.*]] = sdiv i32 [[SUB25]], 1
6518 // CHECK11-NEXT: [[SUB27:%.*]] = sub nsw i32 [[DIV26]], 1
6519 // CHECK11-NEXT: store i32 [[SUB27]], i32* [[DOTCAPTURE_EXPR_24]], align 4
6520 // CHECK11-NEXT: [[TMP99:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
6521 // CHECK11-NEXT: [[ADD28:%.*]] = add nsw i32 [[TMP99]], 1
6522 // CHECK11-NEXT: [[TMP100:%.*]] = zext i32 [[ADD28]] to i64
6523 // CHECK11-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP100]])
6524 // CHECK11-NEXT: [[TMP101:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118.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)
6525 // CHECK11-NEXT: [[TMP102:%.*]] = icmp ne i32 [[TMP101]], 0
6526 // CHECK11-NEXT: br i1 [[TMP102]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
6527 // CHECK11: omp_offload.failed29:
6528 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118(i32 [[TMP65]], i32 [[TMP67]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
6529 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT30]]
6530 // CHECK11: omp_offload.cont30:
6531 // CHECK11-NEXT: [[TMP103:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
6532 // CHECK11-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP103]])
6533 // CHECK11-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
6534 // CHECK11-NEXT: [[TMP104:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
6535 // CHECK11-NEXT: call void @llvm.stackrestore(i8* [[TMP104]])
6536 // CHECK11-NEXT: [[TMP105:%.*]] = load i32, i32* [[RETVAL]], align 4
6537 // CHECK11-NEXT: ret i32 [[TMP105]]
6538 //
6539 //
6540 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108
6541 // CHECK11-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
6542 // CHECK11-NEXT: entry:
6543 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
6544 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
6545 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
6546 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
6547 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6548 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
6549 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6550 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
6551 // CHECK11-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]])
6552 // CHECK11-NEXT: ret void
6553 //
6554 //
6555 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
6556 // CHECK11-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]] {
6557 // CHECK11-NEXT: entry:
6558 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6559 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6560 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
6561 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
6562 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
6563 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6564 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
6565 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
6566 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
6567 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
6568 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6569 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6570 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6571 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6572 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
6573 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6574 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6575 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
6576 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6577 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
6578 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
6579 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6580 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
6581 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
6582 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
6583 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6584 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
6585 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
6586 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
6587 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
6588 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
6589 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6590 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
6591 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
6592 // CHECK11: omp.precond.then:
6593 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6594 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6595 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
6596 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6597 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6598 // CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6599 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
6600 // CHECK11-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)
6601 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6602 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6603 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
6604 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6605 // CHECK11: cond.true:
6606 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6607 // CHECK11-NEXT: br label [[COND_END:%.*]]
6608 // CHECK11: cond.false:
6609 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6610 // CHECK11-NEXT: br label [[COND_END]]
6611 // CHECK11: cond.end:
6612 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
6613 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6614 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6615 // CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
6616 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6617 // CHECK11: omp.inner.for.cond:
6618 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6619 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6620 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
6621 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6622 // CHECK11: omp.inner.for.body:
6623 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6624 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6625 // CHECK11-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]])
6626 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6627 // CHECK11: omp.inner.for.inc:
6628 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6629 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6630 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
6631 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6632 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
6633 // CHECK11: omp.inner.for.end:
6634 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6635 // CHECK11: omp.loop.exit:
6636 // CHECK11-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6637 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
6638 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
6639 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6640 // CHECK11-NEXT: [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
6641 // CHECK11-NEXT: br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6642 // CHECK11: .omp.final.then:
6643 // CHECK11-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6644 // CHECK11-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
6645 // CHECK11-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
6646 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
6647 // CHECK11-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
6648 // CHECK11-NEXT: store i32 [[ADD8]], i32* [[I3]], align 4
6649 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
6650 // CHECK11: .omp.final.done:
6651 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
6652 // CHECK11: omp.precond.end:
6653 // CHECK11-NEXT: ret void
6654 //
6655 //
6656 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
6657 // CHECK11-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]] {
6658 // CHECK11-NEXT: entry:
6659 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6660 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6661 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
6662 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
6663 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
6664 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
6665 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
6666 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6667 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
6668 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
6669 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
6670 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
6671 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6672 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6673 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6674 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6675 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
6676 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6677 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6678 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6679 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6680 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
6681 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6682 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
6683 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
6684 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6685 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
6686 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
6687 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
6688 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6689 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
6690 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
6691 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
6692 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
6693 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
6694 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6695 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
6696 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
6697 // CHECK11: omp.precond.then:
6698 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6699 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6700 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
6701 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6702 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6703 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
6704 // CHECK11-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
6705 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6706 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6707 // CHECK11-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6708 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
6709 // CHECK11-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)
6710 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6711 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6712 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
6713 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6714 // CHECK11: cond.true:
6715 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6716 // CHECK11-NEXT: br label [[COND_END:%.*]]
6717 // CHECK11: cond.false:
6718 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6719 // CHECK11-NEXT: br label [[COND_END]]
6720 // CHECK11: cond.end:
6721 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
6722 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6723 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6724 // CHECK11-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
6725 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6726 // CHECK11: omp.inner.for.cond:
6727 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6728 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6729 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
6730 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6731 // CHECK11: omp.inner.for.body:
6732 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6733 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
6734 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6735 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
6736 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
6737 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
6738 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
6739 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6740 // CHECK11: omp.body.continue:
6741 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6742 // CHECK11: omp.inner.for.inc:
6743 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6744 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
6745 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
6746 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
6747 // CHECK11: omp.inner.for.end:
6748 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6749 // CHECK11: omp.loop.exit:
6750 // CHECK11-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6751 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
6752 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
6753 // CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6754 // CHECK11-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
6755 // CHECK11-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6756 // CHECK11: .omp.final.then:
6757 // CHECK11-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6758 // CHECK11-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
6759 // CHECK11-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
6760 // CHECK11-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
6761 // CHECK11-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
6762 // CHECK11-NEXT: store i32 [[ADD10]], i32* [[I3]], align 4
6763 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
6764 // CHECK11: .omp.final.done:
6765 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
6766 // CHECK11: omp.precond.end:
6767 // CHECK11-NEXT: ret void
6768 //
6769 //
6770 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113
6771 // CHECK11-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
6772 // CHECK11-NEXT: entry:
6773 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
6774 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
6775 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
6776 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
6777 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6778 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
6779 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6780 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
6781 // CHECK11-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]])
6782 // CHECK11-NEXT: ret void
6783 //
6784 //
6785 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
6786 // CHECK11-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]] {
6787 // CHECK11-NEXT: entry:
6788 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6789 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6790 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
6791 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
6792 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
6793 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6794 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
6795 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
6796 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
6797 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
6798 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6799 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6800 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6801 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6802 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
6803 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6804 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6805 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
6806 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6807 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
6808 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
6809 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6810 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
6811 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
6812 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
6813 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6814 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
6815 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
6816 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
6817 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
6818 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
6819 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6820 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
6821 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
6822 // CHECK11: omp.precond.then:
6823 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6824 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6825 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
6826 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6827 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6828 // CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6829 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
6830 // CHECK11-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)
6831 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6832 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6833 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
6834 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6835 // CHECK11: cond.true:
6836 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6837 // CHECK11-NEXT: br label [[COND_END:%.*]]
6838 // CHECK11: cond.false:
6839 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6840 // CHECK11-NEXT: br label [[COND_END]]
6841 // CHECK11: cond.end:
6842 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
6843 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6844 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6845 // CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
6846 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6847 // CHECK11: omp.inner.for.cond:
6848 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6849 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6850 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
6851 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6852 // CHECK11: omp.inner.for.body:
6853 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6854 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6855 // CHECK11-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]])
6856 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6857 // CHECK11: omp.inner.for.inc:
6858 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6859 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6860 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
6861 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6862 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
6863 // CHECK11: omp.inner.for.end:
6864 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6865 // CHECK11: omp.loop.exit:
6866 // CHECK11-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6867 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
6868 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
6869 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6870 // CHECK11-NEXT: [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
6871 // CHECK11-NEXT: br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6872 // CHECK11: .omp.final.then:
6873 // CHECK11-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6874 // CHECK11-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
6875 // CHECK11-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
6876 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
6877 // CHECK11-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
6878 // CHECK11-NEXT: store i32 [[ADD8]], i32* [[I3]], align 4
6879 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
6880 // CHECK11: .omp.final.done:
6881 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
6882 // CHECK11: omp.precond.end:
6883 // CHECK11-NEXT: ret void
6884 //
6885 //
6886 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
6887 // CHECK11-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]] {
6888 // CHECK11-NEXT: entry:
6889 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
6890 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
6891 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
6892 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
6893 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
6894 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
6895 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
6896 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
6897 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
6898 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
6899 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
6900 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
6901 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
6902 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
6903 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6904 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6905 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
6906 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
6907 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
6908 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6909 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6910 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
6911 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
6912 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
6913 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
6914 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
6915 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
6916 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
6917 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
6918 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6919 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
6920 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
6921 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
6922 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
6923 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
6924 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6925 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
6926 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
6927 // CHECK11: omp.precond.then:
6928 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
6929 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6930 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
6931 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
6932 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
6933 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
6934 // CHECK11-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
6935 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6936 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6937 // CHECK11-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6938 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
6939 // CHECK11-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)
6940 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6941 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6942 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
6943 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6944 // CHECK11: cond.true:
6945 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
6946 // CHECK11-NEXT: br label [[COND_END:%.*]]
6947 // CHECK11: cond.false:
6948 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6949 // CHECK11-NEXT: br label [[COND_END]]
6950 // CHECK11: cond.end:
6951 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
6952 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6953 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6954 // CHECK11-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
6955 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
6956 // CHECK11: omp.inner.for.cond:
6957 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6958 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6959 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
6960 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6961 // CHECK11: omp.inner.for.body:
6962 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6963 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
6964 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6965 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
6966 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
6967 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
6968 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
6969 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
6970 // CHECK11: omp.body.continue:
6971 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
6972 // CHECK11: omp.inner.for.inc:
6973 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6974 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
6975 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
6976 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
6977 // CHECK11: omp.inner.for.end:
6978 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
6979 // CHECK11: omp.loop.exit:
6980 // CHECK11-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
6981 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
6982 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
6983 // CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6984 // CHECK11-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
6985 // CHECK11-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6986 // CHECK11: .omp.final.then:
6987 // CHECK11-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
6988 // CHECK11-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
6989 // CHECK11-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
6990 // CHECK11-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
6991 // CHECK11-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
6992 // CHECK11-NEXT: store i32 [[ADD10]], i32* [[I3]], align 4
6993 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
6994 // CHECK11: .omp.final.done:
6995 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
6996 // CHECK11: omp.precond.end:
6997 // CHECK11-NEXT: ret void
6998 //
6999 //
7000 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118
7001 // CHECK11-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
7002 // CHECK11-NEXT: entry:
7003 // CHECK11-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
7004 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
7005 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
7006 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
7007 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
7008 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
7009 // CHECK11-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
7010 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
7011 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7012 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
7013 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7014 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
7015 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
7016 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
7017 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7018 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
7019 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
7020 // CHECK11-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]])
7021 // CHECK11-NEXT: ret void
7022 //
7023 //
7024 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
7025 // CHECK11-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]] {
7026 // CHECK11-NEXT: entry:
7027 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7028 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7029 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
7030 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
7031 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
7032 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
7033 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7034 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
7035 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
7036 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
7037 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
7038 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7039 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7040 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7041 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7042 // CHECK11-NEXT: [[I4:%.*]] = alloca i32, align 4
7043 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
7044 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7045 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7046 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
7047 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7048 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
7049 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
7050 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
7051 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7052 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
7053 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
7054 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
7055 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7056 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
7057 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
7058 // CHECK11-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
7059 // CHECK11-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
7060 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
7061 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7062 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
7063 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
7064 // CHECK11: omp.precond.then:
7065 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7066 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
7067 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
7068 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7069 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7070 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
7071 // CHECK11-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7072 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
7073 // CHECK11-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]])
7074 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7075 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
7076 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
7077 // CHECK11-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7078 // CHECK11: cond.true:
7079 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
7080 // CHECK11-NEXT: br label [[COND_END:%.*]]
7081 // CHECK11: cond.false:
7082 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7083 // CHECK11-NEXT: br label [[COND_END]]
7084 // CHECK11: cond.end:
7085 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
7086 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7087 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7088 // CHECK11-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
7089 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7090 // CHECK11: omp.inner.for.cond:
7091 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7092 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
7093 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
7094 // CHECK11-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
7095 // CHECK11-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7096 // CHECK11: omp.inner.for.body:
7097 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7098 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7099 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
7100 // CHECK11-NEXT: store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
7101 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
7102 // CHECK11-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]])
7103 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7104 // CHECK11: omp.inner.for.inc:
7105 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7106 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7107 // CHECK11-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
7108 // CHECK11-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
7109 // CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7110 // CHECK11-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7111 // CHECK11-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
7112 // CHECK11-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4
7113 // CHECK11-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7114 // CHECK11-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7115 // CHECK11-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
7116 // CHECK11-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4
7117 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7118 // CHECK11-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
7119 // CHECK11-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
7120 // CHECK11-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
7121 // CHECK11: cond.true11:
7122 // CHECK11-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
7123 // CHECK11-NEXT: br label [[COND_END13:%.*]]
7124 // CHECK11: cond.false12:
7125 // CHECK11-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7126 // CHECK11-NEXT: br label [[COND_END13]]
7127 // CHECK11: cond.end13:
7128 // CHECK11-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
7129 // CHECK11-NEXT: store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4
7130 // CHECK11-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7131 // CHECK11-NEXT: store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
7132 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
7133 // CHECK11: omp.inner.for.end:
7134 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7135 // CHECK11: omp.loop.exit:
7136 // CHECK11-NEXT: [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7137 // CHECK11-NEXT: [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
7138 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
7139 // CHECK11-NEXT: [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7140 // CHECK11-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
7141 // CHECK11-NEXT: br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7142 // CHECK11: .omp.final.then:
7143 // CHECK11-NEXT: [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7144 // CHECK11-NEXT: [[SUB15:%.*]] = sub nsw i32 [[TMP36]], 0
7145 // CHECK11-NEXT: [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
7146 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV16]], 1
7147 // CHECK11-NEXT: [[ADD17:%.*]] = add nsw i32 0, [[MUL]]
7148 // CHECK11-NEXT: store i32 [[ADD17]], i32* [[I4]], align 4
7149 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
7150 // CHECK11: .omp.final.done:
7151 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
7152 // CHECK11: omp.precond.end:
7153 // CHECK11-NEXT: ret void
7154 //
7155 //
7156 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..6
7157 // CHECK11-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]] {
7158 // CHECK11-NEXT: entry:
7159 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7160 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7161 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
7162 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
7163 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
7164 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
7165 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
7166 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
7167 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7168 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
7169 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
7170 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
7171 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
7172 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7173 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7174 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7175 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7176 // CHECK11-NEXT: [[I4:%.*]] = alloca i32, align 4
7177 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7178 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7179 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7180 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7181 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
7182 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
7183 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
7184 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
7185 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
7186 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
7187 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
7188 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
7189 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
7190 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7191 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
7192 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
7193 // CHECK11-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
7194 // CHECK11-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
7195 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
7196 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7197 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
7198 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
7199 // CHECK11: omp.precond.then:
7200 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7201 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
7202 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
7203 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7204 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7205 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
7206 // CHECK11-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
7207 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7208 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7209 // CHECK11-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7210 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
7211 // CHECK11-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)
7212 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7213 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
7214 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
7215 // CHECK11-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7216 // CHECK11: cond.true:
7217 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
7218 // CHECK11-NEXT: br label [[COND_END:%.*]]
7219 // CHECK11: cond.false:
7220 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7221 // CHECK11-NEXT: br label [[COND_END]]
7222 // CHECK11: cond.end:
7223 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
7224 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7225 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7226 // CHECK11-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
7227 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7228 // CHECK11: omp.inner.for.cond:
7229 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7230 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7231 // CHECK11-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
7232 // CHECK11-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7233 // CHECK11: omp.inner.for.body:
7234 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7235 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
7236 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7237 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
7238 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
7239 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
7240 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
7241 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7242 // CHECK11: omp.body.continue:
7243 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7244 // CHECK11: omp.inner.for.inc:
7245 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7246 // CHECK11-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
7247 // CHECK11-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
7248 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
7249 // CHECK11: omp.inner.for.end:
7250 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7251 // CHECK11: omp.loop.exit:
7252 // CHECK11-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7253 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
7254 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
7255 // CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7256 // CHECK11-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
7257 // CHECK11-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7258 // CHECK11: .omp.final.then:
7259 // CHECK11-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7260 // CHECK11-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
7261 // CHECK11-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
7262 // CHECK11-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
7263 // CHECK11-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
7264 // CHECK11-NEXT: store i32 [[ADD11]], i32* [[I4]], align 4
7265 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
7266 // CHECK11: .omp.final.done:
7267 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
7268 // CHECK11: omp.precond.end:
7269 // CHECK11-NEXT: ret void
7270 //
7271 //
7272 // CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
7273 // CHECK11-SAME: (i32 [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
7274 // CHECK11-NEXT: entry:
7275 // CHECK11-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
7276 // CHECK11-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
7277 // CHECK11-NEXT: [[M:%.*]] = alloca i32, align 4
7278 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
7279 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
7280 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
7281 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
7282 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
7283 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
7284 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
7285 // CHECK11-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
7286 // CHECK11-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
7287 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 4
7288 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 4
7289 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 4
7290 // CHECK11-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
7291 // CHECK11-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
7292 // CHECK11-NEXT: store i32 10, i32* [[M]], align 4
7293 // CHECK11-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7294 // CHECK11-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
7295 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
7296 // CHECK11-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7297 // CHECK11-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
7298 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
7299 // CHECK11-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
7300 // CHECK11-NEXT: store i8* null, i8** [[TMP4]], align 4
7301 // CHECK11-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7302 // CHECK11-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7303 // CHECK11-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
7304 // CHECK11-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86.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.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
7305 // CHECK11-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
7306 // CHECK11-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
7307 // CHECK11: omp_offload.failed:
7308 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86([10 x i32]* [[A]]) #[[ATTR3]]
7309 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
7310 // CHECK11: omp_offload.cont:
7311 // CHECK11-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
7312 // CHECK11-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
7313 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 4
7314 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
7315 // CHECK11-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
7316 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 4
7317 // CHECK11-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
7318 // CHECK11-NEXT: store i8* null, i8** [[TMP13]], align 4
7319 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
7320 // CHECK11-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
7321 // CHECK11-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
7322 // CHECK11-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.13, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.14, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
7323 // CHECK11-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
7324 // CHECK11-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
7325 // CHECK11: omp_offload.failed5:
7326 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91([10 x i32]* [[A]]) #[[ATTR3]]
7327 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT6]]
7328 // CHECK11: omp_offload.cont6:
7329 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[M]], align 4
7330 // CHECK11-NEXT: store i32 [[TMP18]], i32* [[M_CASTED]], align 4
7331 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[M_CASTED]], align 4
7332 // CHECK11-NEXT: [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
7333 // CHECK11-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
7334 // CHECK11-NEXT: store i32 [[TMP19]], i32* [[TMP21]], align 4
7335 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
7336 // CHECK11-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
7337 // CHECK11-NEXT: store i32 [[TMP19]], i32* [[TMP23]], align 4
7338 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 0
7339 // CHECK11-NEXT: store i8* null, i8** [[TMP24]], align 4
