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 // Test host codegen.
6 // 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
7 // 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
8 // 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
9 // 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
10 // 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
11 // 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
12
13 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
14 // 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
15 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
16 // RUN: %clang_cc1 -DCK1 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
17 // 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
18 // RUN: %clang_cc1 -DCK1 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
19 #ifdef CK1
20
21 int a[100];
22
teams_argument_global(int n)23 int teams_argument_global(int n) {
24 int te = n / 128;
25 int th = 128;
26 // discard n_addr
27
28
29 #pragma omp target
30 #pragma omp teams distribute num_teams(te), thread_limit(th)
31 for(int i = 0; i < n; i++) {
32 a[i] = 0;
33 }
34
35 #pragma omp target
36 {{{
37 #pragma omp teams distribute
38 for(int i = 0; i < n; i++) {
39 a[i] = 0;
40 }
41 }}}
42
43 // outlined target regions
44
45
46
47
48 return a[0];
49 }
50
51 #endif // CK1
52
53 // Test host codegen.
54 // 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
55 // 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
56 // 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
57 // 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
58 // 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
59 // 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
60
61 // RUN: %clang_cc1 -DCK2 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
62 // 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
63 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
64 // RUN: %clang_cc1 -DCK2 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
65 // 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
66 // RUN: %clang_cc1 -DCK2 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
67 #ifdef CK2
68
teams_local_arg(void)69 int teams_local_arg(void) {
70 int n = 100;
71 int a[n];
72
73 #pragma omp target
74 #pragma omp teams distribute
75 for(int i = 0; i < n; i++) {
76 a[i] = 0;
77 }
78
79 // outlined target region
80
81
82 return a[0];
83 }
84 #endif // CK2
85
86 // Test host codegen.
87 // RUN: %clang_cc1 -DCK3 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK17
88 // RUN: %clang_cc1 -DCK3 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
89 // RUN: %clang_cc1 -DCK3 -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=CHECK18
90 // RUN: %clang_cc1 -DCK3 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK19
91 // RUN: %clang_cc1 -DCK3 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
92 // RUN: %clang_cc1 -DCK3 -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=CHECK20
93
94 // RUN: %clang_cc1 -DCK3 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
95 // RUN: %clang_cc1 -DCK3 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
96 // RUN: %clang_cc1 -DCK3 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
97 // RUN: %clang_cc1 -DCK3 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
98 // RUN: %clang_cc1 -DCK3 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
99 // RUN: %clang_cc1 -DCK3 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
100 #ifdef CK3
101
102
103 template <typename T, int X, long long Y>
104 struct SS{
105 T a[X];
106 float b;
fooSS107 int foo(void) {
108
109 #pragma omp target
110 #pragma omp teams distribute
111 for(int i = 0; i < X; i++) {
112 a[i] = (T)0;
113 }
114
115 // outlined target region
116
117
118 return a[0];
119 }
120 };
121
teams_template_struct(void)122 int teams_template_struct(void) {
123 SS<int, 123, 456> V;
124 return V.foo();
125
126 }
127 #endif // CK3
128
129 // Test host codegen.
130 // RUN: %clang_cc1 -DCK4 -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK25
131 // RUN: %clang_cc1 -DCK4 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
132 // RUN: %clang_cc1 -DCK4 -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=CHECK26
133 // RUN: %clang_cc1 -DCK4 -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK27
134 // RUN: %clang_cc1 -DCK4 -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
135 // RUN: %clang_cc1 -DCK4 -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=CHECK28
136
137 // RUN: %clang_cc1 -DCK4 -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
138 // RUN: %clang_cc1 -DCK4 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
139 // RUN: %clang_cc1 -DCK4 -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
140 // RUN: %clang_cc1 -DCK4 -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
141 // RUN: %clang_cc1 -DCK4 -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
142 // RUN: %clang_cc1 -DCK4 -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
143
144 #ifdef CK4
145
146 template <typename T, int n>
tmain(T argc)147 int tmain(T argc) {
148 T a[n];
149 int te = n/128;
150 int th = 128;
151 #pragma omp target
152 #pragma omp teams distribute num_teams(te) thread_limit(th)
153 for(int i = 0; i < n; i++) {
154 a[i] = (T)0;
155 }
156 return 0;
157 }
158
main(int argc,char ** argv)159 int main (int argc, char **argv) {
160 int n = 100;
161 int a[n];
162 #pragma omp target
163 #pragma omp teams distribute
164 for(int i = 0; i < n; i++) {
165 a[i] = 0;
166 }
167 return tmain<int, 10>(argc);
168 }
169
170
171
172
173
174
175
176 #endif // CK4
177 #endif
178 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_argument_globali
179 // CHECK1-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
180 // CHECK1-NEXT: entry:
181 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
182 // CHECK1-NEXT: [[TE:%.*]] = alloca i32, align 4
183 // CHECK1-NEXT: [[TH:%.*]] = alloca i32, align 4
184 // CHECK1-NEXT: [[TE_CASTED:%.*]] = alloca i64, align 8
185 // CHECK1-NEXT: [[TH_CASTED:%.*]] = alloca i64, align 8
186 // CHECK1-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
187 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
188 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
189 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
190 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
191 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
192 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
193 // CHECK1-NEXT: [[N_CASTED6:%.*]] = alloca i64, align 8
194 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x i8*], align 8
195 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 8
196 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 8
197 // CHECK1-NEXT: [[_TMP11:%.*]] = alloca i32, align 4
198 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
199 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_13:%.*]] = alloca i32, align 4
200 // CHECK1-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
201 // CHECK1-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