7340 // CHECK11-NEXT: [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
7341 // CHECK11-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
7342 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 4
7343 // CHECK11-NEXT: [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
7344 // CHECK11-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
7345 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 4
7346 // CHECK11-NEXT: [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 1
7347 // CHECK11-NEXT: store i8* null, i8** [[TMP29]], align 4
7348 // CHECK11-NEXT: [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
7349 // CHECK11-NEXT: [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
7350 // CHECK11-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
7351 // CHECK11-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.17, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.18, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
7352 // CHECK11-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
7353 // CHECK11-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
7354 // CHECK11: omp_offload.failed11:
7355 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96(i32 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
7356 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT12]]
7357 // CHECK11: omp_offload.cont12:
7358 // CHECK11-NEXT: ret i32 0
7359 //
7360 //
7361 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86
7362 // CHECK11-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
7363 // CHECK11-NEXT: entry:
7364 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
7365 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
7366 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
7367 // CHECK11-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..8 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
7368 // CHECK11-NEXT: ret void
7369 //
7370 //
7371 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..8
7372 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
7373 // CHECK11-NEXT: entry:
7374 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7375 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7376 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
7377 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7378 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
7379 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7380 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7381 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7382 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7383 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
7384 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7385 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7386 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
7387 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
7388 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7389 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
7390 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7391 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7392 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7393 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
7394 // CHECK11-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)
7395 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7396 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
7397 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7398 // CHECK11: cond.true:
7399 // CHECK11-NEXT: br label [[COND_END:%.*]]
7400 // CHECK11: cond.false:
7401 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7402 // CHECK11-NEXT: br label [[COND_END]]
7403 // CHECK11: cond.end:
7404 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
7405 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7406 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7407 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
7408 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7409 // CHECK11: omp.inner.for.cond:
7410 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7411 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7412 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7413 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7414 // CHECK11: omp.inner.for.body:
7415 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7416 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7417 // CHECK11-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..9 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
7418 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7419 // CHECK11: omp.inner.for.inc:
7420 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7421 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7422 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
7423 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7424 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
7425 // CHECK11: omp.inner.for.end:
7426 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7427 // CHECK11: omp.loop.exit:
7428 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
7429 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7430 // CHECK11-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
7431 // CHECK11-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7432 // CHECK11: .omp.final.then:
7433 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
7434 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
7435 // CHECK11: .omp.final.done:
7436 // CHECK11-NEXT: ret void
7437 //
7438 //
7439 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..9
7440 // CHECK11-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]] {
7441 // CHECK11-NEXT: entry:
7442 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7443 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7444 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
7445 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
7446 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
7447 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7448 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
7449 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7450 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7451 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7452 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7453 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
7454 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7455 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7456 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7457 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7458 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
7459 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
7460 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7461 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
7462 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7463 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7464 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
7465 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
7466 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7467 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7468 // CHECK11-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7469 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
7470 // CHECK11-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)
7471 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7472 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
7473 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7474 // CHECK11: cond.true:
7475 // CHECK11-NEXT: br label [[COND_END:%.*]]
7476 // CHECK11: cond.false:
7477 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7478 // CHECK11-NEXT: br label [[COND_END]]
7479 // CHECK11: cond.end:
7480 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
7481 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7482 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7483 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
7484 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7485 // CHECK11: omp.inner.for.cond:
7486 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7487 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7488 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
7489 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7490 // CHECK11: omp.inner.for.body:
7491 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7492 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
7493 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7494 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4
7495 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
7496 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
7497 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
7498 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7499 // CHECK11: omp.body.continue:
7500 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7501 // CHECK11: omp.inner.for.inc:
7502 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7503 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
7504 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
7505 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
7506 // CHECK11: omp.inner.for.end:
7507 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7508 // CHECK11: omp.loop.exit:
7509 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
7510 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7511 // CHECK11-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
7512 // CHECK11-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7513 // CHECK11: .omp.final.then:
7514 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
7515 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
7516 // CHECK11: .omp.final.done:
7517 // CHECK11-NEXT: ret void
7518 //
7519 //
7520 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91
7521 // CHECK11-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
7522 // CHECK11-NEXT: entry:
7523 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
7524 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
7525 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
7526 // CHECK11-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..11 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
7527 // CHECK11-NEXT: ret void
7528 //
7529 //
7530 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..11
7531 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
7532 // CHECK11-NEXT: entry:
7533 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7534 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7535 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
7536 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7537 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
7538 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7539 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7540 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7541 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7542 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
7543 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7544 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7545 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
7546 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
7547 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7548 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
7549 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7550 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7551 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7552 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
7553 // CHECK11-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)
7554 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7555 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
7556 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7557 // CHECK11: cond.true:
7558 // CHECK11-NEXT: br label [[COND_END:%.*]]
7559 // CHECK11: cond.false:
7560 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7561 // CHECK11-NEXT: br label [[COND_END]]
7562 // CHECK11: cond.end:
7563 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
7564 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7565 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7566 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
7567 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7568 // CHECK11: omp.inner.for.cond:
7569 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7570 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7571 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7572 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7573 // CHECK11: omp.inner.for.body:
7574 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7575 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7576 // CHECK11-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..12 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
7577 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7578 // CHECK11: omp.inner.for.inc:
7579 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7580 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7581 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
7582 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7583 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
7584 // CHECK11: omp.inner.for.end:
7585 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7586 // CHECK11: omp.loop.exit:
7587 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
7588 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7589 // CHECK11-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
7590 // CHECK11-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7591 // CHECK11: .omp.final.then:
7592 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
7593 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
7594 // CHECK11: .omp.final.done:
7595 // CHECK11-NEXT: ret void
7596 //
7597 //
7598 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..12
7599 // CHECK11-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]] {
7600 // CHECK11-NEXT: entry:
7601 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7602 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7603 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
7604 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
7605 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
7606 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7607 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
7608 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7609 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7610 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7611 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7612 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
7613 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7614 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7615 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7616 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7617 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
7618 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
7619 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7620 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
7621 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7622 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7623 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
7624 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
7625 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7626 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7627 // CHECK11-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7628 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
7629 // CHECK11-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)
7630 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7631 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
7632 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7633 // CHECK11: cond.true:
7634 // CHECK11-NEXT: br label [[COND_END:%.*]]
7635 // CHECK11: cond.false:
7636 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7637 // CHECK11-NEXT: br label [[COND_END]]
7638 // CHECK11: cond.end:
7639 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
7640 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7641 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7642 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
7643 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7644 // CHECK11: omp.inner.for.cond:
7645 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7646 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7647 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
7648 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7649 // CHECK11: omp.inner.for.body:
7650 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7651 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
7652 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7653 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4
7654 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
7655 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
7656 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
7657 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7658 // CHECK11: omp.body.continue:
7659 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7660 // CHECK11: omp.inner.for.inc:
7661 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7662 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
7663 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
7664 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
7665 // CHECK11: omp.inner.for.end:
7666 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7667 // CHECK11: omp.loop.exit:
7668 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
7669 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7670 // CHECK11-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
7671 // CHECK11-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7672 // CHECK11: .omp.final.then:
7673 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
7674 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
7675 // CHECK11: .omp.final.done:
7676 // CHECK11-NEXT: ret void
7677 //
7678 //
7679 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96
7680 // CHECK11-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
7681 // CHECK11-NEXT: entry:
7682 // CHECK11-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
7683 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
7684 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
7685 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
7686 // CHECK11-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
7687 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
7688 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
7689 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
7690 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
7691 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7692 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
7693 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
7694 // CHECK11-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..15 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
7695 // CHECK11-NEXT: ret void
7696 //
7697 //
7698 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..15
7699 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
7700 // CHECK11-NEXT: entry:
7701 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7702 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7703 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
7704 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
7705 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7706 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
7707 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7708 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7709 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7710 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7711 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
7712 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
7713 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7714 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7715 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
7716 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
7717 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
7718 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7719 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
7720 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7721 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7722 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
7723 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7724 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
7725 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP1]])
7726 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7727 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
7728 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7729 // CHECK11: cond.true:
7730 // CHECK11-NEXT: br label [[COND_END:%.*]]
7731 // CHECK11: cond.false:
7732 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7733 // CHECK11-NEXT: br label [[COND_END]]
7734 // CHECK11: cond.end:
7735 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7736 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7737 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7738 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
7739 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7740 // CHECK11: omp.inner.for.cond:
7741 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7742 // CHECK11-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP7]], 10
7743 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7744 // CHECK11: omp.inner.for.body:
7745 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7746 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7747 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
7748 // CHECK11-NEXT: store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
7749 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
7750 // CHECK11-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..16 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
7751 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7752 // CHECK11: omp.inner.for.inc:
7753 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7754 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7755 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
7756 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7757 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7758 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7759 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
7760 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4
7761 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7762 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7763 // CHECK11-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
7764 // CHECK11-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4
7765 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7766 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
7767 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
7768 // CHECK11: cond.true5:
7769 // CHECK11-NEXT: br label [[COND_END7:%.*]]
7770 // CHECK11: cond.false6:
7771 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7772 // CHECK11-NEXT: br label [[COND_END7]]
7773 // CHECK11: cond.end7:
7774 // CHECK11-NEXT: [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
7775 // CHECK11-NEXT: store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4
7776 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7777 // CHECK11-NEXT: store i32 [[TMP20]], i32* [[DOTOMP_IV]], align 4
7778 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
7779 // CHECK11: omp.inner.for.end:
7780 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7781 // CHECK11: omp.loop.exit:
7782 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
7783 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7784 // CHECK11-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
7785 // CHECK11-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7786 // CHECK11: .omp.final.then:
7787 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
7788 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
7789 // CHECK11: .omp.final.done:
7790 // CHECK11-NEXT: ret void
7791 //
7792 //
7793 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..16
7794 // CHECK11-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]] {
7795 // CHECK11-NEXT: entry:
7796 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
7797 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
7798 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
7799 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
7800 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
7801 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
7802 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
7803 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
7804 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
7805 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
7806 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7807 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7808 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
7809 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
7810 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
7811 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7812 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7813 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
7814 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
7815 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
7816 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
7817 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
7818 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
7819 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
7820 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
7821 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
7822 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7823 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7824 // CHECK11-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
7825 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
7826 // CHECK11-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)
7827 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7828 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
7829 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7830 // CHECK11: cond.true:
7831 // CHECK11-NEXT: br label [[COND_END:%.*]]
7832 // CHECK11: cond.false:
7833 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7834 // CHECK11-NEXT: br label [[COND_END]]
7835 // CHECK11: cond.end:
7836 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
7837 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7838 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7839 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
7840 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
7841 // CHECK11: omp.inner.for.cond:
7842 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7843 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7844 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
7845 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7846 // CHECK11: omp.inner.for.body:
7847 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7848 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
7849 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7850 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4
7851 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
7852 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
7853 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
7854 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
7855 // CHECK11: omp.body.continue:
7856 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
7857 // CHECK11: omp.inner.for.inc:
7858 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7859 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
7860 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
7861 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
7862 // CHECK11: omp.inner.for.end:
7863 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
7864 // CHECK11: omp.loop.exit:
7865 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
7866 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7867 // CHECK11-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
7868 // CHECK11-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7869 // CHECK11: .omp.final.then:
7870 // CHECK11-NEXT: store i32 10, i32* [[I]], align 4
7871 // CHECK11-NEXT: br label [[DOTOMP_FINAL_DONE]]
7872 // CHECK11: .omp.final.done:
7873 // CHECK11-NEXT: ret void
7874 //
7875 //
7876 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
7877 // CHECK11-SAME: () #[[ATTR5:[0-9]+]] {
7878 // CHECK11-NEXT: entry:
7879 // CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
7880 // CHECK11-NEXT: ret void
7881 //
7882 //
7883 // CHECK12-LABEL: define {{[^@]+}}@main
7884 // CHECK12-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
7885 // CHECK12-NEXT: entry:
7886 // CHECK12-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
7887 // CHECK12-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
7888 // CHECK12-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
7889 // CHECK12-NEXT: [[N:%.*]] = alloca i32, align 4
7890 // CHECK12-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
7891 // CHECK12-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
7892 // CHECK12-NEXT: [[M:%.*]] = alloca i32, align 4
7893 // CHECK12-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
7894 // CHECK12-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
7895 // CHECK12-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
7896 // CHECK12-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
7897 // CHECK12-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
7898 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
7899 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
7900 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
7901 // CHECK12-NEXT: [[N_CASTED3:%.*]] = alloca i32, align 4
7902 // CHECK12-NEXT: [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
7903 // CHECK12-NEXT: [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
7904 // CHECK12-NEXT: [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
7905 // CHECK12-NEXT: [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
7906 // CHECK12-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
7907 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
7908 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
7909 // CHECK12-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
7910 // CHECK12-NEXT: [[N_CASTED17:%.*]] = alloca i32, align 4
7911 // CHECK12-NEXT: [[DOTOFFLOAD_BASEPTRS18:%.*]] = alloca [4 x i8*], align 4
7912 // CHECK12-NEXT: [[DOTOFFLOAD_PTRS19:%.*]] = alloca [4 x i8*], align 4
7913 // CHECK12-NEXT: [[DOTOFFLOAD_MAPPERS20:%.*]] = alloca [4 x i8*], align 4
7914 // CHECK12-NEXT: [[DOTOFFLOAD_SIZES21:%.*]] = alloca [4 x i64], align 4
7915 // CHECK12-NEXT: [[_TMP22:%.*]] = alloca i32, align 4
7916 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_23:%.*]] = alloca i32, align 4
7917 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
7918 // CHECK12-NEXT: store i32 0, i32* [[RETVAL]], align 4
7919 // CHECK12-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
7920 // CHECK12-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
7921 // CHECK12-NEXT: store i32 100, i32* [[N]], align 4
7922 // CHECK12-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
7923 // CHECK12-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
7924 // CHECK12-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
7925 // CHECK12-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
7926 // CHECK12-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
7927 // CHECK12-NEXT: store i32 10, i32* [[M]], align 4
7928 // CHECK12-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
7929 // CHECK12-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
7930 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
7931 // CHECK12-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
7932 // CHECK12-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
7933 // CHECK12-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7934 // CHECK12-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
7935 // CHECK12-NEXT: store i32 [[TMP3]], i32* [[TMP7]], align 4
7936 // CHECK12-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7937 // CHECK12-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
7938 // CHECK12-NEXT: store i32 [[TMP3]], i32* [[TMP9]], align 4
7939 // CHECK12-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
7940 // CHECK12-NEXT: store i64 4, i64* [[TMP10]], align 4
7941 // CHECK12-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
7942 // CHECK12-NEXT: store i8* null, i8** [[TMP11]], align 4
7943 // CHECK12-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
7944 // CHECK12-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
7945 // CHECK12-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
7946 // CHECK12-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
7947 // CHECK12-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
7948 // CHECK12-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
7949 // CHECK12-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
7950 // CHECK12-NEXT: store i64 4, i64* [[TMP16]], align 4
7951 // CHECK12-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
7952 // CHECK12-NEXT: store i8* null, i8** [[TMP17]], align 4
7953 // CHECK12-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
7954 // CHECK12-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
7955 // CHECK12-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 4
7956 // CHECK12-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
7957 // CHECK12-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
7958 // CHECK12-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 4
7959 // CHECK12-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
7960 // CHECK12-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 4
7961 // CHECK12-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
7962 // CHECK12-NEXT: store i8* null, i8** [[TMP23]], align 4
7963 // CHECK12-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7964 // CHECK12-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7965 // CHECK12-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
7966 // CHECK12-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
7967 // CHECK12-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
7968 // CHECK12-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
7969 // CHECK12-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
7970 // CHECK12-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
7971 // CHECK12-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
7972 // CHECK12-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
7973 // CHECK12-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
7974 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
7975 // CHECK12-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
7976 // CHECK12-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 [[TMP30]])
7977 // CHECK12-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108.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)
7978 // CHECK12-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
7979 // CHECK12-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
7980 // CHECK12: omp_offload.failed:
7981 // CHECK12-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
7982 // CHECK12-NEXT: br label [[OMP_OFFLOAD_CONT]]
7983 // CHECK12: omp_offload.cont:
7984 // CHECK12-NEXT: [[TMP33:%.*]] = load i32, i32* [[N]], align 4
7985 // CHECK12-NEXT: store i32 [[TMP33]], i32* [[N_CASTED3]], align 4
7986 // CHECK12-NEXT: [[TMP34:%.*]] = load i32, i32* [[N_CASTED3]], align 4
7987 // CHECK12-NEXT: [[TMP35:%.*]] = mul nuw i32 [[TMP0]], 4
7988 // CHECK12-NEXT: [[TMP36:%.*]] = sext i32 [[TMP35]] to i64
7989 // CHECK12-NEXT: [[TMP37:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
7990 // CHECK12-NEXT: [[TMP38:%.*]] = bitcast i8** [[TMP37]] to i32*
7991 // CHECK12-NEXT: store i32 [[TMP34]], i32* [[TMP38]], align 4
7992 // CHECK12-NEXT: [[TMP39:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
7993 // CHECK12-NEXT: [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
7994 // CHECK12-NEXT: store i32 [[TMP34]], i32* [[TMP40]], align 4
7995 // CHECK12-NEXT: [[TMP41:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
7996 // CHECK12-NEXT: store i64 4, i64* [[TMP41]], align 4
7997 // CHECK12-NEXT: [[TMP42:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
7998 // CHECK12-NEXT: store i8* null, i8** [[TMP42]], align 4
7999 // CHECK12-NEXT: [[TMP43:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
8000 // CHECK12-NEXT: [[TMP44:%.*]] = bitcast i8** [[TMP43]] to i32*
8001 // CHECK12-NEXT: store i32 [[TMP0]], i32* [[TMP44]], align 4
8002 // CHECK12-NEXT: [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
8003 // CHECK12-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i32*
8004 // CHECK12-NEXT: store i32 [[TMP0]], i32* [[TMP46]], align 4
8005 // CHECK12-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 1
8006 // CHECK12-NEXT: store i64 4, i64* [[TMP47]], align 4
8007 // CHECK12-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
8008 // CHECK12-NEXT: store i8* null, i8** [[TMP48]], align 4
8009 // CHECK12-NEXT: [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
8010 // CHECK12-NEXT: [[TMP50:%.*]] = bitcast i8** [[TMP49]] to i32**
8011 // CHECK12-NEXT: store i32* [[VLA]], i32** [[TMP50]], align 4
8012 // CHECK12-NEXT: [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
8013 // CHECK12-NEXT: [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32**
8014 // CHECK12-NEXT: store i32* [[VLA]], i32** [[TMP52]], align 4
8015 // CHECK12-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
8016 // CHECK12-NEXT: store i64 [[TMP36]], i64* [[TMP53]], align 4
8017 // CHECK12-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
8018 // CHECK12-NEXT: store i8* null, i8** [[TMP54]], align 4
8019 // CHECK12-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
8020 // CHECK12-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
8021 // CHECK12-NEXT: [[TMP57:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
8022 // CHECK12-NEXT: [[TMP58:%.*]] = load i32, i32* [[N]], align 4
8023 // CHECK12-NEXT: store i32 [[TMP58]], i32* [[DOTCAPTURE_EXPR_9]], align 4
8024 // CHECK12-NEXT: [[TMP59:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
8025 // CHECK12-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP59]], 0
8026 // CHECK12-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
8027 // CHECK12-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
8028 // CHECK12-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
8029 // CHECK12-NEXT: [[TMP60:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
8030 // CHECK12-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP60]], 1
8031 // CHECK12-NEXT: [[TMP61:%.*]] = zext i32 [[ADD14]] to i64
8032 // CHECK12-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP61]])
8033 // CHECK12-NEXT: [[TMP62:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113.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)
8034 // CHECK12-NEXT: [[TMP63:%.*]] = icmp ne i32 [[TMP62]], 0
8035 // CHECK12-NEXT: br i1 [[TMP63]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
8036 // CHECK12: omp_offload.failed15:
8037 // CHECK12-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113(i32 [[TMP34]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
8038 // CHECK12-NEXT: br label [[OMP_OFFLOAD_CONT16]]
8039 // CHECK12: omp_offload.cont16:
8040 // CHECK12-NEXT: [[TMP64:%.*]] = load i32, i32* [[M]], align 4
8041 // CHECK12-NEXT: store i32 [[TMP64]], i32* [[M_CASTED]], align 4
8042 // CHECK12-NEXT: [[TMP65:%.*]] = load i32, i32* [[M_CASTED]], align 4
8043 // CHECK12-NEXT: [[TMP66:%.*]] = load i32, i32* [[N]], align 4
8044 // CHECK12-NEXT: store i32 [[TMP66]], i32* [[N_CASTED17]], align 4
8045 // CHECK12-NEXT: [[TMP67:%.*]] = load i32, i32* [[N_CASTED17]], align 4
8046 // CHECK12-NEXT: [[TMP68:%.*]] = mul nuw i32 [[TMP0]], 4
8047 // CHECK12-NEXT: [[TMP69:%.*]] = sext i32 [[TMP68]] to i64
8048 // CHECK12-NEXT: [[TMP70:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
8049 // CHECK12-NEXT: [[TMP71:%.*]] = bitcast i8** [[TMP70]] to i32*
8050 // CHECK12-NEXT: store i32 [[TMP65]], i32* [[TMP71]], align 4
8051 // CHECK12-NEXT: [[TMP72:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
8052 // CHECK12-NEXT: [[TMP73:%.*]] = bitcast i8** [[TMP72]] to i32*
8053 // CHECK12-NEXT: store i32 [[TMP65]], i32* [[TMP73]], align 4
8054 // CHECK12-NEXT: [[TMP74:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
8055 // CHECK12-NEXT: store i64 4, i64* [[TMP74]], align 4
8056 // CHECK12-NEXT: [[TMP75:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 0
8057 // CHECK12-NEXT: store i8* null, i8** [[TMP75]], align 4
8058 // CHECK12-NEXT: [[TMP76:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 1
8059 // CHECK12-NEXT: [[TMP77:%.*]] = bitcast i8** [[TMP76]] to i32*
8060 // CHECK12-NEXT: store i32 [[TMP67]], i32* [[TMP77]], align 4
8061 // CHECK12-NEXT: [[TMP78:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 1
8062 // CHECK12-NEXT: [[TMP79:%.*]] = bitcast i8** [[TMP78]] to i32*
8063 // CHECK12-NEXT: store i32 [[TMP67]], i32* [[TMP79]], align 4
8064 // CHECK12-NEXT: [[TMP80:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 1
8065 // CHECK12-NEXT: store i64 4, i64* [[TMP80]], align 4
8066 // CHECK12-NEXT: [[TMP81:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 1
8067 // CHECK12-NEXT: store i8* null, i8** [[TMP81]], align 4
8068 // CHECK12-NEXT: [[TMP82:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 2
8069 // CHECK12-NEXT: [[TMP83:%.*]] = bitcast i8** [[TMP82]] to i32*
8070 // CHECK12-NEXT: store i32 [[TMP0]], i32* [[TMP83]], align 4
8071 // CHECK12-NEXT: [[TMP84:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 2
8072 // CHECK12-NEXT: [[TMP85:%.*]] = bitcast i8** [[TMP84]] to i32*
8073 // CHECK12-NEXT: store i32 [[TMP0]], i32* [[TMP85]], align 4
8074 // CHECK12-NEXT: [[TMP86:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 2
8075 // CHECK12-NEXT: store i64 4, i64* [[TMP86]], align 4
8076 // CHECK12-NEXT: [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 2
8077 // CHECK12-NEXT: store i8* null, i8** [[TMP87]], align 4
8078 // CHECK12-NEXT: [[TMP88:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 3
8079 // CHECK12-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i32**
8080 // CHECK12-NEXT: store i32* [[VLA]], i32** [[TMP89]], align 4
8081 // CHECK12-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 3
8082 // CHECK12-NEXT: [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i32**
8083 // CHECK12-NEXT: store i32* [[VLA]], i32** [[TMP91]], align 4
8084 // CHECK12-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 3
8085 // CHECK12-NEXT: store i64 [[TMP69]], i64* [[TMP92]], align 4
8086 // CHECK12-NEXT: [[TMP93:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS20]], i32 0, i32 3
8087 // CHECK12-NEXT: store i8* null, i8** [[TMP93]], align 4
8088 // CHECK12-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS18]], i32 0, i32 0
8089 // CHECK12-NEXT: [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS19]], i32 0, i32 0
8090 // CHECK12-NEXT: [[TMP96:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES21]], i32 0, i32 0
8091 // CHECK12-NEXT: [[TMP97:%.*]] = load i32, i32* [[N]], align 4
8092 // CHECK12-NEXT: store i32 [[TMP97]], i32* [[DOTCAPTURE_EXPR_23]], align 4
8093 // CHECK12-NEXT: [[TMP98:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_23]], align 4
8094 // CHECK12-NEXT: [[SUB25:%.*]] = sub nsw i32 [[TMP98]], 0
8095 // CHECK12-NEXT: [[DIV26:%.*]] = sdiv i32 [[SUB25]], 1
8096 // CHECK12-NEXT: [[SUB27:%.*]] = sub nsw i32 [[DIV26]], 1
8097 // CHECK12-NEXT: store i32 [[SUB27]], i32* [[DOTCAPTURE_EXPR_24]], align 4
8098 // CHECK12-NEXT: [[TMP99:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
8099 // CHECK12-NEXT: [[ADD28:%.*]] = add nsw i32 [[TMP99]], 1
8100 // CHECK12-NEXT: [[TMP100:%.*]] = zext i32 [[ADD28]] to i64
8101 // CHECK12-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 [[TMP100]])
8102 // CHECK12-NEXT: [[TMP101:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118.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)
8103 // CHECK12-NEXT: [[TMP102:%.*]] = icmp ne i32 [[TMP101]], 0
8104 // CHECK12-NEXT: br i1 [[TMP102]], label [[OMP_OFFLOAD_FAILED29:%.*]], label [[OMP_OFFLOAD_CONT30:%.*]]
8105 // CHECK12: omp_offload.failed29:
8106 // CHECK12-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118(i32 [[TMP65]], i32 [[TMP67]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
8107 // CHECK12-NEXT: br label [[OMP_OFFLOAD_CONT30]]
8108 // CHECK12: omp_offload.cont30:
8109 // CHECK12-NEXT: [[TMP103:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
8110 // CHECK12-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP103]])
8111 // CHECK12-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
8112 // CHECK12-NEXT: [[TMP104:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
8113 // CHECK12-NEXT: call void @llvm.stackrestore(i8* [[TMP104]])
8114 // CHECK12-NEXT: [[TMP105:%.*]] = load i32, i32* [[RETVAL]], align 4
8115 // CHECK12-NEXT: ret i32 [[TMP105]]
8116 //
8117 //
8118 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l108
8119 // CHECK12-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
8120 // CHECK12-NEXT: entry:
8121 // CHECK12-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
8122 // CHECK12-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
8123 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
8124 // CHECK12-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
8125 // CHECK12-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
8126 // CHECK12-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
8127 // CHECK12-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
8128 // CHECK12-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
8129 // CHECK12-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]])
8130 // CHECK12-NEXT: ret void
8131 //
8132 //
8133 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined.
8134 // CHECK12-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]] {
8135 // CHECK12-NEXT: entry:
8136 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8137 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8138 // CHECK12-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
8139 // CHECK12-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
8140 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
8141 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8142 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
8143 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8144 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8145 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
8146 // CHECK12-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8147 // CHECK12-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8148 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8149 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8150 // CHECK12-NEXT: [[I3:%.*]] = alloca i32, align 4
8151 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8152 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8153 // CHECK12-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
8154 // CHECK12-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
8155 // CHECK12-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
8156 // CHECK12-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
8157 // CHECK12-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
8158 // CHECK12-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
8159 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8160 // CHECK12-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
8161 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8162 // CHECK12-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8163 // CHECK12-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8164 // CHECK12-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
8165 // CHECK12-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8166 // CHECK12-NEXT: store i32 0, i32* [[I]], align 4
8167 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8168 // CHECK12-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8169 // CHECK12-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8170 // CHECK12: omp.precond.then:
8171 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8172 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8173 // CHECK12-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
8174 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8175 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8176 // CHECK12-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8177 // CHECK12-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
8178 // CHECK12-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)
8179 // CHECK12-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8180 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8181 // CHECK12-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
8182 // CHECK12-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8183 // CHECK12: cond.true:
8184 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8185 // CHECK12-NEXT: br label [[COND_END:%.*]]
8186 // CHECK12: cond.false:
8187 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8188 // CHECK12-NEXT: br label [[COND_END]]
8189 // CHECK12: cond.end:
8190 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
8191 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8192 // CHECK12-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8193 // CHECK12-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
8194 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8195 // CHECK12: omp.inner.for.cond:
8196 // CHECK12-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8197 // CHECK12-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8198 // CHECK12-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
8199 // CHECK12-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8200 // CHECK12: omp.inner.for.body:
8201 // CHECK12-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8202 // CHECK12-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8203 // CHECK12-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]])
8204 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8205 // CHECK12: omp.inner.for.inc:
8206 // CHECK12-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8207 // CHECK12-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8208 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
8209 // CHECK12-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
8210 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
8211 // CHECK12: omp.inner.for.end:
8212 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8213 // CHECK12: omp.loop.exit:
8214 // CHECK12-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8215 // CHECK12-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
8216 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
8217 // CHECK12-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8218 // CHECK12-NEXT: [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
8219 // CHECK12-NEXT: br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8220 // CHECK12: .omp.final.then:
8221 // CHECK12-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8222 // CHECK12-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
8223 // CHECK12-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
8224 // CHECK12-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
8225 // CHECK12-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
8226 // CHECK12-NEXT: store i32 [[ADD8]], i32* [[I3]], align 4
8227 // CHECK12-NEXT: br label [[DOTOMP_FINAL_DONE]]
8228 // CHECK12: .omp.final.done:
8229 // CHECK12-NEXT: br label [[OMP_PRECOND_END]]
8230 // CHECK12: omp.precond.end:
8231 // CHECK12-NEXT: ret void
8232 //
8233 //
8234 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..1
8235 // CHECK12-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]] {
8236 // CHECK12-NEXT: entry:
8237 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8238 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8239 // CHECK12-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
8240 // CHECK12-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
8241 // CHECK12-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
8242 // CHECK12-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
8243 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
8244 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8245 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
8246 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8247 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8248 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
8249 // CHECK12-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8250 // CHECK12-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8251 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8252 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8253 // CHECK12-NEXT: [[I3:%.*]] = alloca i32, align 4
8254 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8255 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8256 // CHECK12-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
8257 // CHECK12-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
8258 // CHECK12-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
8259 // CHECK12-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
8260 // CHECK12-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
8261 // CHECK12-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
8262 // CHECK12-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
8263 // CHECK12-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
8264 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8265 // CHECK12-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
8266 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8267 // CHECK12-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8268 // CHECK12-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8269 // CHECK12-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
8270 // CHECK12-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8271 // CHECK12-NEXT: store i32 0, i32* [[I]], align 4
8272 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8273 // CHECK12-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8274 // CHECK12-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8275 // CHECK12: omp.precond.then:
8276 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8277 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8278 // CHECK12-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
8279 // CHECK12-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
8280 // CHECK12-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
8281 // CHECK12-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
8282 // CHECK12-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
8283 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8284 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8285 // CHECK12-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8286 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
8287 // CHECK12-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)
8288 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8289 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8290 // CHECK12-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
8291 // CHECK12-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8292 // CHECK12: cond.true:
8293 // CHECK12-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8294 // CHECK12-NEXT: br label [[COND_END:%.*]]
8295 // CHECK12: cond.false:
8296 // CHECK12-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8297 // CHECK12-NEXT: br label [[COND_END]]
8298 // CHECK12: cond.end:
8299 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
8300 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8301 // CHECK12-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8302 // CHECK12-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
8303 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8304 // CHECK12: omp.inner.for.cond:
8305 // CHECK12-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8306 // CHECK12-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8307 // CHECK12-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
8308 // CHECK12-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8309 // CHECK12: omp.inner.for.body:
8310 // CHECK12-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8311 // CHECK12-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
8312 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8313 // CHECK12-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
8314 // CHECK12-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
8315 // CHECK12-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
8316 // CHECK12-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
8317 // CHECK12-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8318 // CHECK12: omp.body.continue:
8319 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8320 // CHECK12: omp.inner.for.inc:
8321 // CHECK12-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8322 // CHECK12-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
8323 // CHECK12-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
8324 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP11:![0-9]+]]
8325 // CHECK12: omp.inner.for.end:
8326 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8327 // CHECK12: omp.loop.exit:
8328 // CHECK12-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8329 // CHECK12-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
8330 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
8331 // CHECK12-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8332 // CHECK12-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
8333 // CHECK12-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8334 // CHECK12: .omp.final.then:
8335 // CHECK12-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8336 // CHECK12-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
8337 // CHECK12-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
8338 // CHECK12-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
8339 // CHECK12-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
8340 // CHECK12-NEXT: store i32 [[ADD10]], i32* [[I3]], align 4
8341 // CHECK12-NEXT: br label [[DOTOMP_FINAL_DONE]]
8342 // CHECK12: .omp.final.done:
8343 // CHECK12-NEXT: br label [[OMP_PRECOND_END]]
8344 // CHECK12: omp.precond.end:
8345 // CHECK12-NEXT: ret void
8346 //
8347 //
8348 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l113
8349 // CHECK12-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
8350 // CHECK12-NEXT: entry:
8351 // CHECK12-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
8352 // CHECK12-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
8353 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
8354 // CHECK12-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
8355 // CHECK12-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
8356 // CHECK12-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
8357 // CHECK12-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
8358 // CHECK12-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
8359 // CHECK12-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]])
8360 // CHECK12-NEXT: ret void
8361 //
8362 //
8363 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..2
8364 // CHECK12-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]] {
8365 // CHECK12-NEXT: entry:
8366 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8367 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8368 // CHECK12-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
8369 // CHECK12-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
8370 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
8371 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8372 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
8373 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8374 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8375 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
8376 // CHECK12-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8377 // CHECK12-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8378 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8379 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8380 // CHECK12-NEXT: [[I3:%.*]] = alloca i32, align 4
8381 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8382 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8383 // CHECK12-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
8384 // CHECK12-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
8385 // CHECK12-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
8386 // CHECK12-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
8387 // CHECK12-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
8388 // CHECK12-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
8389 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8390 // CHECK12-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
8391 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8392 // CHECK12-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8393 // CHECK12-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8394 // CHECK12-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
8395 // CHECK12-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8396 // CHECK12-NEXT: store i32 0, i32* [[I]], align 4
8397 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8398 // CHECK12-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8399 // CHECK12-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8400 // CHECK12: omp.precond.then:
8401 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8402 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8403 // CHECK12-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
8404 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8405 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8406 // CHECK12-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8407 // CHECK12-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
8408 // CHECK12-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)
8409 // CHECK12-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8410 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8411 // CHECK12-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
8412 // CHECK12-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8413 // CHECK12: cond.true:
8414 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8415 // CHECK12-NEXT: br label [[COND_END:%.*]]
8416 // CHECK12: cond.false:
8417 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8418 // CHECK12-NEXT: br label [[COND_END]]
8419 // CHECK12: cond.end:
8420 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
8421 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8422 // CHECK12-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8423 // CHECK12-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
8424 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8425 // CHECK12: omp.inner.for.cond:
8426 // CHECK12-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8427 // CHECK12-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8428 // CHECK12-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
8429 // CHECK12-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8430 // CHECK12: omp.inner.for.body:
8431 // CHECK12-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8432 // CHECK12-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8433 // CHECK12-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]])
8434 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8435 // CHECK12: omp.inner.for.inc:
8436 // CHECK12-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8437 // CHECK12-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8438 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
8439 // CHECK12-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
8440 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
8441 // CHECK12: omp.inner.for.end:
8442 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8443 // CHECK12: omp.loop.exit:
8444 // CHECK12-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8445 // CHECK12-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
8446 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
8447 // CHECK12-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8448 // CHECK12-NEXT: [[TMP23:%.*]] = icmp ne i32 [[TMP22]], 0
8449 // CHECK12-NEXT: br i1 [[TMP23]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8450 // CHECK12: .omp.final.then:
8451 // CHECK12-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8452 // CHECK12-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP24]], 0
8453 // CHECK12-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
8454 // CHECK12-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV7]], 1
8455 // CHECK12-NEXT: [[ADD8:%.*]] = add nsw i32 0, [[MUL]]
8456 // CHECK12-NEXT: store i32 [[ADD8]], i32* [[I3]], align 4
8457 // CHECK12-NEXT: br label [[DOTOMP_FINAL_DONE]]
8458 // CHECK12: .omp.final.done:
8459 // CHECK12-NEXT: br label [[OMP_PRECOND_END]]
8460 // CHECK12: omp.precond.end:
8461 // CHECK12-NEXT: ret void
8462 //
8463 //
8464 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..3
8465 // CHECK12-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]] {
8466 // CHECK12-NEXT: entry:
8467 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8468 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8469 // CHECK12-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
8470 // CHECK12-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
8471 // CHECK12-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
8472 // CHECK12-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
8473 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
8474 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8475 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
8476 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8477 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8478 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
8479 // CHECK12-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8480 // CHECK12-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8481 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8482 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8483 // CHECK12-NEXT: [[I3:%.*]] = alloca i32, align 4
8484 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8485 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8486 // CHECK12-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
8487 // CHECK12-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
8488 // CHECK12-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
8489 // CHECK12-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
8490 // CHECK12-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
8491 // CHECK12-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
8492 // CHECK12-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
8493 // CHECK12-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
8494 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8495 // CHECK12-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
8496 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8497 // CHECK12-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8498 // CHECK12-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8499 // CHECK12-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
8500 // CHECK12-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8501 // CHECK12-NEXT: store i32 0, i32* [[I]], align 4
8502 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8503 // CHECK12-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8504 // CHECK12-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8505 // CHECK12: omp.precond.then:
8506 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8507 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8508 // CHECK12-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
8509 // CHECK12-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
8510 // CHECK12-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
8511 // CHECK12-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
8512 // CHECK12-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
8513 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8514 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8515 // CHECK12-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8516 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
8517 // CHECK12-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)
8518 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8519 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8520 // CHECK12-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
8521 // CHECK12-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8522 // CHECK12: cond.true:
8523 // CHECK12-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8524 // CHECK12-NEXT: br label [[COND_END:%.*]]
8525 // CHECK12: cond.false:
8526 // CHECK12-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8527 // CHECK12-NEXT: br label [[COND_END]]
8528 // CHECK12: cond.end:
8529 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
8530 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8531 // CHECK12-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8532 // CHECK12-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
8533 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8534 // CHECK12: omp.inner.for.cond:
8535 // CHECK12-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8536 // CHECK12-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8537 // CHECK12-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
8538 // CHECK12-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8539 // CHECK12: omp.inner.for.body:
8540 // CHECK12-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8541 // CHECK12-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
8542 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8543 // CHECK12-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
8544 // CHECK12-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
8545 // CHECK12-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
8546 // CHECK12-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
8547 // CHECK12-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8548 // CHECK12: omp.body.continue:
8549 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8550 // CHECK12: omp.inner.for.inc:
8551 // CHECK12-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8552 // CHECK12-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
8553 // CHECK12-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
8554 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
8555 // CHECK12: omp.inner.for.end:
8556 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8557 // CHECK12: omp.loop.exit:
8558 // CHECK12-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8559 // CHECK12-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
8560 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
8561 // CHECK12-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8562 // CHECK12-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
8563 // CHECK12-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8564 // CHECK12: .omp.final.then:
8565 // CHECK12-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8566 // CHECK12-NEXT: [[SUB7:%.*]] = sub nsw i32 [[TMP25]], 0
8567 // CHECK12-NEXT: [[DIV8:%.*]] = sdiv i32 [[SUB7]], 1
8568 // CHECK12-NEXT: [[MUL9:%.*]] = mul nsw i32 [[DIV8]], 1
8569 // CHECK12-NEXT: [[ADD10:%.*]] = add nsw i32 0, [[MUL9]]
8570 // CHECK12-NEXT: store i32 [[ADD10]], i32* [[I3]], align 4
8571 // CHECK12-NEXT: br label [[DOTOMP_FINAL_DONE]]
8572 // CHECK12: .omp.final.done:
8573 // CHECK12-NEXT: br label [[OMP_PRECOND_END]]
8574 // CHECK12: omp.precond.end:
8575 // CHECK12-NEXT: ret void
8576 //
8577 //
8578 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l118
8579 // CHECK12-SAME: (i32 [[M:%.*]], i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
8580 // CHECK12-NEXT: entry:
8581 // CHECK12-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
8582 // CHECK12-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
8583 // CHECK12-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
8584 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
8585 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
8586 // CHECK12-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
8587 // CHECK12-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
8588 // CHECK12-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