202 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
203 // CHECK1-NEXT: store i32 [[DIV]], i32* [[TE]], align 4
204 // CHECK1-NEXT: store i32 128, i32* [[TH]], align 4
205 // CHECK1-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
206 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_CASTED]] to i32*
207 // CHECK1-NEXT: store i32 [[TMP1]], i32* [[CONV]], align 4
208 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[TE_CASTED]], align 8
209 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH]], align 4
210 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_CASTED]] to i32*
211 // CHECK1-NEXT: store i32 [[TMP3]], i32* [[CONV1]], align 4
212 // CHECK1-NEXT: [[TMP4:%.*]] = load i64, i64* [[TH_CASTED]], align 8
213 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[N_ADDR]], align 4
214 // CHECK1-NEXT: [[CONV2:%.*]] = bitcast i64* [[N_CASTED]] to i32*
215 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[CONV2]], align 4
216 // CHECK1-NEXT: [[TMP6:%.*]] = load i64, i64* [[N_CASTED]], align 8
217 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
218 // CHECK1-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
219 // CHECK1-NEXT: store i64 [[TMP2]], i64* [[TMP8]], align 8
220 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
221 // CHECK1-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
222 // CHECK1-NEXT: store i64 [[TMP2]], i64* [[TMP10]], align 8
223 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
224 // CHECK1-NEXT: store i8* null, i8** [[TMP11]], align 8
225 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
226 // CHECK1-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
227 // CHECK1-NEXT: store i64 [[TMP4]], i64* [[TMP13]], align 8
228 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
229 // CHECK1-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
230 // CHECK1-NEXT: store i64 [[TMP4]], i64* [[TMP15]], align 8
231 // CHECK1-NEXT: [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
232 // CHECK1-NEXT: store i8* null, i8** [[TMP16]], align 8
233 // CHECK1-NEXT: [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
234 // CHECK1-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
235 // CHECK1-NEXT: store i64 [[TMP6]], i64* [[TMP18]], align 8
236 // CHECK1-NEXT: [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
237 // CHECK1-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
238 // CHECK1-NEXT: store i64 [[TMP6]], i64* [[TMP20]], align 8
239 // CHECK1-NEXT: [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
240 // CHECK1-NEXT: store i8* null, i8** [[TMP21]], align 8
241 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
242 // CHECK1-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [100 x i32]**
243 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP23]], align 8
244 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
245 // CHECK1-NEXT: [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [100 x i32]**
246 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP25]], align 8
247 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
248 // CHECK1-NEXT: store i8* null, i8** [[TMP26]], align 8
249 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
250 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
251 // CHECK1-NEXT: [[TMP29:%.*]] = load i32, i32* [[TE]], align 4
252 // CHECK1-NEXT: [[TMP30:%.*]] = load i32, i32* [[TH]], align 4
253 // CHECK1-NEXT: [[TMP31:%.*]] = load i32, i32* [[N_ADDR]], align 4
254 // CHECK1-NEXT: store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR_]], align 4
255 // CHECK1-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
256 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP32]], 0
257 // CHECK1-NEXT: [[DIV4:%.*]] = sdiv i32 [[SUB]], 1
258 // CHECK1-NEXT: [[SUB5:%.*]] = sub nsw i32 [[DIV4]], 1
259 // CHECK1-NEXT: store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
260 // CHECK1-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
261 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP33]], 1
262 // CHECK1-NEXT: [[TMP34:%.*]] = zext i32 [[ADD]] to i64
263 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 [[TMP34]])
264 // CHECK1-NEXT: [[TMP35:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP29]], i32 [[TMP30]])
265 // CHECK1-NEXT: [[TMP36:%.*]] = icmp ne i32 [[TMP35]], 0
266 // CHECK1-NEXT: br i1 [[TMP36]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
267 // CHECK1: omp_offload.failed:
268 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29(i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [100 x i32]* @a) #[[ATTR2:[0-9]+]]
269 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
270 // CHECK1: omp_offload.cont:
271 // CHECK1-NEXT: [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
272 // CHECK1-NEXT: [[CONV7:%.*]] = bitcast i64* [[N_CASTED6]] to i32*
273 // CHECK1-NEXT: store i32 [[TMP37]], i32* [[CONV7]], align 4
274 // CHECK1-NEXT: [[TMP38:%.*]] = load i64, i64* [[N_CASTED6]], align 8
275 // CHECK1-NEXT: [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
276 // CHECK1-NEXT: [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i64*
277 // CHECK1-NEXT: store i64 [[TMP38]], i64* [[TMP40]], align 8
278 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
279 // CHECK1-NEXT: [[TMP42:%.*]] = bitcast i8** [[TMP41]] to i64*
280 // CHECK1-NEXT: store i64 [[TMP38]], i64* [[TMP42]], align 8
281 // CHECK1-NEXT: [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
282 // CHECK1-NEXT: store i8* null, i8** [[TMP43]], align 8
283 // CHECK1-NEXT: [[TMP44:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
284 // CHECK1-NEXT: [[TMP45:%.*]] = bitcast i8** [[TMP44]] to [100 x i32]**
285 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP45]], align 8
286 // CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
287 // CHECK1-NEXT: [[TMP47:%.*]] = bitcast i8** [[TMP46]] to [100 x i32]**
288 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP47]], align 8
289 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 1
290 // CHECK1-NEXT: store i8* null, i8** [[TMP48]], align 8
291 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
292 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
293 // CHECK1-NEXT: [[TMP51:%.*]] = load i32, i32* [[N_ADDR]], align 4
294 // CHECK1-NEXT: store i32 [[TMP51]], i32* [[DOTCAPTURE_EXPR_12]], align 4
295 // CHECK1-NEXT: [[TMP52:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
296 // CHECK1-NEXT: [[SUB14:%.*]] = sub nsw i32 [[TMP52]], 0
297 // CHECK1-NEXT: [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1
298 // CHECK1-NEXT: [[SUB16:%.*]] = sub nsw i32 [[DIV15]], 1
299 // CHECK1-NEXT: store i32 [[SUB16]], i32* [[DOTCAPTURE_EXPR_13]], align 4
300 // CHECK1-NEXT: [[TMP53:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_13]], align 4
301 // CHECK1-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP53]], 1
302 // CHECK1-NEXT: [[TMP54:%.*]] = zext i32 [[ADD17]] to i64
303 // CHECK1-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 [[TMP54]])