8589 // CHECK12-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
8590 // CHECK12-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
8591 // CHECK12-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
8592 // CHECK12-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
8593 // CHECK12-NEXT: [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
8594 // CHECK12-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
8595 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
8596 // CHECK12-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
8597 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
8598 // CHECK12-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]])
8599 // CHECK12-NEXT: ret void
8600 //
8601 //
8602 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..5
8603 // CHECK12-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]] {
8604 // CHECK12-NEXT: entry:
8605 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8606 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8607 // CHECK12-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
8608 // CHECK12-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
8609 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
8610 // CHECK12-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
8611 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8612 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
8613 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8614 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
8615 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
8616 // CHECK12-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8617 // CHECK12-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8618 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8619 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8620 // CHECK12-NEXT: [[I4:%.*]] = alloca i32, align 4
8621 // CHECK12-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
8622 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8623 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8624 // CHECK12-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
8625 // CHECK12-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
8626 // CHECK12-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
8627 // CHECK12-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
8628 // CHECK12-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
8629 // CHECK12-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
8630 // CHECK12-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
8631 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8632 // CHECK12-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8633 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8634 // CHECK12-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8635 // CHECK12-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8636 // CHECK12-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
8637 // CHECK12-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
8638 // CHECK12-NEXT: store i32 0, i32* [[I]], align 4
8639 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8640 // CHECK12-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8641 // CHECK12-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8642 // CHECK12: omp.precond.then:
8643 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8644 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8645 // CHECK12-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
8646 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8647 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8648 // CHECK12-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
8649 // CHECK12-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8650 // CHECK12-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
8651 // CHECK12-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]])
8652 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8653 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8654 // CHECK12-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
8655 // CHECK12-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8656 // CHECK12: cond.true:
8657 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8658 // CHECK12-NEXT: br label [[COND_END:%.*]]
8659 // CHECK12: cond.false:
8660 // CHECK12-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8661 // CHECK12-NEXT: br label [[COND_END]]
8662 // CHECK12: cond.end:
8663 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
8664 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8665 // CHECK12-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8666 // CHECK12-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
8667 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8668 // CHECK12: omp.inner.for.cond:
8669 // CHECK12-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8670 // CHECK12-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8671 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
8672 // CHECK12-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
8673 // CHECK12-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8674 // CHECK12: omp.inner.for.body:
8675 // CHECK12-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8676 // CHECK12-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8677 // CHECK12-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
8678 // CHECK12-NEXT: store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
8679 // CHECK12-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
8680 // CHECK12-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]])
8681 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8682 // CHECK12: omp.inner.for.inc:
8683 // CHECK12-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8684 // CHECK12-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8685 // CHECK12-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
8686 // CHECK12-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
8687 // CHECK12-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8688 // CHECK12-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8689 // CHECK12-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
8690 // CHECK12-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4
8691 // CHECK12-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8692 // CHECK12-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
8693 // CHECK12-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
8694 // CHECK12-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4
8695 // CHECK12-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8696 // CHECK12-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8697 // CHECK12-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
8698 // CHECK12-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
8699 // CHECK12: cond.true11:
8700 // CHECK12-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8701 // CHECK12-NEXT: br label [[COND_END13:%.*]]
8702 // CHECK12: cond.false12:
8703 // CHECK12-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8704 // CHECK12-NEXT: br label [[COND_END13]]
8705 // CHECK12: cond.end13:
8706 // CHECK12-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
8707 // CHECK12-NEXT: store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4
8708 // CHECK12-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8709 // CHECK12-NEXT: store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
8710 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
8711 // CHECK12: omp.inner.for.end:
8712 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8713 // CHECK12: omp.loop.exit:
8714 // CHECK12-NEXT: [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8715 // CHECK12-NEXT: [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
8716 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
8717 // CHECK12-NEXT: [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8718 // CHECK12-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
8719 // CHECK12-NEXT: br i1 [[TMP35]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8720 // CHECK12: .omp.final.then:
8721 // CHECK12-NEXT: [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8722 // CHECK12-NEXT: [[SUB15:%.*]] = sub nsw i32 [[TMP36]], 0
8723 // CHECK12-NEXT: [[DIV16:%.*]] = sdiv i32 [[SUB15]], 1
8724 // CHECK12-NEXT: [[MUL:%.*]] = mul nsw i32 [[DIV16]], 1
8725 // CHECK12-NEXT: [[ADD17:%.*]] = add nsw i32 0, [[MUL]]
8726 // CHECK12-NEXT: store i32 [[ADD17]], i32* [[I4]], align 4
8727 // CHECK12-NEXT: br label [[DOTOMP_FINAL_DONE]]
8728 // CHECK12: .omp.final.done:
8729 // CHECK12-NEXT: br label [[OMP_PRECOND_END]]
8730 // CHECK12: omp.precond.end:
8731 // CHECK12-NEXT: ret void
8732 //
8733 //
8734 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..6
8735 // CHECK12-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]] {
8736 // CHECK12-NEXT: entry:
8737 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8738 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8739 // CHECK12-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
8740 // CHECK12-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
8741 // CHECK12-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
8742 // CHECK12-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
8743 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
8744 // CHECK12-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
8745 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8746 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
8747 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
8748 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
8749 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
8750 // CHECK12-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
8751 // CHECK12-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
8752 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8753 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8754 // CHECK12-NEXT: [[I4:%.*]] = alloca i32, align 4
8755 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8756 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8757 // CHECK12-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
8758 // CHECK12-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
8759 // CHECK12-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
8760 // CHECK12-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
8761 // CHECK12-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
8762 // CHECK12-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
8763 // CHECK12-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
8764 // CHECK12-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
8765 // CHECK12-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
8766 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
8767 // CHECK12-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
8768 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8769 // CHECK12-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
8770 // CHECK12-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
8771 // CHECK12-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
8772 // CHECK12-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
8773 // CHECK12-NEXT: store i32 0, i32* [[I]], align 4
8774 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8775 // CHECK12-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
8776 // CHECK12-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
8777 // CHECK12: omp.precond.then:
8778 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
8779 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8780 // CHECK12-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
8781 // CHECK12-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
8782 // CHECK12-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
8783 // CHECK12-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
8784 // CHECK12-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
8785 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8786 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8787 // CHECK12-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8788 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
8789 // CHECK12-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)
8790 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8791 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8792 // CHECK12-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
8793 // CHECK12-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8794 // CHECK12: cond.true:
8795 // CHECK12-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
8796 // CHECK12-NEXT: br label [[COND_END:%.*]]
8797 // CHECK12: cond.false:
8798 // CHECK12-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8799 // CHECK12-NEXT: br label [[COND_END]]
8800 // CHECK12: cond.end:
8801 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
8802 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8803 // CHECK12-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8804 // CHECK12-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
8805 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8806 // CHECK12: omp.inner.for.cond:
8807 // CHECK12-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8808 // CHECK12-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8809 // CHECK12-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
8810 // CHECK12-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8811 // CHECK12: omp.inner.for.body:
8812 // CHECK12-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8813 // CHECK12-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
8814 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8815 // CHECK12-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
8816 // CHECK12-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
8817 // CHECK12-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
8818 // CHECK12-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
8819 // CHECK12-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
8820 // CHECK12: omp.body.continue:
8821 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8822 // CHECK12: omp.inner.for.inc:
8823 // CHECK12-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8824 // CHECK12-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
8825 // CHECK12-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
8826 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP17:![0-9]+]]
8827 // CHECK12: omp.inner.for.end:
8828 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
8829 // CHECK12: omp.loop.exit:
8830 // CHECK12-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8831 // CHECK12-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
8832 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
8833 // CHECK12-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8834 // CHECK12-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
8835 // CHECK12-NEXT: br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8836 // CHECK12: .omp.final.then:
8837 // CHECK12-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
8838 // CHECK12-NEXT: [[SUB8:%.*]] = sub nsw i32 [[TMP25]], 0
8839 // CHECK12-NEXT: [[DIV9:%.*]] = sdiv i32 [[SUB8]], 1
8840 // CHECK12-NEXT: [[MUL10:%.*]] = mul nsw i32 [[DIV9]], 1
8841 // CHECK12-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
8842 // CHECK12-NEXT: store i32 [[ADD11]], i32* [[I4]], align 4
8843 // CHECK12-NEXT: br label [[DOTOMP_FINAL_DONE]]
8844 // CHECK12: .omp.final.done:
8845 // CHECK12-NEXT: br label [[OMP_PRECOND_END]]
8846 // CHECK12: omp.precond.end:
8847 // CHECK12-NEXT: ret void
8848 //
8849 //
8850 // CHECK12-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
8851 // CHECK12-SAME: (i32 [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
8852 // CHECK12-NEXT: entry:
8853 // CHECK12-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
8854 // CHECK12-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
8855 // CHECK12-NEXT: [[M:%.*]] = alloca i32, align 4
8856 // CHECK12-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
8857 // CHECK12-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
8858 // CHECK12-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
8859 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
8860 // CHECK12-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
8861 // CHECK12-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
8862 // CHECK12-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
8863 // CHECK12-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
8864 // CHECK12-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
8865 // CHECK12-NEXT: [[DOTOFFLOAD_BASEPTRS7:%.*]] = alloca [2 x i8*], align 4
8866 // CHECK12-NEXT: [[DOTOFFLOAD_PTRS8:%.*]] = alloca [2 x i8*], align 4
8867 // CHECK12-NEXT: [[DOTOFFLOAD_MAPPERS9:%.*]] = alloca [2 x i8*], align 4
8868 // CHECK12-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
8869 // CHECK12-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
8870 // CHECK12-NEXT: store i32 10, i32* [[M]], align 4
8871 // CHECK12-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8872 // CHECK12-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
8873 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
8874 // CHECK12-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8875 // CHECK12-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
8876 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
8877 // CHECK12-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
8878 // CHECK12-NEXT: store i8* null, i8** [[TMP4]], align 4
8879 // CHECK12-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8880 // CHECK12-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8881 // CHECK12-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
8882 // CHECK12-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86.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.10, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8883 // CHECK12-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
8884 // CHECK12-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8885 // CHECK12: omp_offload.failed:
8886 // CHECK12-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86([10 x i32]* [[A]]) #[[ATTR3]]
8887 // CHECK12-NEXT: br label [[OMP_OFFLOAD_CONT]]
8888 // CHECK12: omp_offload.cont:
8889 // CHECK12-NEXT: [[TMP9:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
8890 // CHECK12-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [10 x i32]**
8891 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP10]], align 4
8892 // CHECK12-NEXT: [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
8893 // CHECK12-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to [10 x i32]**
8894 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP12]], align 4
8895 // CHECK12-NEXT: [[TMP13:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
8896 // CHECK12-NEXT: store i8* null, i8** [[TMP13]], align 4
8897 // CHECK12-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
8898 // CHECK12-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
8899 // CHECK12-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
8900 // CHECK12-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91.region_id, i32 1, i8** [[TMP14]], i8** [[TMP15]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.13, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.14, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8901 // CHECK12-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
8902 // CHECK12-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
8903 // CHECK12: omp_offload.failed5:
8904 // CHECK12-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91([10 x i32]* [[A]]) #[[ATTR3]]
8905 // CHECK12-NEXT: br label [[OMP_OFFLOAD_CONT6]]
8906 // CHECK12: omp_offload.cont6:
8907 // CHECK12-NEXT: [[TMP18:%.*]] = load i32, i32* [[M]], align 4
8908 // CHECK12-NEXT: store i32 [[TMP18]], i32* [[M_CASTED]], align 4
8909 // CHECK12-NEXT: [[TMP19:%.*]] = load i32, i32* [[M_CASTED]], align 4
8910 // CHECK12-NEXT: [[TMP20:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
8911 // CHECK12-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32*
8912 // CHECK12-NEXT: store i32 [[TMP19]], i32* [[TMP21]], align 4
8913 // CHECK12-NEXT: [[TMP22:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
8914 // CHECK12-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to i32*
8915 // CHECK12-NEXT: store i32 [[TMP19]], i32* [[TMP23]], align 4
8916 // CHECK12-NEXT: [[TMP24:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 0
8917 // CHECK12-NEXT: store i8* null, i8** [[TMP24]], align 4
8918 // CHECK12-NEXT: [[TMP25:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 1
8919 // CHECK12-NEXT: [[TMP26:%.*]] = bitcast i8** [[TMP25]] to [10 x i32]**
8920 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP26]], align 4
8921 // CHECK12-NEXT: [[TMP27:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 1
8922 // CHECK12-NEXT: [[TMP28:%.*]] = bitcast i8** [[TMP27]] to [10 x i32]**
8923 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP28]], align 4
8924 // CHECK12-NEXT: [[TMP29:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS9]], i32 0, i32 1
8925 // CHECK12-NEXT: store i8* null, i8** [[TMP29]], align 4
8926 // CHECK12-NEXT: [[TMP30:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS7]], i32 0, i32 0
8927 // CHECK12-NEXT: [[TMP31:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS8]], i32 0, i32 0
8928 // CHECK12-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 10)
8929 // CHECK12-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96.region_id, i32 2, i8** [[TMP30]], i8** [[TMP31]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.17, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.18, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
8930 // CHECK12-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
8931 // CHECK12-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED11:%.*]], label [[OMP_OFFLOAD_CONT12:%.*]]
8932 // CHECK12: omp_offload.failed11:
8933 // CHECK12-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96(i32 [[TMP19]], [10 x i32]* [[A]]) #[[ATTR3]]
8934 // CHECK12-NEXT: br label [[OMP_OFFLOAD_CONT12]]
8935 // CHECK12: omp_offload.cont12:
8936 // CHECK12-NEXT: ret i32 0
8937 //
8938 //
8939 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l86
8940 // CHECK12-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
8941 // CHECK12-NEXT: entry:
8942 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
8943 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
8944 // CHECK12-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
8945 // CHECK12-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..8 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
8946 // CHECK12-NEXT: ret void
8947 //
8948 //
8949 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..8
8950 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
8951 // CHECK12-NEXT: entry:
8952 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
8953 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
8954 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
8955 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
8956 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
8957 // CHECK12-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8958 // CHECK12-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8959 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8960 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8961 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
8962 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
8963 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
8964 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
8965 // CHECK12-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
8966 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8967 // CHECK12-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
8968 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8969 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8970 // CHECK12-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
8971 // CHECK12-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
8972 // CHECK12-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)
8973 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8974 // CHECK12-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
8975 // CHECK12-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8976 // CHECK12: cond.true:
8977 // CHECK12-NEXT: br label [[COND_END:%.*]]
8978 // CHECK12: cond.false:
8979 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8980 // CHECK12-NEXT: br label [[COND_END]]
8981 // CHECK12: cond.end:
8982 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
8983 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8984 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8985 // CHECK12-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
8986 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
8987 // CHECK12: omp.inner.for.cond:
8988 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8989 // CHECK12-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8990 // CHECK12-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
8991 // CHECK12-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8992 // CHECK12: omp.inner.for.body:
8993 // CHECK12-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8994 // CHECK12-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8995 // CHECK12-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..9 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
8996 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
8997 // CHECK12: omp.inner.for.inc:
8998 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
8999 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9000 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
9001 // CHECK12-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9002 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP18:![0-9]+]]
9003 // CHECK12: omp.inner.for.end:
9004 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9005 // CHECK12: omp.loop.exit:
9006 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
9007 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9008 // CHECK12-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
9009 // CHECK12-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9010 // CHECK12: .omp.final.then:
9011 // CHECK12-NEXT: store i32 10, i32* [[I]], align 4
9012 // CHECK12-NEXT: br label [[DOTOMP_FINAL_DONE]]
9013 // CHECK12: .omp.final.done:
9014 // CHECK12-NEXT: ret void
9015 //
9016 //
9017 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..9
9018 // CHECK12-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]] {
9019 // CHECK12-NEXT: entry:
9020 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9021 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9022 // CHECK12-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