304 // CHECK1-NEXT: [[TMP55:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35.region_id, i32 2, i8** [[TMP49]], i8** [[TMP50]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
305 // CHECK1-NEXT: [[TMP56:%.*]] = icmp ne i32 [[TMP55]], 0
306 // CHECK1-NEXT: br i1 [[TMP56]], label [[OMP_OFFLOAD_FAILED18:%.*]], label [[OMP_OFFLOAD_CONT19:%.*]]
307 // CHECK1: omp_offload.failed18:
308 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35(i64 [[TMP38]], [100 x i32]* @a) #[[ATTR2]]
309 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT19]]
310 // CHECK1: omp_offload.cont19:
311 // CHECK1-NEXT: [[TMP57:%.*]] = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i64 0, i64 0), align 4
312 // CHECK1-NEXT: ret i32 [[TMP57]]
313 //
314 //
315 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29
316 // CHECK1-SAME: (i64 [[TE:%.*]], i64 [[TH:%.*]], i64 [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
317 // CHECK1-NEXT: entry:
318 // CHECK1-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8
319 // CHECK1-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8
320 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
321 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
322 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
323 // CHECK1-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8
324 // CHECK1-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8
325 // CHECK1-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
326 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
327 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_ADDR]] to i32*
328 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_ADDR]] to i32*
329 // CHECK1-NEXT: [[CONV2:%.*]] = bitcast i64* [[N_ADDR]] to i32*
330 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
331 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
332 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 8
333 // CHECK1-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
334 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV2]], [100 x i32]* [[TMP1]])
335 // CHECK1-NEXT: ret void
336 //
337 //
338 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
339 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
340 // CHECK1-NEXT: entry:
341 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
342 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
343 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
344 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
345 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
346 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
347 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
348 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
349 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
350 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
351 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
352 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
353 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
354 // CHECK1-NEXT: [[I3:%.*]] = alloca i32, align 4
355 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
356 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
357 // CHECK1-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
358 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
359 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
360 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
361 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
362 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
363 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
364 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
365 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
366 // CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
367 // CHECK1-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
368 // CHECK1-NEXT: store i32 0, i32* [[I]], align 4
369 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
370 // CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
371 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
372 // CHECK1: omp.precond.then:
373 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
374 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
375 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
376 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
377 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
378 // CHECK1-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
379 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
380 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
381 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
382 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
383 // CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
384 // CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
385 // CHECK1: cond.true:
386 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
387 // CHECK1-NEXT: br label [[COND_END:%.*]]
388 // CHECK1: cond.false:
389 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
390 // CHECK1-NEXT: br label [[COND_END]]
391 // CHECK1: cond.end:
392 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
393 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
394 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
395 // CHECK1-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
396 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
397 // CHECK1: omp.inner.for.cond:
398 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
399 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
400 // CHECK1-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
401 // CHECK1-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
402 // CHECK1: omp.inner.for.body:
403 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
404 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
405 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
406 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
407 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
408 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
409 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]]
410 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
411 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
412 // CHECK1: omp.body.continue:
413 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
414 // CHECK1: omp.inner.for.inc:
415 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
416 // CHECK1-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
417 // CHECK1-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
418 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
419 // CHECK1: omp.inner.for.end:
420 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
421 // CHECK1: omp.loop.exit:
422 // CHECK1-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
423 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
424 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
425 // CHECK1-NEXT: br label [[OMP_PRECOND_END]]
426 // CHECK1: omp.precond.end:
427 // CHECK1-NEXT: ret void
428 //
429 //