9023 // CHECK12-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
9024 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
9025 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9026 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
9027 // CHECK12-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9028 // CHECK12-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9029 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9030 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9031 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
9032 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9033 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9034 // CHECK12-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
9035 // CHECK12-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
9036 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
9037 // CHECK12-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
9038 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9039 // CHECK12-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
9040 // CHECK12-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
9041 // CHECK12-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
9042 // CHECK12-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
9043 // CHECK12-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
9044 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9045 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9046 // CHECK12-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9047 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
9048 // CHECK12-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)
9049 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9050 // CHECK12-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
9051 // CHECK12-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9052 // CHECK12: cond.true:
9053 // CHECK12-NEXT: br label [[COND_END:%.*]]
9054 // CHECK12: cond.false:
9055 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9056 // CHECK12-NEXT: br label [[COND_END]]
9057 // CHECK12: cond.end:
9058 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
9059 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9060 // CHECK12-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9061 // CHECK12-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
9062 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9063 // CHECK12: omp.inner.for.cond:
9064 // CHECK12-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9065 // CHECK12-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9066 // CHECK12-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
9067 // CHECK12-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9068 // CHECK12: omp.inner.for.body:
9069 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9070 // CHECK12-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
9071 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9072 // CHECK12-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9073 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
9074 // CHECK12-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
9075 // CHECK12-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
9076 // CHECK12-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9077 // CHECK12: omp.body.continue:
9078 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9079 // CHECK12: omp.inner.for.inc:
9080 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9081 // CHECK12-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
9082 // CHECK12-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
9083 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
9084 // CHECK12: omp.inner.for.end:
9085 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9086 // CHECK12: omp.loop.exit:
9087 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
9088 // CHECK12-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9089 // CHECK12-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
9090 // CHECK12-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9091 // CHECK12: .omp.final.then:
9092 // CHECK12-NEXT: store i32 10, i32* [[I]], align 4
9093 // CHECK12-NEXT: br label [[DOTOMP_FINAL_DONE]]
9094 // CHECK12: .omp.final.done:
9095 // CHECK12-NEXT: ret void
9096 //
9097 //
9098 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l91
9099 // CHECK12-SAME: ([10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9100 // CHECK12-NEXT: entry:
9101 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
9102 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
9103 // CHECK12-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
9104 // CHECK12-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..11 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
9105 // CHECK12-NEXT: ret void
9106 //
9107 //
9108 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..11
9109 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9110 // CHECK12-NEXT: entry:
9111 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9112 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9113 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
9114 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9115 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
9116 // CHECK12-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9117 // CHECK12-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9118 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9119 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9120 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
9121 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9122 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9123 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
9124 // CHECK12-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
9125 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9126 // CHECK12-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
9127 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9128 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9129 // CHECK12-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9130 // CHECK12-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9131 // CHECK12-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)
9132 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9133 // CHECK12-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
9134 // CHECK12-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9135 // CHECK12: cond.true:
9136 // CHECK12-NEXT: br label [[COND_END:%.*]]
9137 // CHECK12: cond.false:
9138 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9139 // CHECK12-NEXT: br label [[COND_END]]
9140 // CHECK12: cond.end:
9141 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9142 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9143 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9144 // CHECK12-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
9145 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9146 // CHECK12: omp.inner.for.cond:
9147 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9148 // CHECK12-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9149 // CHECK12-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9150 // CHECK12-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9151 // CHECK12: omp.inner.for.body:
9152 // CHECK12-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9153 // CHECK12-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9154 // CHECK12-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..12 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
9155 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9156 // CHECK12: omp.inner.for.inc:
9157 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9158 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9159 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
9160 // CHECK12-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9161 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
9162 // CHECK12: omp.inner.for.end:
9163 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9164 // CHECK12: omp.loop.exit:
9165 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
9166 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9167 // CHECK12-NEXT: [[TMP13:%.*]] = icmp ne i32 [[TMP12]], 0
9168 // CHECK12-NEXT: br i1 [[TMP13]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9169 // CHECK12: .omp.final.then:
9170 // CHECK12-NEXT: store i32 10, i32* [[I]], align 4
9171 // CHECK12-NEXT: br label [[DOTOMP_FINAL_DONE]]
9172 // CHECK12: .omp.final.done:
9173 // CHECK12-NEXT: ret void
9174 //
9175 //
9176 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..12
9177 // CHECK12-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]] {
9178 // CHECK12-NEXT: entry:
9179 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9180 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9181 // CHECK12-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
9182 // CHECK12-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
9183 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
9184 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9185 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
9186 // CHECK12-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9187 // CHECK12-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9188 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9189 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9190 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
9191 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9192 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9193 // CHECK12-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
9194 // CHECK12-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
9195 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
9196 // CHECK12-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
9197 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9198 // CHECK12-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
9199 // CHECK12-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
9200 // CHECK12-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
9201 // CHECK12-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
9202 // CHECK12-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
9203 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9204 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9205 // CHECK12-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9206 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
9207 // CHECK12-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)
9208 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9209 // CHECK12-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
9210 // CHECK12-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9211 // CHECK12: cond.true:
9212 // CHECK12-NEXT: br label [[COND_END:%.*]]
9213 // CHECK12: cond.false:
9214 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9215 // CHECK12-NEXT: br label [[COND_END]]
9216 // CHECK12: cond.end:
9217 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
9218 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9219 // CHECK12-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9220 // CHECK12-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
9221 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9222 // CHECK12: omp.inner.for.cond:
9223 // CHECK12-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9224 // CHECK12-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9225 // CHECK12-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
9226 // CHECK12-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9227 // CHECK12: omp.inner.for.body:
9228 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9229 // CHECK12-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
9230 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9231 // CHECK12-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9232 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
9233 // CHECK12-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
9234 // CHECK12-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
9235 // CHECK12-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9236 // CHECK12: omp.body.continue:
9237 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9238 // CHECK12: omp.inner.for.inc:
9239 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9240 // CHECK12-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
9241 // CHECK12-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
9242 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
9243 // CHECK12: omp.inner.for.end:
9244 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9245 // CHECK12: omp.loop.exit:
9246 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
9247 // CHECK12-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9248 // CHECK12-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
9249 // CHECK12-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9250 // CHECK12: .omp.final.then:
9251 // CHECK12-NEXT: store i32 10, i32* [[I]], align 4
9252 // CHECK12-NEXT: br label [[DOTOMP_FINAL_DONE]]
9253 // CHECK12: .omp.final.done:
9254 // CHECK12-NEXT: ret void
9255 //
9256 //
9257 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l96
9258 // CHECK12-SAME: (i32 [[M:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
9259 // CHECK12-NEXT: entry:
9260 // CHECK12-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
9261 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
9262 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
9263 // CHECK12-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
9264 // CHECK12-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
9265 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
9266 // CHECK12-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
9267 // CHECK12-NEXT: [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
9268 // CHECK12-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
9269 // CHECK12-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9270 // CHECK12-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
9271 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
9272 // CHECK12-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..15 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
9273 // CHECK12-NEXT: ret void
9274 //
9275 //
9276 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..15
9277 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]], i32 [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
9278 // CHECK12-NEXT: entry:
9279 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9280 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9281 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
9282 // CHECK12-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
9283 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9284 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
9285 // CHECK12-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9286 // CHECK12-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9287 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9288 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9289 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
9290 // CHECK12-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
9291 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9292 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9293 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
9294 // CHECK12-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9295 // CHECK12-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
9296 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9297 // CHECK12-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
9298 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9299 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9300 // CHECK12-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9301 // CHECK12-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9302 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9303 // CHECK12-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP1]])
9304 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9305 // CHECK12-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
9306 // CHECK12-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9307 // CHECK12: cond.true:
9308 // CHECK12-NEXT: br label [[COND_END:%.*]]
9309 // CHECK12: cond.false:
9310 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9311 // CHECK12-NEXT: br label [[COND_END]]
9312 // CHECK12: cond.end:
9313 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9314 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9315 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9316 // CHECK12-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9317 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9318 // CHECK12: omp.inner.for.cond:
9319 // CHECK12-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9320 // CHECK12-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP7]], 10
9321 // CHECK12-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9322 // CHECK12: omp.inner.for.body:
9323 // CHECK12-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9324 // CHECK12-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9325 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9326 // CHECK12-NEXT: store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
9327 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
9328 // CHECK12-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..16 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
9329 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9330 // CHECK12: omp.inner.for.inc:
9331 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9332 // CHECK12-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9333 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
9334 // CHECK12-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
9335 // CHECK12-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9336 // CHECK12-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9337 // CHECK12-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
9338 // CHECK12-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4
9339 // CHECK12-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9340 // CHECK12-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
9341 // CHECK12-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
9342 // CHECK12-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4
9343 // CHECK12-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9344 // CHECK12-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
9345 // CHECK12-NEXT: br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
9346 // CHECK12: cond.true5:
9347 // CHECK12-NEXT: br label [[COND_END7:%.*]]
9348 // CHECK12: cond.false6:
9349 // CHECK12-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9350 // CHECK12-NEXT: br label [[COND_END7]]
9351 // CHECK12: cond.end7:
9352 // CHECK12-NEXT: [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
9353 // CHECK12-NEXT: store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4
9354 // CHECK12-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9355 // CHECK12-NEXT: store i32 [[TMP20]], i32* [[DOTOMP_IV]], align 4
9356 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]]
9357 // CHECK12: omp.inner.for.end:
9358 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9359 // CHECK12: omp.loop.exit:
9360 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9361 // CHECK12-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9362 // CHECK12-NEXT: [[TMP22:%.*]] = icmp ne i32 [[TMP21]], 0
9363 // CHECK12-NEXT: br i1 [[TMP22]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9364 // CHECK12: .omp.final.then:
9365 // CHECK12-NEXT: store i32 10, i32* [[I]], align 4
9366 // CHECK12-NEXT: br label [[DOTOMP_FINAL_DONE]]
9367 // CHECK12: .omp.final.done:
9368 // CHECK12-NEXT: ret void
9369 //
9370 //
9371 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..16
9372 // CHECK12-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]] {
9373 // CHECK12-NEXT: entry:
9374 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
9375 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
9376 // CHECK12-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
9377 // CHECK12-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
9378 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
9379 // CHECK12-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
9380 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9381 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
9382 // CHECK12-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9383 // CHECK12-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9384 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9385 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9386 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
9387 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
9388 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
9389 // CHECK12-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
9390 // CHECK12-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
9391 // CHECK12-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
9392 // CHECK12-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
9393 // CHECK12-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
9394 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9395 // CHECK12-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
9396 // CHECK12-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
9397 // CHECK12-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
9398 // CHECK12-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
9399 // CHECK12-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
9400 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9401 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9402 // CHECK12-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
9403 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
9404 // CHECK12-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)
9405 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9406 // CHECK12-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
9407 // CHECK12-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9408 // CHECK12: cond.true:
9409 // CHECK12-NEXT: br label [[COND_END:%.*]]
9410 // CHECK12: cond.false:
9411 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9412 // CHECK12-NEXT: br label [[COND_END]]
9413 // CHECK12: cond.end:
9414 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
9415 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9416 // CHECK12-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9417 // CHECK12-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
9418 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9419 // CHECK12: omp.inner.for.cond:
9420 // CHECK12-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9421 // CHECK12-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9422 // CHECK12-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
9423 // CHECK12-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9424 // CHECK12: omp.inner.for.body:
9425 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9426 // CHECK12-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
9427 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9428 // CHECK12-NEXT: store i32 [[ADD]], i32* [[I]], align 4
9429 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
9430 // CHECK12-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
9431 // CHECK12-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
9432 // CHECK12-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9433 // CHECK12: omp.body.continue:
9434 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9435 // CHECK12: omp.inner.for.inc:
9436 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
9437 // CHECK12-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
9438 // CHECK12-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
9439 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
9440 // CHECK12: omp.inner.for.end:
9441 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
9442 // CHECK12: omp.loop.exit:
9443 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
9444 // CHECK12-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9445 // CHECK12-NEXT: [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
9446 // CHECK12-NEXT: br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9447 // CHECK12: .omp.final.then:
9448 // CHECK12-NEXT: store i32 10, i32* [[I]], align 4
9449 // CHECK12-NEXT: br label [[DOTOMP_FINAL_DONE]]
9450 // CHECK12: .omp.final.done:
9451 // CHECK12-NEXT: ret void
9452 //
9453 //
9454 // CHECK12-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
9455 // CHECK12-SAME: () #[[ATTR5:[0-9]+]] {
9456 // CHECK12-NEXT: entry:
9457 // CHECK12-NEXT: call void @__tgt_register_requires(i64 1)
9458 // CHECK12-NEXT: ret void
9459 //
9460 //
9461 // CHECK13-LABEL: define {{[^@]+}}@main
9462 // CHECK13-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
9463 // CHECK13-NEXT: entry:
9464 // CHECK13-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
9465 // CHECK13-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
9466 // CHECK13-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
9467 // CHECK13-NEXT: [[N:%.*]] = alloca i32, align 4
9468 // CHECK13-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
9469 // CHECK13-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
9470 // CHECK13-NEXT: [[M:%.*]] = alloca i32, align 4
9471 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9472 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
9473 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
9474 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9475 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9476 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9477 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9478 // CHECK13-NEXT: [[I3:%.*]] = alloca i32, align 4
9479 // CHECK13-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
9480 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
9481 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
9482 // CHECK13-NEXT: [[DOTOMP_LB16:%.*]] = alloca i32, align 4
9483 // CHECK13-NEXT: [[DOTOMP_UB17:%.*]] = alloca i32, align 4
9484 // CHECK13-NEXT: [[I18:%.*]] = alloca i32, align 4
9485 // CHECK13-NEXT: [[DOTOMP_IV21:%.*]] = alloca i32, align 4
9486 // CHECK13-NEXT: [[I22:%.*]] = alloca i32, align 4
9487 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_39:%.*]] = alloca i32, align 4
9488 // CHECK13-NEXT: [[_TMP40:%.*]] = alloca i32, align 4
9489 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
9490 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
9491 // CHECK13-NEXT: [[DOTOMP_LB46:%.*]] = alloca i32, align 4
9492 // CHECK13-NEXT: [[DOTOMP_UB47:%.*]] = alloca i32, align 4
9493 // CHECK13-NEXT: [[I48:%.*]] = alloca i32, align 4
9494 // CHECK13-NEXT: [[DOTOMP_IV51:%.*]] = alloca i32, align 4
9495 // CHECK13-NEXT: [[I52:%.*]] = alloca i32, align 4
9496 // CHECK13-NEXT: store i32 0, i32* [[RETVAL]], align 4
9497 // CHECK13-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
9498 // CHECK13-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
9499 // CHECK13-NEXT: store i32 100, i32* [[N]], align 4
9500 // CHECK13-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
9501 // CHECK13-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
9502 // CHECK13-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
9503 // CHECK13-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
9504 // CHECK13-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
9505 // CHECK13-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
9506 // CHECK13-NEXT: store i32 10, i32* [[M]], align 4
9507 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
9508 // CHECK13-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
9509 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9510 // CHECK13-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
9511 // CHECK13-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
9512 // CHECK13-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
9513 // CHECK13-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
9514 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9515 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
9516 // CHECK13-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
9517 // CHECK13-NEXT: store i32 0, i32* [[I]], align 4
9518 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9519 // CHECK13-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
9520 // CHECK13-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
9521 // CHECK13: simd.if.then:
9522 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9523 // CHECK13-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
9524 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9525 // CHECK13: omp.inner.for.cond:
9526 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
9527 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
9528 // CHECK13-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
9529 // CHECK13-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9530 // CHECK13: omp.inner.for.body:
9531 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
9532 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
9533 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9534 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !2
9535 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !2
9536 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
9537 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM]]
9538 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
9539 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9540 // CHECK13: omp.body.continue:
9541 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9542 // CHECK13: omp.inner.for.inc:
9543 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
9544 // CHECK13-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
9545 // CHECK13-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
9546 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
9547 // CHECK13: omp.inner.for.end:
9548 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9549 // CHECK13-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP13]], 0
9550 // CHECK13-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
9551 // CHECK13-NEXT: [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
9552 // CHECK13-NEXT: [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
9553 // CHECK13-NEXT: store i32 [[ADD9]], i32* [[I3]], align 4
9554 // CHECK13-NEXT: br label [[SIMD_IF_END]]
9555 // CHECK13: simd.if.end:
9556 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[N]], align 4
9557 // CHECK13-NEXT: store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
9558 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
9559 // CHECK13-NEXT: [[SUB13:%.*]] = sub nsw i32 [[TMP15]], 0
9560 // CHECK13-NEXT: [[DIV14:%.*]] = sdiv i32 [[SUB13]], 1
9561 // CHECK13-NEXT: [[SUB15:%.*]] = sub nsw i32 [[DIV14]], 1
9562 // CHECK13-NEXT: store i32 [[SUB15]], i32* [[DOTCAPTURE_EXPR_12]], align 4
9563 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB16]], align 4
9564 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
9565 // CHECK13-NEXT: store i32 [[TMP16]], i32* [[DOTOMP_UB17]], align 4
9566 // CHECK13-NEXT: store i32 0, i32* [[I18]], align 4
9567 // CHECK13-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
9568 // CHECK13-NEXT: [[CMP19:%.*]] = icmp slt i32 0, [[TMP17]]
9569 // CHECK13-NEXT: br i1 [[CMP19]], label [[SIMD_IF_THEN20:%.*]], label [[SIMD_IF_END38:%.*]]
9570 // CHECK13: simd.if.then20:
9571 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB16]], align 4
9572 // CHECK13-NEXT: store i32 [[TMP18]], i32* [[DOTOMP_IV21]], align 4
9573 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND23:%.*]]
9574 // CHECK13: omp.inner.for.cond23:
9575 // CHECK13-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
9576 // CHECK13-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB17]], align 4, !llvm.access.group !6
9577 // CHECK13-NEXT: [[CMP24:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
9578 // CHECK13-NEXT: br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
9579 // CHECK13: omp.inner.for.body25:
9580 // CHECK13-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
9581 // CHECK13-NEXT: [[MUL26:%.*]] = mul nsw i32 [[TMP21]], 1
9582 // CHECK13-NEXT: [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
9583 // CHECK13-NEXT: store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !6
9584 // CHECK13-NEXT: [[TMP22:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !6
9585 // CHECK13-NEXT: [[IDXPROM28:%.*]] = sext i32 [[TMP22]] to i64
9586 // CHECK13-NEXT: [[ARRAYIDX29:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM28]]
9587 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX29]], align 4, !llvm.access.group !6
9588 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE30:%.*]]
9589 // CHECK13: omp.body.continue30:
9590 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC31:%.*]]
9591 // CHECK13: omp.inner.for.inc31:
9592 // CHECK13-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
9593 // CHECK13-NEXT: [[ADD32:%.*]] = add nsw i32 [[TMP23]], 1
9594 // CHECK13-NEXT: store i32 [[ADD32]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
9595 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP7:![0-9]+]]
9596 // CHECK13: omp.inner.for.end33:
9597 // CHECK13-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
9598 // CHECK13-NEXT: [[SUB34:%.*]] = sub nsw i32 [[TMP24]], 0
9599 // CHECK13-NEXT: [[DIV35:%.*]] = sdiv i32 [[SUB34]], 1
9600 // CHECK13-NEXT: [[MUL36:%.*]] = mul nsw i32 [[DIV35]], 1
9601 // CHECK13-NEXT: [[ADD37:%.*]] = add nsw i32 0, [[MUL36]]
9602 // CHECK13-NEXT: store i32 [[ADD37]], i32* [[I22]], align 4
9603 // CHECK13-NEXT: br label [[SIMD_IF_END38]]
9604 // CHECK13: simd.if.end38:
9605 // CHECK13-NEXT: [[TMP25:%.*]] = load i32, i32* [[M]], align 4
9606 // CHECK13-NEXT: store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_39]], align 4
9607 // CHECK13-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
9608 // CHECK13-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_41]], align 4
9609 // CHECK13-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
9610 // CHECK13-NEXT: [[SUB43:%.*]] = sub nsw i32 [[TMP27]], 0
9611 // CHECK13-NEXT: [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
9612 // CHECK13-NEXT: [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
9613 // CHECK13-NEXT: store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
9614 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB46]], align 4
9615 // CHECK13-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
9616 // CHECK13-NEXT: store i32 [[TMP28]], i32* [[DOTOMP_UB47]], align 4
9617 // CHECK13-NEXT: store i32 0, i32* [[I48]], align 4
9618 // CHECK13-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
9619 // CHECK13-NEXT: [[CMP49:%.*]] = icmp slt i32 0, [[TMP29]]
9620 // CHECK13-NEXT: br i1 [[CMP49]], label [[SIMD_IF_THEN50:%.*]], label [[SIMD_IF_END68:%.*]]
9621 // CHECK13: simd.if.then50:
9622 // CHECK13-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_LB46]], align 4
9623 // CHECK13-NEXT: store i32 [[TMP30]], i32* [[DOTOMP_IV51]], align 4
9624 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND53:%.*]]
9625 // CHECK13: omp.inner.for.cond53:
9626 // CHECK13-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
9627 // CHECK13-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_UB47]], align 4, !llvm.access.group !9
9628 // CHECK13-NEXT: [[CMP54:%.*]] = icmp sle i32 [[TMP31]], [[TMP32]]
9629 // CHECK13-NEXT: br i1 [[CMP54]], label [[OMP_INNER_FOR_BODY55:%.*]], label [[OMP_INNER_FOR_END63:%.*]]
9630 // CHECK13: omp.inner.for.body55:
9631 // CHECK13-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
9632 // CHECK13-NEXT: [[MUL56:%.*]] = mul nsw i32 [[TMP33]], 1
9633 // CHECK13-NEXT: [[ADD57:%.*]] = add nsw i32 0, [[MUL56]]
9634 // CHECK13-NEXT: store i32 [[ADD57]], i32* [[I52]], align 4, !llvm.access.group !9
9635 // CHECK13-NEXT: [[TMP34:%.*]] = load i32, i32* [[I52]], align 4, !llvm.access.group !9
9636 // CHECK13-NEXT: [[IDXPROM58:%.*]] = sext i32 [[TMP34]] to i64
9637 // CHECK13-NEXT: [[ARRAYIDX59:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM58]]
9638 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX59]], align 4, !llvm.access.group !9
9639 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE60:%.*]]
9640 // CHECK13: omp.body.continue60:
9641 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC61:%.*]]
9642 // CHECK13: omp.inner.for.inc61:
9643 // CHECK13-NEXT: [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
9644 // CHECK13-NEXT: [[ADD62:%.*]] = add nsw i32 [[TMP35]], 1
9645 // CHECK13-NEXT: store i32 [[ADD62]], i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
9646 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND53]], !llvm.loop [[LOOP10:![0-9]+]]
9647 // CHECK13: omp.inner.for.end63:
9648 // CHECK13-NEXT: [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
9649 // CHECK13-NEXT: [[SUB64:%.*]] = sub nsw i32 [[TMP36]], 0
9650 // CHECK13-NEXT: [[DIV65:%.*]] = sdiv i32 [[SUB64]], 1
9651 // CHECK13-NEXT: [[MUL66:%.*]] = mul nsw i32 [[DIV65]], 1
9652 // CHECK13-NEXT: [[ADD67:%.*]] = add nsw i32 0, [[MUL66]]
9653 // CHECK13-NEXT: store i32 [[ADD67]], i32* [[I52]], align 4
9654 // CHECK13-NEXT: br label [[SIMD_IF_END68]]
9655 // CHECK13: simd.if.end68:
9656 // CHECK13-NEXT: [[TMP37:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
9657 // CHECK13-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP37]])
9658 // CHECK13-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
9659 // CHECK13-NEXT: [[TMP38:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
9660 // CHECK13-NEXT: call void @llvm.stackrestore(i8* [[TMP38]])
9661 // CHECK13-NEXT: [[TMP39:%.*]] = load i32, i32* [[RETVAL]], align 4
9662 // CHECK13-NEXT: ret i32 [[TMP39]]
9663 //
9664 //
9665 // CHECK13-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
9666 // CHECK13-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
9667 // CHECK13-NEXT: entry:
9668 // CHECK13-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
9669 // CHECK13-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
9670 // CHECK13-NEXT: [[M:%.*]] = alloca i32, align 4
9671 // CHECK13-NEXT: [[TMP:%.*]] = alloca i32, align 4
9672 // CHECK13-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9673 // CHECK13-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9674 // CHECK13-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9675 // CHECK13-NEXT: [[I:%.*]] = alloca i32, align 4
9676 // CHECK13-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
9677 // CHECK13-NEXT: [[DOTOMP_LB3:%.*]] = alloca i32, align 4
9678 // CHECK13-NEXT: [[DOTOMP_UB4:%.*]] = alloca i32, align 4
9679 // CHECK13-NEXT: [[DOTOMP_IV5:%.*]] = alloca i32, align 4
9680 // CHECK13-NEXT: [[I6:%.*]] = alloca i32, align 4
9681 // CHECK13-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
9682 // CHECK13-NEXT: [[_TMP18:%.*]] = alloca i32, align 4
9683 // CHECK13-NEXT: [[DOTOMP_LB19:%.*]] = alloca i32, align 4
9684 // CHECK13-NEXT: [[DOTOMP_UB20:%.*]] = alloca i32, align 4
9685 // CHECK13-NEXT: [[DOTOMP_IV21:%.*]] = alloca i32, align 4
9686 // CHECK13-NEXT: [[I22:%.*]] = alloca i32, align 4
9687 // CHECK13-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
9688 // CHECK13-NEXT: store i32 10, i32* [[M]], align 4
9689 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9690 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
9691 // CHECK13-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9692 // CHECK13-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
9693 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9694 // CHECK13: omp.inner.for.cond:
9695 // CHECK13-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
9696 // CHECK13-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
9697 // CHECK13-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
9698 // CHECK13-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9699 // CHECK13: omp.inner.for.body:
9700 // CHECK13-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
9701 // CHECK13-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
9702 // CHECK13-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9703 // CHECK13-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
9704 // CHECK13-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12
9705 // CHECK13-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
9706 // CHECK13-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
9707 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !12
9708 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9709 // CHECK13: omp.body.continue:
9710 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9711 // CHECK13: omp.inner.for.inc:
9712 // CHECK13-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
9713 // CHECK13-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
9714 // CHECK13-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
9715 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
9716 // CHECK13: omp.inner.for.end:
9717 // CHECK13-NEXT: store i32 10, i32* [[I]], align 4
9718 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB3]], align 4
9719 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_UB4]], align 4
9720 // CHECK13-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
9721 // CHECK13-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV5]], align 4
9722 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND7:%.*]]
9723 // CHECK13: omp.inner.for.cond7:
9724 // CHECK13-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !15
9725 // CHECK13-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !15
9726 // CHECK13-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9727 // CHECK13-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END17:%.*]]
9728 // CHECK13: omp.inner.for.body9:
9729 // CHECK13-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !15
9730 // CHECK13-NEXT: [[MUL10:%.*]] = mul nsw i32 [[TMP9]], 1
9731 // CHECK13-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
9732 // CHECK13-NEXT: store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !15
9733 // CHECK13-NEXT: [[TMP10:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !15
9734 // CHECK13-NEXT: [[IDXPROM12:%.*]] = sext i32 [[TMP10]] to i64
9735 // CHECK13-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM12]]
9736 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX13]], align 4, !llvm.access.group !15
9737 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE14:%.*]]
9738 // CHECK13: omp.body.continue14:
9739 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC15:%.*]]
9740 // CHECK13: omp.inner.for.inc15:
9741 // CHECK13-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !15
9742 // CHECK13-NEXT: [[ADD16:%.*]] = add nsw i32 [[TMP11]], 1
9743 // CHECK13-NEXT: store i32 [[ADD16]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !15
9744 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP16:![0-9]+]]
9745 // CHECK13: omp.inner.for.end17:
9746 // CHECK13-NEXT: store i32 10, i32* [[I6]], align 4
9747 // CHECK13-NEXT: [[TMP12:%.*]] = load i32, i32* [[M]], align 4
9748 // CHECK13-NEXT: store i32 [[TMP12]], i32* [[DOTCAPTURE_EXPR_]], align 4
9749 // CHECK13-NEXT: store i32 0, i32* [[DOTOMP_LB19]], align 4
9750 // CHECK13-NEXT: store i32 9, i32* [[DOTOMP_UB20]], align 4
9751 // CHECK13-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB19]], align 4
9752 // CHECK13-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV21]], align 4
9753 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND23:%.*]]
9754 // CHECK13: omp.inner.for.cond23:
9755 // CHECK13-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !18
9756 // CHECK13-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB20]], align 4, !llvm.access.group !18
9757 // CHECK13-NEXT: [[CMP24:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
9758 // CHECK13-NEXT: br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
9759 // CHECK13: omp.inner.for.body25:
9760 // CHECK13-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !18
9761 // CHECK13-NEXT: [[MUL26:%.*]] = mul nsw i32 [[TMP16]], 1
9762 // CHECK13-NEXT: [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
9763 // CHECK13-NEXT: store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !18
9764 // CHECK13-NEXT: [[TMP17:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !18
9765 // CHECK13-NEXT: [[IDXPROM28:%.*]] = sext i32 [[TMP17]] to i64
9766 // CHECK13-NEXT: [[ARRAYIDX29:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM28]]
9767 // CHECK13-NEXT: store i32 0, i32* [[ARRAYIDX29]], align 4, !llvm.access.group !18
9768 // CHECK13-NEXT: br label [[OMP_BODY_CONTINUE30:%.*]]
9769 // CHECK13: omp.body.continue30:
9770 // CHECK13-NEXT: br label [[OMP_INNER_FOR_INC31:%.*]]
9771 // CHECK13: omp.inner.for.inc31:
9772 // CHECK13-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !18
9773 // CHECK13-NEXT: [[ADD32:%.*]] = add nsw i32 [[TMP18]], 1
9774 // CHECK13-NEXT: store i32 [[ADD32]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !18
9775 // CHECK13-NEXT: br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP19:![0-9]+]]
9776 // CHECK13: omp.inner.for.end33:
9777 // CHECK13-NEXT: store i32 10, i32* [[I22]], align 4
9778 // CHECK13-NEXT: ret i32 0
9779 //
9780 //
9781 // CHECK14-LABEL: define {{[^@]+}}@main
9782 // CHECK14-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
9783 // CHECK14-NEXT: entry:
9784 // CHECK14-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
9785 // CHECK14-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
9786 // CHECK14-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
9787 // CHECK14-NEXT: [[N:%.*]] = alloca i32, align 4
9788 // CHECK14-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
9789 // CHECK14-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
9790 // CHECK14-NEXT: [[M:%.*]] = alloca i32, align 4
9791 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9792 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
9793 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
9794 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9795 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9796 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
9797 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9798 // CHECK14-NEXT: [[I3:%.*]] = alloca i32, align 4
9799 // CHECK14-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
9800 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
9801 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
9802 // CHECK14-NEXT: [[DOTOMP_LB16:%.*]] = alloca i32, align 4
9803 // CHECK14-NEXT: [[DOTOMP_UB17:%.*]] = alloca i32, align 4
9804 // CHECK14-NEXT: [[I18:%.*]] = alloca i32, align 4
9805 // CHECK14-NEXT: [[DOTOMP_IV21:%.*]] = alloca i32, align 4
9806 // CHECK14-NEXT: [[I22:%.*]] = alloca i32, align 4
9807 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_39:%.*]] = alloca i32, align 4
9808 // CHECK14-NEXT: [[_TMP40:%.*]] = alloca i32, align 4
9809 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
9810 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_42:%.*]] = alloca i32, align 4
9811 // CHECK14-NEXT: [[DOTOMP_LB46:%.*]] = alloca i32, align 4
9812 // CHECK14-NEXT: [[DOTOMP_UB47:%.*]] = alloca i32, align 4
9813 // CHECK14-NEXT: [[I48:%.*]] = alloca i32, align 4
9814 // CHECK14-NEXT: [[DOTOMP_IV51:%.*]] = alloca i32, align 4
9815 // CHECK14-NEXT: [[I52:%.*]] = alloca i32, align 4
9816 // CHECK14-NEXT: store i32 0, i32* [[RETVAL]], align 4
9817 // CHECK14-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
9818 // CHECK14-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
9819 // CHECK14-NEXT: store i32 100, i32* [[N]], align 4
9820 // CHECK14-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
9821 // CHECK14-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
9822 // CHECK14-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
9823 // CHECK14-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
9824 // CHECK14-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
9825 // CHECK14-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
9826 // CHECK14-NEXT: store i32 10, i32* [[M]], align 4
9827 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
9828 // CHECK14-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
9829 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9830 // CHECK14-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
9831 // CHECK14-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
9832 // CHECK14-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
9833 // CHECK14-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
9834 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
9835 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
9836 // CHECK14-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
9837 // CHECK14-NEXT: store i32 0, i32* [[I]], align 4
9838 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9839 // CHECK14-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
9840 // CHECK14-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
9841 // CHECK14: simd.if.then:
9842 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9843 // CHECK14-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
9844 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
9845 // CHECK14: omp.inner.for.cond:
9846 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
9847 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
9848 // CHECK14-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
9849 // CHECK14-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9850 // CHECK14: omp.inner.for.body:
9851 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
9852 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
9853 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9854 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !2
9855 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !2
9856 // CHECK14-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
9857 // CHECK14-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM]]
9858 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !2
9859 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
9860 // CHECK14: omp.body.continue:
9861 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
9862 // CHECK14: omp.inner.for.inc:
9863 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
9864 // CHECK14-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP12]], 1
9865 // CHECK14-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
9866 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
9867 // CHECK14: omp.inner.for.end:
9868 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
9869 // CHECK14-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP13]], 0
9870 // CHECK14-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
9871 // CHECK14-NEXT: [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
9872 // CHECK14-NEXT: [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
9873 // CHECK14-NEXT: store i32 [[ADD9]], i32* [[I3]], align 4
9874 // CHECK14-NEXT: br label [[SIMD_IF_END]]
9875 // CHECK14: simd.if.end:
9876 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[N]], align 4
9877 // CHECK14-NEXT: store i32 [[TMP14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
9878 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
9879 // CHECK14-NEXT: [[SUB13:%.*]] = sub nsw i32 [[TMP15]], 0
9880 // CHECK14-NEXT: [[DIV14:%.*]] = sdiv i32 [[SUB13]], 1
9881 // CHECK14-NEXT: [[SUB15:%.*]] = sub nsw i32 [[DIV14]], 1
9882 // CHECK14-NEXT: store i32 [[SUB15]], i32* [[DOTCAPTURE_EXPR_12]], align 4
9883 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB16]], align 4
9884 // CHECK14-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
9885 // CHECK14-NEXT: store i32 [[TMP16]], i32* [[DOTOMP_UB17]], align 4
9886 // CHECK14-NEXT: store i32 0, i32* [[I18]], align 4
9887 // CHECK14-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
9888 // CHECK14-NEXT: [[CMP19:%.*]] = icmp slt i32 0, [[TMP17]]
9889 // CHECK14-NEXT: br i1 [[CMP19]], label [[SIMD_IF_THEN20:%.*]], label [[SIMD_IF_END38:%.*]]
9890 // CHECK14: simd.if.then20:
9891 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_LB16]], align 4
9892 // CHECK14-NEXT: store i32 [[TMP18]], i32* [[DOTOMP_IV21]], align 4
9893 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND23:%.*]]
9894 // CHECK14: omp.inner.for.cond23:
9895 // CHECK14-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
9896 // CHECK14-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_UB17]], align 4, !llvm.access.group !6
9897 // CHECK14-NEXT: [[CMP24:%.*]] = icmp sle i32 [[TMP19]], [[TMP20]]
9898 // CHECK14-NEXT: br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
9899 // CHECK14: omp.inner.for.body25:
9900 // CHECK14-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
9901 // CHECK14-NEXT: [[MUL26:%.*]] = mul nsw i32 [[TMP21]], 1
9902 // CHECK14-NEXT: [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
9903 // CHECK14-NEXT: store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !6
9904 // CHECK14-NEXT: [[TMP22:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !6
9905 // CHECK14-NEXT: [[IDXPROM28:%.*]] = sext i32 [[TMP22]] to i64
9906 // CHECK14-NEXT: [[ARRAYIDX29:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM28]]
9907 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX29]], align 4, !llvm.access.group !6
9908 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE30:%.*]]
9909 // CHECK14: omp.body.continue30:
9910 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC31:%.*]]
9911 // CHECK14: omp.inner.for.inc31:
9912 // CHECK14-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
9913 // CHECK14-NEXT: [[ADD32:%.*]] = add nsw i32 [[TMP23]], 1
9914 // CHECK14-NEXT: store i32 [[ADD32]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !6
9915 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP7:![0-9]+]]
9916 // CHECK14: omp.inner.for.end33:
9917 // CHECK14-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
9918 // CHECK14-NEXT: [[SUB34:%.*]] = sub nsw i32 [[TMP24]], 0
9919 // CHECK14-NEXT: [[DIV35:%.*]] = sdiv i32 [[SUB34]], 1
9920 // CHECK14-NEXT: [[MUL36:%.*]] = mul nsw i32 [[DIV35]], 1
9921 // CHECK14-NEXT: [[ADD37:%.*]] = add nsw i32 0, [[MUL36]]
9922 // CHECK14-NEXT: store i32 [[ADD37]], i32* [[I22]], align 4
9923 // CHECK14-NEXT: br label [[SIMD_IF_END38]]
9924 // CHECK14: simd.if.end38:
9925 // CHECK14-NEXT: [[TMP25:%.*]] = load i32, i32* [[M]], align 4
9926 // CHECK14-NEXT: store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_39]], align 4
9927 // CHECK14-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
9928 // CHECK14-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_41]], align 4
9929 // CHECK14-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
9930 // CHECK14-NEXT: [[SUB43:%.*]] = sub nsw i32 [[TMP27]], 0
9931 // CHECK14-NEXT: [[DIV44:%.*]] = sdiv i32 [[SUB43]], 1
9932 // CHECK14-NEXT: [[SUB45:%.*]] = sub nsw i32 [[DIV44]], 1
9933 // CHECK14-NEXT: store i32 [[SUB45]], i32* [[DOTCAPTURE_EXPR_42]], align 4
9934 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB46]], align 4
9935 // CHECK14-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_42]], align 4
9936 // CHECK14-NEXT: store i32 [[TMP28]], i32* [[DOTOMP_UB47]], align 4
9937 // CHECK14-NEXT: store i32 0, i32* [[I48]], align 4
9938 // CHECK14-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
9939 // CHECK14-NEXT: [[CMP49:%.*]] = icmp slt i32 0, [[TMP29]]
9940 // CHECK14-NEXT: br i1 [[CMP49]], label [[SIMD_IF_THEN50:%.*]], label [[SIMD_IF_END68:%.*]]
9941 // CHECK14: simd.if.then50:
9942 // CHECK14-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_LB46]], align 4
9943 // CHECK14-NEXT: store i32 [[TMP30]], i32* [[DOTOMP_IV51]], align 4
9944 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND53:%.*]]
9945 // CHECK14: omp.inner.for.cond53:
9946 // CHECK14-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
9947 // CHECK14-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_UB47]], align 4, !llvm.access.group !9
9948 // CHECK14-NEXT: [[CMP54:%.*]] = icmp sle i32 [[TMP31]], [[TMP32]]
9949 // CHECK14-NEXT: br i1 [[CMP54]], label [[OMP_INNER_FOR_BODY55:%.*]], label [[OMP_INNER_FOR_END63:%.*]]
9950 // CHECK14: omp.inner.for.body55:
9951 // CHECK14-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
9952 // CHECK14-NEXT: [[MUL56:%.*]] = mul nsw i32 [[TMP33]], 1
9953 // CHECK14-NEXT: [[ADD57:%.*]] = add nsw i32 0, [[MUL56]]
9954 // CHECK14-NEXT: store i32 [[ADD57]], i32* [[I52]], align 4, !llvm.access.group !9
9955 // CHECK14-NEXT: [[TMP34:%.*]] = load i32, i32* [[I52]], align 4, !llvm.access.group !9
9956 // CHECK14-NEXT: [[IDXPROM58:%.*]] = sext i32 [[TMP34]] to i64
9957 // CHECK14-NEXT: [[ARRAYIDX59:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 [[IDXPROM58]]
9958 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX59]], align 4, !llvm.access.group !9
9959 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE60:%.*]]
9960 // CHECK14: omp.body.continue60:
9961 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC61:%.*]]
9962 // CHECK14: omp.inner.for.inc61:
9963 // CHECK14-NEXT: [[TMP35:%.*]] = load i32, i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
9964 // CHECK14-NEXT: [[ADD62:%.*]] = add nsw i32 [[TMP35]], 1
9965 // CHECK14-NEXT: store i32 [[ADD62]], i32* [[DOTOMP_IV51]], align 4, !llvm.access.group !9
9966 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND53]], !llvm.loop [[LOOP10:![0-9]+]]
9967 // CHECK14: omp.inner.for.end63:
9968 // CHECK14-NEXT: [[TMP36:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
9969 // CHECK14-NEXT: [[SUB64:%.*]] = sub nsw i32 [[TMP36]], 0
9970 // CHECK14-NEXT: [[DIV65:%.*]] = sdiv i32 [[SUB64]], 1
9971 // CHECK14-NEXT: [[MUL66:%.*]] = mul nsw i32 [[DIV65]], 1
9972 // CHECK14-NEXT: [[ADD67:%.*]] = add nsw i32 0, [[MUL66]]
9973 // CHECK14-NEXT: store i32 [[ADD67]], i32* [[I52]], align 4
9974 // CHECK14-NEXT: br label [[SIMD_IF_END68]]
9975 // CHECK14: simd.if.end68:
9976 // CHECK14-NEXT: [[TMP37:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
9977 // CHECK14-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP37]])
9978 // CHECK14-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
9979 // CHECK14-NEXT: [[TMP38:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
9980 // CHECK14-NEXT: call void @llvm.stackrestore(i8* [[TMP38]])
9981 // CHECK14-NEXT: [[TMP39:%.*]] = load i32, i32* [[RETVAL]], align 4
9982 // CHECK14-NEXT: ret i32 [[TMP39]]
9983 //
9984 //
9985 // CHECK14-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
9986 // CHECK14-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
9987 // CHECK14-NEXT: entry:
9988 // CHECK14-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
9989 // CHECK14-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
9990 // CHECK14-NEXT: [[M:%.*]] = alloca i32, align 4
9991 // CHECK14-NEXT: [[TMP:%.*]] = alloca i32, align 4
9992 // CHECK14-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
9993 // CHECK14-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
9994 // CHECK14-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
9995 // CHECK14-NEXT: [[I:%.*]] = alloca i32, align 4
9996 // CHECK14-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
9997 // CHECK14-NEXT: [[DOTOMP_LB3:%.*]] = alloca i32, align 4
9998 // CHECK14-NEXT: [[DOTOMP_UB4:%.*]] = alloca i32, align 4
9999 // CHECK14-NEXT: [[DOTOMP_IV5:%.*]] = alloca i32, align 4
10000 // CHECK14-NEXT: [[I6:%.*]] = alloca i32, align 4
10001 // CHECK14-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10002 // CHECK14-NEXT: [[_TMP18:%.*]] = alloca i32, align 4
10003 // CHECK14-NEXT: [[DOTOMP_LB19:%.*]] = alloca i32, align 4
10004 // CHECK14-NEXT: [[DOTOMP_UB20:%.*]] = alloca i32, align 4
10005 // CHECK14-NEXT: [[DOTOMP_IV21:%.*]] = alloca i32, align 4
10006 // CHECK14-NEXT: [[I22:%.*]] = alloca i32, align 4
10007 // CHECK14-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
10008 // CHECK14-NEXT: store i32 10, i32* [[M]], align 4
10009 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
10010 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
10011 // CHECK14-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10012 // CHECK14-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
10013 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10014 // CHECK14: omp.inner.for.cond:
10015 // CHECK14-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
10016 // CHECK14-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
10017 // CHECK14-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
10018 // CHECK14-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10019 // CHECK14: omp.inner.for.body:
10020 // CHECK14-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
10021 // CHECK14-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
10022 // CHECK14-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10023 // CHECK14-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !12
10024 // CHECK14-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !12
10025 // CHECK14-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP4]] to i64
10026 // CHECK14-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
10027 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !12
10028 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
10029 // CHECK14: omp.body.continue:
10030 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10031 // CHECK14: omp.inner.for.inc:
10032 // CHECK14-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
10033 // CHECK14-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
10034 // CHECK14-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
10035 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
10036 // CHECK14: omp.inner.for.end:
10037 // CHECK14-NEXT: store i32 10, i32* [[I]], align 4
10038 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB3]], align 4
10039 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_UB4]], align 4
10040 // CHECK14-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
10041 // CHECK14-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV5]], align 4
10042 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND7:%.*]]
10043 // CHECK14: omp.inner.for.cond7:
10044 // CHECK14-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !15
10045 // CHECK14-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !15
10046 // CHECK14-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10047 // CHECK14-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END17:%.*]]
10048 // CHECK14: omp.inner.for.body9:
10049 // CHECK14-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !15
10050 // CHECK14-NEXT: [[MUL10:%.*]] = mul nsw i32 [[TMP9]], 1
10051 // CHECK14-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
10052 // CHECK14-NEXT: store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !15
10053 // CHECK14-NEXT: [[TMP10:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !15
10054 // CHECK14-NEXT: [[IDXPROM12:%.*]] = sext i32 [[TMP10]] to i64
10055 // CHECK14-NEXT: [[ARRAYIDX13:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM12]]
10056 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX13]], align 4, !llvm.access.group !15
10057 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE14:%.*]]
10058 // CHECK14: omp.body.continue14:
10059 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC15:%.*]]
10060 // CHECK14: omp.inner.for.inc15:
10061 // CHECK14-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !15
10062 // CHECK14-NEXT: [[ADD16:%.*]] = add nsw i32 [[TMP11]], 1
10063 // CHECK14-NEXT: store i32 [[ADD16]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !15
10064 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP16:![0-9]+]]
10065 // CHECK14: omp.inner.for.end17:
10066 // CHECK14-NEXT: store i32 10, i32* [[I6]], align 4
10067 // CHECK14-NEXT: [[TMP12:%.*]] = load i32, i32* [[M]], align 4
10068 // CHECK14-NEXT: store i32 [[TMP12]], i32* [[DOTCAPTURE_EXPR_]], align 4
10069 // CHECK14-NEXT: store i32 0, i32* [[DOTOMP_LB19]], align 4
10070 // CHECK14-NEXT: store i32 9, i32* [[DOTOMP_UB20]], align 4
10071 // CHECK14-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB19]], align 4
10072 // CHECK14-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV21]], align 4
10073 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND23:%.*]]
10074 // CHECK14: omp.inner.for.cond23:
10075 // CHECK14-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !18
10076 // CHECK14-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB20]], align 4, !llvm.access.group !18
10077 // CHECK14-NEXT: [[CMP24:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
10078 // CHECK14-NEXT: br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END33:%.*]]
10079 // CHECK14: omp.inner.for.body25:
10080 // CHECK14-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !18
10081 // CHECK14-NEXT: [[MUL26:%.*]] = mul nsw i32 [[TMP16]], 1
10082 // CHECK14-NEXT: [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
10083 // CHECK14-NEXT: store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !18
10084 // CHECK14-NEXT: [[TMP17:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !18
10085 // CHECK14-NEXT: [[IDXPROM28:%.*]] = sext i32 [[TMP17]] to i64
10086 // CHECK14-NEXT: [[ARRAYIDX29:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i64 0, i64 [[IDXPROM28]]
10087 // CHECK14-NEXT: store i32 0, i32* [[ARRAYIDX29]], align 4, !llvm.access.group !18
10088 // CHECK14-NEXT: br label [[OMP_BODY_CONTINUE30:%.*]]
10089 // CHECK14: omp.body.continue30:
10090 // CHECK14-NEXT: br label [[OMP_INNER_FOR_INC31:%.*]]
10091 // CHECK14: omp.inner.for.inc31:
10092 // CHECK14-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !18
10093 // CHECK14-NEXT: [[ADD32:%.*]] = add nsw i32 [[TMP18]], 1
10094 // CHECK14-NEXT: store i32 [[ADD32]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !18
10095 // CHECK14-NEXT: br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP19:![0-9]+]]
10096 // CHECK14: omp.inner.for.end33:
10097 // CHECK14-NEXT: store i32 10, i32* [[I22]], align 4
10098 // CHECK14-NEXT: ret i32 0
10099 //
10100 //
10101 // CHECK15-LABEL: define {{[^@]+}}@main
10102 // CHECK15-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
10103 // CHECK15-NEXT: entry:
10104 // CHECK15-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
10105 // CHECK15-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
10106 // CHECK15-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
10107 // CHECK15-NEXT: [[N:%.*]] = alloca i32, align 4
10108 // CHECK15-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
10109 // CHECK15-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
10110 // CHECK15-NEXT: [[M:%.*]] = alloca i32, align 4
10111 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
10112 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10113 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
10114 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
10115 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
10116 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
10117 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10118 // CHECK15-NEXT: [[I3:%.*]] = alloca i32, align 4
10119 // CHECK15-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
10120 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
10121 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
10122 // CHECK15-NEXT: [[DOTOMP_LB16:%.*]] = alloca i32, align 4
10123 // CHECK15-NEXT: [[DOTOMP_UB17:%.*]] = alloca i32, align 4
10124 // CHECK15-NEXT: [[I18:%.*]] = alloca i32, align 4
10125 // CHECK15-NEXT: [[DOTOMP_IV21:%.*]] = alloca i32, align 4
10126 // CHECK15-NEXT: [[I22:%.*]] = alloca i32, align 4
10127 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
10128 // CHECK15-NEXT: [[_TMP39:%.*]] = alloca i32, align 4
10129 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_40:%.*]] = alloca i32, align 4
10130 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
10131 // CHECK15-NEXT: [[DOTOMP_LB45:%.*]] = alloca i32, align 4
10132 // CHECK15-NEXT: [[DOTOMP_UB46:%.*]] = alloca i32, align 4
10133 // CHECK15-NEXT: [[I47:%.*]] = alloca i32, align 4
10134 // CHECK15-NEXT: [[DOTOMP_IV50:%.*]] = alloca i32, align 4
10135 // CHECK15-NEXT: [[I51:%.*]] = alloca i32, align 4
10136 // CHECK15-NEXT: store i32 0, i32* [[RETVAL]], align 4
10137 // CHECK15-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
10138 // CHECK15-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
10139 // CHECK15-NEXT: store i32 100, i32* [[N]], align 4
10140 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
10141 // CHECK15-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
10142 // CHECK15-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
10143 // CHECK15-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
10144 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
10145 // CHECK15-NEXT: store i32 10, i32* [[M]], align 4
10146 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
10147 // CHECK15-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
10148 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10149 // CHECK15-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
10150 // CHECK15-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
10151 // CHECK15-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
10152 // CHECK15-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
10153 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
10154 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10155 // CHECK15-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
10156 // CHECK15-NEXT: store i32 0, i32* [[I]], align 4
10157 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10158 // CHECK15-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
10159 // CHECK15-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
10160 // CHECK15: simd.if.then:
10161 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10162 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
10163 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10164 // CHECK15: omp.inner.for.cond:
10165 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
10166 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
10167 // CHECK15-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10168 // CHECK15-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10169 // CHECK15: omp.inner.for.body:
10170 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
10171 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
10172 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10173 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !3
10174 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !3
10175 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP10]]
10176 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
10177 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
10178 // CHECK15: omp.body.continue:
10179 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10180 // CHECK15: omp.inner.for.inc:
10181 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
10182 // CHECK15-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP11]], 1
10183 // CHECK15-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
10184 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
10185 // CHECK15: omp.inner.for.end:
10186 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10187 // CHECK15-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP12]], 0
10188 // CHECK15-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
10189 // CHECK15-NEXT: [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
10190 // CHECK15-NEXT: [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
10191 // CHECK15-NEXT: store i32 [[ADD9]], i32* [[I3]], align 4
10192 // CHECK15-NEXT: br label [[SIMD_IF_END]]
10193 // CHECK15: simd.if.end:
10194 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[N]], align 4
10195 // CHECK15-NEXT: store i32 [[TMP13]], i32* [[DOTCAPTURE_EXPR_11]], align 4
10196 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
10197 // CHECK15-NEXT: [[SUB13:%.*]] = sub nsw i32 [[TMP14]], 0
10198 // CHECK15-NEXT: [[DIV14:%.*]] = sdiv i32 [[SUB13]], 1
10199 // CHECK15-NEXT: [[SUB15:%.*]] = sub nsw i32 [[DIV14]], 1
10200 // CHECK15-NEXT: store i32 [[SUB15]], i32* [[DOTCAPTURE_EXPR_12]], align 4
10201 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB16]], align 4
10202 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
10203 // CHECK15-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_UB17]], align 4
10204 // CHECK15-NEXT: store i32 0, i32* [[I18]], align 4
10205 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
10206 // CHECK15-NEXT: [[CMP19:%.*]] = icmp slt i32 0, [[TMP16]]
10207 // CHECK15-NEXT: br i1 [[CMP19]], label [[SIMD_IF_THEN20:%.*]], label [[SIMD_IF_END37:%.*]]
10208 // CHECK15: simd.if.then20:
10209 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB16]], align 4
10210 // CHECK15-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV21]], align 4
10211 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND23:%.*]]
10212 // CHECK15: omp.inner.for.cond23:
10213 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
10214 // CHECK15-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB17]], align 4, !llvm.access.group !7
10215 // CHECK15-NEXT: [[CMP24:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
10216 // CHECK15-NEXT: br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END32:%.*]]
10217 // CHECK15: omp.inner.for.body25:
10218 // CHECK15-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
10219 // CHECK15-NEXT: [[MUL26:%.*]] = mul nsw i32 [[TMP20]], 1
10220 // CHECK15-NEXT: [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
10221 // CHECK15-NEXT: store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !7
10222 // CHECK15-NEXT: [[TMP21:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !7
10223 // CHECK15-NEXT: [[ARRAYIDX28:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP21]]
10224 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX28]], align 4, !llvm.access.group !7
10225 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE29:%.*]]
10226 // CHECK15: omp.body.continue29:
10227 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC30:%.*]]
10228 // CHECK15: omp.inner.for.inc30:
10229 // CHECK15-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
10230 // CHECK15-NEXT: [[ADD31:%.*]] = add nsw i32 [[TMP22]], 1
10231 // CHECK15-NEXT: store i32 [[ADD31]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
10232 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP8:![0-9]+]]
10233 // CHECK15: omp.inner.for.end32:
10234 // CHECK15-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
10235 // CHECK15-NEXT: [[SUB33:%.*]] = sub nsw i32 [[TMP23]], 0
10236 // CHECK15-NEXT: [[DIV34:%.*]] = sdiv i32 [[SUB33]], 1
10237 // CHECK15-NEXT: [[MUL35:%.*]] = mul nsw i32 [[DIV34]], 1
10238 // CHECK15-NEXT: [[ADD36:%.*]] = add nsw i32 0, [[MUL35]]
10239 // CHECK15-NEXT: store i32 [[ADD36]], i32* [[I22]], align 4
10240 // CHECK15-NEXT: br label [[SIMD_IF_END37]]
10241 // CHECK15: simd.if.end37:
10242 // CHECK15-NEXT: [[TMP24:%.*]] = load i32, i32* [[M]], align 4
10243 // CHECK15-NEXT: store i32 [[TMP24]], i32* [[DOTCAPTURE_EXPR_38]], align 4
10244 // CHECK15-NEXT: [[TMP25:%.*]] = load i32, i32* [[N]], align 4
10245 // CHECK15-NEXT: store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_40]], align 4
10246 // CHECK15-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
10247 // CHECK15-NEXT: [[SUB42:%.*]] = sub nsw i32 [[TMP26]], 0
10248 // CHECK15-NEXT: [[DIV43:%.*]] = sdiv i32 [[SUB42]], 1
10249 // CHECK15-NEXT: [[SUB44:%.*]] = sub nsw i32 [[DIV43]], 1
10250 // CHECK15-NEXT: store i32 [[SUB44]], i32* [[DOTCAPTURE_EXPR_41]], align 4
10251 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB45]], align 4
10252 // CHECK15-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
10253 // CHECK15-NEXT: store i32 [[TMP27]], i32* [[DOTOMP_UB46]], align 4
10254 // CHECK15-NEXT: store i32 0, i32* [[I47]], align 4
10255 // CHECK15-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
10256 // CHECK15-NEXT: [[CMP48:%.*]] = icmp slt i32 0, [[TMP28]]
10257 // CHECK15-NEXT: br i1 [[CMP48]], label [[SIMD_IF_THEN49:%.*]], label [[SIMD_IF_END66:%.*]]
10258 // CHECK15: simd.if.then49:
10259 // CHECK15-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTOMP_LB45]], align 4
10260 // CHECK15-NEXT: store i32 [[TMP29]], i32* [[DOTOMP_IV50]], align 4
10261 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND52:%.*]]
10262 // CHECK15: omp.inner.for.cond52:
10263 // CHECK15-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
10264 // CHECK15-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_UB46]], align 4, !llvm.access.group !10
10265 // CHECK15-NEXT: [[CMP53:%.*]] = icmp sle i32 [[TMP30]], [[TMP31]]
10266 // CHECK15-NEXT: br i1 [[CMP53]], label [[OMP_INNER_FOR_BODY54:%.*]], label [[OMP_INNER_FOR_END61:%.*]]
10267 // CHECK15: omp.inner.for.body54:
10268 // CHECK15-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
10269 // CHECK15-NEXT: [[MUL55:%.*]] = mul nsw i32 [[TMP32]], 1
10270 // CHECK15-NEXT: [[ADD56:%.*]] = add nsw i32 0, [[MUL55]]
10271 // CHECK15-NEXT: store i32 [[ADD56]], i32* [[I51]], align 4, !llvm.access.group !10
10272 // CHECK15-NEXT: [[TMP33:%.*]] = load i32, i32* [[I51]], align 4, !llvm.access.group !10
10273 // CHECK15-NEXT: [[ARRAYIDX57:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP33]]
10274 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX57]], align 4, !llvm.access.group !10
10275 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE58:%.*]]
10276 // CHECK15: omp.body.continue58:
10277 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC59:%.*]]
10278 // CHECK15: omp.inner.for.inc59:
10279 // CHECK15-NEXT: [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
10280 // CHECK15-NEXT: [[ADD60:%.*]] = add nsw i32 [[TMP34]], 1
10281 // CHECK15-NEXT: store i32 [[ADD60]], i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
10282 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND52]], !llvm.loop [[LOOP11:![0-9]+]]
10283 // CHECK15: omp.inner.for.end61:
10284 // CHECK15-NEXT: [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
10285 // CHECK15-NEXT: [[SUB62:%.*]] = sub nsw i32 [[TMP35]], 0
10286 // CHECK15-NEXT: [[DIV63:%.*]] = sdiv i32 [[SUB62]], 1
10287 // CHECK15-NEXT: [[MUL64:%.*]] = mul nsw i32 [[DIV63]], 1
10288 // CHECK15-NEXT: [[ADD65:%.*]] = add nsw i32 0, [[MUL64]]
10289 // CHECK15-NEXT: store i32 [[ADD65]], i32* [[I51]], align 4
10290 // CHECK15-NEXT: br label [[SIMD_IF_END66]]
10291 // CHECK15: simd.if.end66:
10292 // CHECK15-NEXT: [[TMP36:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
10293 // CHECK15-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP36]])
10294 // CHECK15-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
10295 // CHECK15-NEXT: [[TMP37:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
10296 // CHECK15-NEXT: call void @llvm.stackrestore(i8* [[TMP37]])
10297 // CHECK15-NEXT: [[TMP38:%.*]] = load i32, i32* [[RETVAL]], align 4
10298 // CHECK15-NEXT: ret i32 [[TMP38]]
10299 //
10300 //
10301 // CHECK15-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
10302 // CHECK15-SAME: (i32 [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
10303 // CHECK15-NEXT: entry:
10304 // CHECK15-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
10305 // CHECK15-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
10306 // CHECK15-NEXT: [[M:%.*]] = alloca i32, align 4
10307 // CHECK15-NEXT: [[TMP:%.*]] = alloca i32, align 4
10308 // CHECK15-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
10309 // CHECK15-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
10310 // CHECK15-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10311 // CHECK15-NEXT: [[I:%.*]] = alloca i32, align 4
10312 // CHECK15-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
10313 // CHECK15-NEXT: [[DOTOMP_LB3:%.*]] = alloca i32, align 4
10314 // CHECK15-NEXT: [[DOTOMP_UB4:%.*]] = alloca i32, align 4
10315 // CHECK15-NEXT: [[DOTOMP_IV5:%.*]] = alloca i32, align 4
10316 // CHECK15-NEXT: [[I6:%.*]] = alloca i32, align 4
10317 // CHECK15-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10318 // CHECK15-NEXT: [[_TMP17:%.*]] = alloca i32, align 4
10319 // CHECK15-NEXT: [[DOTOMP_LB18:%.*]] = alloca i32, align 4
10320 // CHECK15-NEXT: [[DOTOMP_UB19:%.*]] = alloca i32, align 4
10321 // CHECK15-NEXT: [[DOTOMP_IV20:%.*]] = alloca i32, align 4
10322 // CHECK15-NEXT: [[I21:%.*]] = alloca i32, align 4
10323 // CHECK15-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
10324 // CHECK15-NEXT: store i32 10, i32* [[M]], align 4
10325 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
10326 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
10327 // CHECK15-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10328 // CHECK15-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
10329 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10330 // CHECK15: omp.inner.for.cond:
10331 // CHECK15-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
10332 // CHECK15-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
10333 // CHECK15-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
10334 // CHECK15-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10335 // CHECK15: omp.inner.for.body:
10336 // CHECK15-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
10337 // CHECK15-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
10338 // CHECK15-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10339 // CHECK15-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
10340 // CHECK15-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
10341 // CHECK15-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP4]]
10342 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
10343 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
10344 // CHECK15: omp.body.continue:
10345 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10346 // CHECK15: omp.inner.for.inc:
10347 // CHECK15-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
10348 // CHECK15-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
10349 // CHECK15-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
10350 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
10351 // CHECK15: omp.inner.for.end:
10352 // CHECK15-NEXT: store i32 10, i32* [[I]], align 4
10353 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB3]], align 4
10354 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_UB4]], align 4
10355 // CHECK15-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
10356 // CHECK15-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV5]], align 4
10357 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND7:%.*]]
10358 // CHECK15: omp.inner.for.cond7:
10359 // CHECK15-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
10360 // CHECK15-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !16
10361 // CHECK15-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10362 // CHECK15-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END16:%.*]]
10363 // CHECK15: omp.inner.for.body9:
10364 // CHECK15-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
10365 // CHECK15-NEXT: [[MUL10:%.*]] = mul nsw i32 [[TMP9]], 1
10366 // CHECK15-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
10367 // CHECK15-NEXT: store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !16
10368 // CHECK15-NEXT: [[TMP10:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !16
10369 // CHECK15-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP10]]
10370 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX12]], align 4, !llvm.access.group !16
10371 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE13:%.*]]
10372 // CHECK15: omp.body.continue13:
10373 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC14:%.*]]
10374 // CHECK15: omp.inner.for.inc14:
10375 // CHECK15-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
10376 // CHECK15-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP11]], 1
10377 // CHECK15-NEXT: store i32 [[ADD15]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
10378 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP17:![0-9]+]]
10379 // CHECK15: omp.inner.for.end16:
10380 // CHECK15-NEXT: store i32 10, i32* [[I6]], align 4
10381 // CHECK15-NEXT: [[TMP12:%.*]] = load i32, i32* [[M]], align 4
10382 // CHECK15-NEXT: store i32 [[TMP12]], i32* [[DOTCAPTURE_EXPR_]], align 4
10383 // CHECK15-NEXT: store i32 0, i32* [[DOTOMP_LB18]], align 4
10384 // CHECK15-NEXT: store i32 9, i32* [[DOTOMP_UB19]], align 4
10385 // CHECK15-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB18]], align 4
10386 // CHECK15-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV20]], align 4
10387 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND22:%.*]]
10388 // CHECK15: omp.inner.for.cond22:
10389 // CHECK15-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !19
10390 // CHECK15-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB19]], align 4, !llvm.access.group !19
10391 // CHECK15-NEXT: [[CMP23:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
10392 // CHECK15-NEXT: br i1 [[CMP23]], label [[OMP_INNER_FOR_BODY24:%.*]], label [[OMP_INNER_FOR_END31:%.*]]
10393 // CHECK15: omp.inner.for.body24:
10394 // CHECK15-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !19
10395 // CHECK15-NEXT: [[MUL25:%.*]] = mul nsw i32 [[TMP16]], 1
10396 // CHECK15-NEXT: [[ADD26:%.*]] = add nsw i32 0, [[MUL25]]
10397 // CHECK15-NEXT: store i32 [[ADD26]], i32* [[I21]], align 4, !llvm.access.group !19
10398 // CHECK15-NEXT: [[TMP17:%.*]] = load i32, i32* [[I21]], align 4, !llvm.access.group !19
10399 // CHECK15-NEXT: [[ARRAYIDX27:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP17]]
10400 // CHECK15-NEXT: store i32 0, i32* [[ARRAYIDX27]], align 4, !llvm.access.group !19
10401 // CHECK15-NEXT: br label [[OMP_BODY_CONTINUE28:%.*]]
10402 // CHECK15: omp.body.continue28:
10403 // CHECK15-NEXT: br label [[OMP_INNER_FOR_INC29:%.*]]
10404 // CHECK15: omp.inner.for.inc29:
10405 // CHECK15-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !19
10406 // CHECK15-NEXT: [[ADD30:%.*]] = add nsw i32 [[TMP18]], 1
10407 // CHECK15-NEXT: store i32 [[ADD30]], i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !19
10408 // CHECK15-NEXT: br label [[OMP_INNER_FOR_COND22]], !llvm.loop [[LOOP20:![0-9]+]]
10409 // CHECK15: omp.inner.for.end31:
10410 // CHECK15-NEXT: store i32 10, i32* [[I21]], align 4
10411 // CHECK15-NEXT: ret i32 0
10412 //
10413 //
10414 // CHECK16-LABEL: define {{[^@]+}}@main
10415 // CHECK16-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
10416 // CHECK16-NEXT: entry:
10417 // CHECK16-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
10418 // CHECK16-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
10419 // CHECK16-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
10420 // CHECK16-NEXT: [[N:%.*]] = alloca i32, align 4
10421 // CHECK16-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
10422 // CHECK16-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
10423 // CHECK16-NEXT: [[M:%.*]] = alloca i32, align 4
10424 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
10425 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10426 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
10427 // CHECK16-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
10428 // CHECK16-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
10429 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
10430 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10431 // CHECK16-NEXT: [[I3:%.*]] = alloca i32, align 4
10432 // CHECK16-NEXT: [[_TMP10:%.*]] = alloca i32, align 4
10433 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
10434 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
10435 // CHECK16-NEXT: [[DOTOMP_LB16:%.*]] = alloca i32, align 4
10436 // CHECK16-NEXT: [[DOTOMP_UB17:%.*]] = alloca i32, align 4
10437 // CHECK16-NEXT: [[I18:%.*]] = alloca i32, align 4
10438 // CHECK16-NEXT: [[DOTOMP_IV21:%.*]] = alloca i32, align 4
10439 // CHECK16-NEXT: [[I22:%.*]] = alloca i32, align 4
10440 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_38:%.*]] = alloca i32, align 4
10441 // CHECK16-NEXT: [[_TMP39:%.*]] = alloca i32, align 4
10442 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_40:%.*]] = alloca i32, align 4
10443 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_41:%.*]] = alloca i32, align 4
10444 // CHECK16-NEXT: [[DOTOMP_LB45:%.*]] = alloca i32, align 4
10445 // CHECK16-NEXT: [[DOTOMP_UB46:%.*]] = alloca i32, align 4
10446 // CHECK16-NEXT: [[I47:%.*]] = alloca i32, align 4
10447 // CHECK16-NEXT: [[DOTOMP_IV50:%.*]] = alloca i32, align 4
10448 // CHECK16-NEXT: [[I51:%.*]] = alloca i32, align 4
10449 // CHECK16-NEXT: store i32 0, i32* [[RETVAL]], align 4
10450 // CHECK16-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
10451 // CHECK16-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
10452 // CHECK16-NEXT: store i32 100, i32* [[N]], align 4
10453 // CHECK16-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
10454 // CHECK16-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
10455 // CHECK16-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
10456 // CHECK16-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
10457 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
10458 // CHECK16-NEXT: store i32 10, i32* [[M]], align 4
10459 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
10460 // CHECK16-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
10461 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10462 // CHECK16-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
10463 // CHECK16-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
10464 // CHECK16-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
10465 // CHECK16-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
10466 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
10467 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
10468 // CHECK16-NEXT: store i32 [[TMP4]], i32* [[DOTOMP_UB]], align 4
10469 // CHECK16-NEXT: store i32 0, i32* [[I]], align 4
10470 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10471 // CHECK16-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
10472 // CHECK16-NEXT: br i1 [[CMP]], label [[SIMD_IF_THEN:%.*]], label [[SIMD_IF_END:%.*]]
10473 // CHECK16: simd.if.then:
10474 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10475 // CHECK16-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
10476 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10477 // CHECK16: omp.inner.for.cond:
10478 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
10479 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !3
10480 // CHECK16-NEXT: [[CMP4:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10481 // CHECK16-NEXT: br i1 [[CMP4]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10482 // CHECK16: omp.inner.for.body:
10483 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
10484 // CHECK16-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
10485 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10486 // CHECK16-NEXT: store i32 [[ADD]], i32* [[I3]], align 4, !llvm.access.group !3
10487 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[I3]], align 4, !llvm.access.group !3
10488 // CHECK16-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP10]]
10489 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !3
10490 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
10491 // CHECK16: omp.body.continue:
10492 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10493 // CHECK16: omp.inner.for.inc:
10494 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
10495 // CHECK16-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP11]], 1
10496 // CHECK16-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !3
10497 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP4:![0-9]+]]
10498 // CHECK16: omp.inner.for.end:
10499 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
10500 // CHECK16-NEXT: [[SUB6:%.*]] = sub nsw i32 [[TMP12]], 0
10501 // CHECK16-NEXT: [[DIV7:%.*]] = sdiv i32 [[SUB6]], 1
10502 // CHECK16-NEXT: [[MUL8:%.*]] = mul nsw i32 [[DIV7]], 1
10503 // CHECK16-NEXT: [[ADD9:%.*]] = add nsw i32 0, [[MUL8]]
10504 // CHECK16-NEXT: store i32 [[ADD9]], i32* [[I3]], align 4
10505 // CHECK16-NEXT: br label [[SIMD_IF_END]]
10506 // CHECK16: simd.if.end:
10507 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[N]], align 4
10508 // CHECK16-NEXT: store i32 [[TMP13]], i32* [[DOTCAPTURE_EXPR_11]], align 4
10509 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
10510 // CHECK16-NEXT: [[SUB13:%.*]] = sub nsw i32 [[TMP14]], 0
10511 // CHECK16-NEXT: [[DIV14:%.*]] = sdiv i32 [[SUB13]], 1
10512 // CHECK16-NEXT: [[SUB15:%.*]] = sub nsw i32 [[DIV14]], 1
10513 // CHECK16-NEXT: store i32 [[SUB15]], i32* [[DOTCAPTURE_EXPR_12]], align 4
10514 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB16]], align 4
10515 // CHECK16-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
10516 // CHECK16-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_UB17]], align 4
10517 // CHECK16-NEXT: store i32 0, i32* [[I18]], align 4
10518 // CHECK16-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
10519 // CHECK16-NEXT: [[CMP19:%.*]] = icmp slt i32 0, [[TMP16]]
10520 // CHECK16-NEXT: br i1 [[CMP19]], label [[SIMD_IF_THEN20:%.*]], label [[SIMD_IF_END37:%.*]]
10521 // CHECK16: simd.if.then20:
10522 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_LB16]], align 4
10523 // CHECK16-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV21]], align 4
10524 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND23:%.*]]
10525 // CHECK16: omp.inner.for.cond23:
10526 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
10527 // CHECK16-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_UB17]], align 4, !llvm.access.group !7
10528 // CHECK16-NEXT: [[CMP24:%.*]] = icmp sle i32 [[TMP18]], [[TMP19]]
10529 // CHECK16-NEXT: br i1 [[CMP24]], label [[OMP_INNER_FOR_BODY25:%.*]], label [[OMP_INNER_FOR_END32:%.*]]
10530 // CHECK16: omp.inner.for.body25:
10531 // CHECK16-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
10532 // CHECK16-NEXT: [[MUL26:%.*]] = mul nsw i32 [[TMP20]], 1
10533 // CHECK16-NEXT: [[ADD27:%.*]] = add nsw i32 0, [[MUL26]]
10534 // CHECK16-NEXT: store i32 [[ADD27]], i32* [[I22]], align 4, !llvm.access.group !7
10535 // CHECK16-NEXT: [[TMP21:%.*]] = load i32, i32* [[I22]], align 4, !llvm.access.group !7
10536 // CHECK16-NEXT: [[ARRAYIDX28:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP21]]
10537 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX28]], align 4, !llvm.access.group !7
10538 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE29:%.*]]
10539 // CHECK16: omp.body.continue29:
10540 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC30:%.*]]
10541 // CHECK16: omp.inner.for.inc30:
10542 // CHECK16-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
10543 // CHECK16-NEXT: [[ADD31:%.*]] = add nsw i32 [[TMP22]], 1
10544 // CHECK16-NEXT: store i32 [[ADD31]], i32* [[DOTOMP_IV21]], align 4, !llvm.access.group !7
10545 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND23]], !llvm.loop [[LOOP8:![0-9]+]]
10546 // CHECK16: omp.inner.for.end32:
10547 // CHECK16-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
10548 // CHECK16-NEXT: [[SUB33:%.*]] = sub nsw i32 [[TMP23]], 0
10549 // CHECK16-NEXT: [[DIV34:%.*]] = sdiv i32 [[SUB33]], 1
10550 // CHECK16-NEXT: [[MUL35:%.*]] = mul nsw i32 [[DIV34]], 1
10551 // CHECK16-NEXT: [[ADD36:%.*]] = add nsw i32 0, [[MUL35]]
10552 // CHECK16-NEXT: store i32 [[ADD36]], i32* [[I22]], align 4
10553 // CHECK16-NEXT: br label [[SIMD_IF_END37]]
10554 // CHECK16: simd.if.end37:
10555 // CHECK16-NEXT: [[TMP24:%.*]] = load i32, i32* [[M]], align 4
10556 // CHECK16-NEXT: store i32 [[TMP24]], i32* [[DOTCAPTURE_EXPR_38]], align 4
10557 // CHECK16-NEXT: [[TMP25:%.*]] = load i32, i32* [[N]], align 4
10558 // CHECK16-NEXT: store i32 [[TMP25]], i32* [[DOTCAPTURE_EXPR_40]], align 4
10559 // CHECK16-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
10560 // CHECK16-NEXT: [[SUB42:%.*]] = sub nsw i32 [[TMP26]], 0
10561 // CHECK16-NEXT: [[DIV43:%.*]] = sdiv i32 [[SUB42]], 1
10562 // CHECK16-NEXT: [[SUB44:%.*]] = sub nsw i32 [[DIV43]], 1
10563 // CHECK16-NEXT: store i32 [[SUB44]], i32* [[DOTCAPTURE_EXPR_41]], align 4
10564 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB45]], align 4
10565 // CHECK16-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_41]], align 4
10566 // CHECK16-NEXT: store i32 [[TMP27]], i32* [[DOTOMP_UB46]], align 4
10567 // CHECK16-NEXT: store i32 0, i32* [[I47]], align 4
10568 // CHECK16-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
10569 // CHECK16-NEXT: [[CMP48:%.*]] = icmp slt i32 0, [[TMP28]]
10570 // CHECK16-NEXT: br i1 [[CMP48]], label [[SIMD_IF_THEN49:%.*]], label [[SIMD_IF_END66:%.*]]
10571 // CHECK16: simd.if.then49:
10572 // CHECK16-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTOMP_LB45]], align 4
10573 // CHECK16-NEXT: store i32 [[TMP29]], i32* [[DOTOMP_IV50]], align 4
10574 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND52:%.*]]
10575 // CHECK16: omp.inner.for.cond52:
10576 // CHECK16-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
10577 // CHECK16-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_UB46]], align 4, !llvm.access.group !10
10578 // CHECK16-NEXT: [[CMP53:%.*]] = icmp sle i32 [[TMP30]], [[TMP31]]
10579 // CHECK16-NEXT: br i1 [[CMP53]], label [[OMP_INNER_FOR_BODY54:%.*]], label [[OMP_INNER_FOR_END61:%.*]]
10580 // CHECK16: omp.inner.for.body54:
10581 // CHECK16-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
10582 // CHECK16-NEXT: [[MUL55:%.*]] = mul nsw i32 [[TMP32]], 1
10583 // CHECK16-NEXT: [[ADD56:%.*]] = add nsw i32 0, [[MUL55]]
10584 // CHECK16-NEXT: store i32 [[ADD56]], i32* [[I51]], align 4, !llvm.access.group !10
10585 // CHECK16-NEXT: [[TMP33:%.*]] = load i32, i32* [[I51]], align 4, !llvm.access.group !10
10586 // CHECK16-NEXT: [[ARRAYIDX57:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 [[TMP33]]
10587 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX57]], align 4, !llvm.access.group !10
10588 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE58:%.*]]
10589 // CHECK16: omp.body.continue58:
10590 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC59:%.*]]
10591 // CHECK16: omp.inner.for.inc59:
10592 // CHECK16-NEXT: [[TMP34:%.*]] = load i32, i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
10593 // CHECK16-NEXT: [[ADD60:%.*]] = add nsw i32 [[TMP34]], 1
10594 // CHECK16-NEXT: store i32 [[ADD60]], i32* [[DOTOMP_IV50]], align 4, !llvm.access.group !10
10595 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND52]], !llvm.loop [[LOOP11:![0-9]+]]
10596 // CHECK16: omp.inner.for.end61:
10597 // CHECK16-NEXT: [[TMP35:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_40]], align 4
10598 // CHECK16-NEXT: [[SUB62:%.*]] = sub nsw i32 [[TMP35]], 0
10599 // CHECK16-NEXT: [[DIV63:%.*]] = sdiv i32 [[SUB62]], 1
10600 // CHECK16-NEXT: [[MUL64:%.*]] = mul nsw i32 [[DIV63]], 1
10601 // CHECK16-NEXT: [[ADD65:%.*]] = add nsw i32 0, [[MUL64]]
10602 // CHECK16-NEXT: store i32 [[ADD65]], i32* [[I51]], align 4
10603 // CHECK16-NEXT: br label [[SIMD_IF_END66]]
10604 // CHECK16: simd.if.end66:
10605 // CHECK16-NEXT: [[TMP36:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
10606 // CHECK16-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP36]])
10607 // CHECK16-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
10608 // CHECK16-NEXT: [[TMP37:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
10609 // CHECK16-NEXT: call void @llvm.stackrestore(i8* [[TMP37]])
10610 // CHECK16-NEXT: [[TMP38:%.*]] = load i32, i32* [[RETVAL]], align 4
10611 // CHECK16-NEXT: ret i32 [[TMP38]]
10612 //
10613 //
10614 // CHECK16-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
10615 // CHECK16-SAME: (i32 [[ARGC:%.*]]) #[[ATTR2:[0-9]+]] comdat {
10616 // CHECK16-NEXT: entry:
10617 // CHECK16-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
10618 // CHECK16-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
10619 // CHECK16-NEXT: [[M:%.*]] = alloca i32, align 4
10620 // CHECK16-NEXT: [[TMP:%.*]] = alloca i32, align 4
10621 // CHECK16-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
10622 // CHECK16-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
10623 // CHECK16-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
10624 // CHECK16-NEXT: [[I:%.*]] = alloca i32, align 4
10625 // CHECK16-NEXT: [[_TMP2:%.*]] = alloca i32, align 4
10626 // CHECK16-NEXT: [[DOTOMP_LB3:%.*]] = alloca i32, align 4
10627 // CHECK16-NEXT: [[DOTOMP_UB4:%.*]] = alloca i32, align 4
10628 // CHECK16-NEXT: [[DOTOMP_IV5:%.*]] = alloca i32, align 4
10629 // CHECK16-NEXT: [[I6:%.*]] = alloca i32, align 4
10630 // CHECK16-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
10631 // CHECK16-NEXT: [[_TMP17:%.*]] = alloca i32, align 4
10632 // CHECK16-NEXT: [[DOTOMP_LB18:%.*]] = alloca i32, align 4
10633 // CHECK16-NEXT: [[DOTOMP_UB19:%.*]] = alloca i32, align 4
10634 // CHECK16-NEXT: [[DOTOMP_IV20:%.*]] = alloca i32, align 4
10635 // CHECK16-NEXT: [[I21:%.*]] = alloca i32, align 4
10636 // CHECK16-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
10637 // CHECK16-NEXT: store i32 10, i32* [[M]], align 4
10638 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
10639 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
10640 // CHECK16-NEXT: [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10641 // CHECK16-NEXT: store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
10642 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
10643 // CHECK16: omp.inner.for.cond:
10644 // CHECK16-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
10645 // CHECK16-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
10646 // CHECK16-NEXT: [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
10647 // CHECK16-NEXT: br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10648 // CHECK16: omp.inner.for.body:
10649 // CHECK16-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
10650 // CHECK16-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
10651 // CHECK16-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10652 // CHECK16-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
10653 // CHECK16-NEXT: [[TMP4:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !13
10654 // CHECK16-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP4]]
10655 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !13
10656 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
10657 // CHECK16: omp.body.continue:
10658 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
10659 // CHECK16: omp.inner.for.inc:
10660 // CHECK16-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
10661 // CHECK16-NEXT: [[ADD1:%.*]] = add nsw i32 [[TMP5]], 1
10662 // CHECK16-NEXT: store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
10663 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
10664 // CHECK16: omp.inner.for.end:
10665 // CHECK16-NEXT: store i32 10, i32* [[I]], align 4
10666 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB3]], align 4
10667 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_UB4]], align 4
10668 // CHECK16-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
10669 // CHECK16-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV5]], align 4
10670 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND7:%.*]]
10671 // CHECK16: omp.inner.for.cond7:
10672 // CHECK16-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
10673 // CHECK16-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !16
10674 // CHECK16-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10675 // CHECK16-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END16:%.*]]
10676 // CHECK16: omp.inner.for.body9:
10677 // CHECK16-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
10678 // CHECK16-NEXT: [[MUL10:%.*]] = mul nsw i32 [[TMP9]], 1
10679 // CHECK16-NEXT: [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
10680 // CHECK16-NEXT: store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !16
10681 // CHECK16-NEXT: [[TMP10:%.*]] = load i32, i32* [[I6]], align 4, !llvm.access.group !16
10682 // CHECK16-NEXT: [[ARRAYIDX12:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP10]]
10683 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX12]], align 4, !llvm.access.group !16
10684 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE13:%.*]]
10685 // CHECK16: omp.body.continue13:
10686 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC14:%.*]]
10687 // CHECK16: omp.inner.for.inc14:
10688 // CHECK16-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
10689 // CHECK16-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP11]], 1
10690 // CHECK16-NEXT: store i32 [[ADD15]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
10691 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP17:![0-9]+]]
10692 // CHECK16: omp.inner.for.end16:
10693 // CHECK16-NEXT: store i32 10, i32* [[I6]], align 4
10694 // CHECK16-NEXT: [[TMP12:%.*]] = load i32, i32* [[M]], align 4
10695 // CHECK16-NEXT: store i32 [[TMP12]], i32* [[DOTCAPTURE_EXPR_]], align 4
10696 // CHECK16-NEXT: store i32 0, i32* [[DOTOMP_LB18]], align 4
10697 // CHECK16-NEXT: store i32 9, i32* [[DOTOMP_UB19]], align 4
10698 // CHECK16-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB18]], align 4
10699 // CHECK16-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV20]], align 4
10700 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND22:%.*]]
10701 // CHECK16: omp.inner.for.cond22:
10702 // CHECK16-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !19
10703 // CHECK16-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB19]], align 4, !llvm.access.group !19
10704 // CHECK16-NEXT: [[CMP23:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
10705 // CHECK16-NEXT: br i1 [[CMP23]], label [[OMP_INNER_FOR_BODY24:%.*]], label [[OMP_INNER_FOR_END31:%.*]]
10706 // CHECK16: omp.inner.for.body24:
10707 // CHECK16-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !19
10708 // CHECK16-NEXT: [[MUL25:%.*]] = mul nsw i32 [[TMP16]], 1
10709 // CHECK16-NEXT: [[ADD26:%.*]] = add nsw i32 0, [[MUL25]]
10710 // CHECK16-NEXT: store i32 [[ADD26]], i32* [[I21]], align 4, !llvm.access.group !19
10711 // CHECK16-NEXT: [[TMP17:%.*]] = load i32, i32* [[I21]], align 4, !llvm.access.group !19
10712 // CHECK16-NEXT: [[ARRAYIDX27:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[A]], i32 0, i32 [[TMP17]]
10713 // CHECK16-NEXT: store i32 0, i32* [[ARRAYIDX27]], align 4, !llvm.access.group !19
10714 // CHECK16-NEXT: br label [[OMP_BODY_CONTINUE28:%.*]]
10715 // CHECK16: omp.body.continue28:
10716 // CHECK16-NEXT: br label [[OMP_INNER_FOR_INC29:%.*]]
10717 // CHECK16: omp.inner.for.inc29:
10718 // CHECK16-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !19
10719 // CHECK16-NEXT: [[ADD30:%.*]] = add nsw i32 [[TMP18]], 1
10720 // CHECK16-NEXT: store i32 [[ADD30]], i32* [[DOTOMP_IV20]], align 4, !llvm.access.group !19
10721 // CHECK16-NEXT: br label [[OMP_INNER_FOR_COND22]], !llvm.loop [[LOOP20:![0-9]+]]
10722 // CHECK16: omp.inner.for.end31:
10723 // CHECK16-NEXT: store i32 10, i32* [[I21]], align 4
10724 // CHECK16-NEXT: ret i32 0
10725 //
10726