430 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35
431 // CHECK1-SAME: (i64 [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
432 // CHECK1-NEXT: entry:
433 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
434 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
435 // CHECK1-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
436 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
437 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
438 // CHECK1-NEXT: [[TMP0:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
439 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32* [[CONV]], [100 x i32]* [[TMP0]])
440 // CHECK1-NEXT: ret void
441 //
442 //
443 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
444 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
445 // CHECK1-NEXT: entry:
446 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
447 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
448 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
449 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
450 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
451 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
452 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
453 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
454 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
455 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
456 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
457 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
458 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
459 // CHECK1-NEXT: [[I3:%.*]] = alloca i32, align 4
460 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
461 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
462 // CHECK1-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
463 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
464 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
465 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
466 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
467 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
468 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
469 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
470 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
471 // CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
472 // CHECK1-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
473 // CHECK1-NEXT: store i32 0, i32* [[I]], align 4
474 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
475 // CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
476 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
477 // CHECK1: omp.precond.then:
478 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
479 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
480 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
481 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
482 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
483 // CHECK1-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
484 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
485 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
486 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
487 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
488 // CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
489 // CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
490 // CHECK1: cond.true:
491 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
492 // CHECK1-NEXT: br label [[COND_END:%.*]]
493 // CHECK1: cond.false:
494 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
495 // CHECK1-NEXT: br label [[COND_END]]
496 // CHECK1: cond.end:
497 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
498 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
499 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
500 // CHECK1-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
501 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
502 // CHECK1: omp.inner.for.cond:
503 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
504 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
505 // CHECK1-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
506 // CHECK1-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
507 // CHECK1: omp.inner.for.body:
508 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
509 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
510 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
511 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
512 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
513 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
514 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]]
515 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
516 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
517 // CHECK1: omp.body.continue:
518 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
519 // CHECK1: omp.inner.for.inc:
520 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
521 // CHECK1-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
522 // CHECK1-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
523 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
524 // CHECK1: omp.inner.for.end:
525 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
526 // CHECK1: omp.loop.exit:
527 // CHECK1-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
528 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
529 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
530 // CHECK1-NEXT: br label [[OMP_PRECOND_END]]
531 // CHECK1: omp.precond.end:
532 // CHECK1-NEXT: ret void
533 //
534 //
535 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
536 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
537 // CHECK1-NEXT: entry:
538 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
539 // CHECK1-NEXT: ret void
540 //
541 //
542 // CHECK2-LABEL: define {{[^@]+}}@_Z21teams_argument_globali
543 // CHECK2-SAME: (i32 signext [[N:%.*]]) #[[ATTR0:[0-9]+]] {
544 // CHECK2-NEXT: entry:
545 // CHECK2-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
546 // CHECK2-NEXT: [[TE:%.*]] = alloca i32, align 4
547 // CHECK2-NEXT: [[TH:%.*]] = alloca i32, align 4
548 // CHECK2-NEXT: [[TE_CASTED:%.*]] = alloca i64, align 8
549 // CHECK2-NEXT: [[TH_CASTED:%.*]] = alloca i64, align 8
550 // CHECK2-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
551 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 8
552 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 8
553 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 8
554 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
555 // CHECK2-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
556 // CHECK2-NEXT: [[DOTCAPTURE_EXPR_3:%.*]] = alloca i32, align 4
557 // CHECK2-NEXT: [[N_CASTED6:%.*]] = alloca i64, align 8
558 // CHECK2-NEXT: [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x i8*], align 8
559 // CHECK2-NEXT: [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 8
560 // CHECK2-NEXT: [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 8
561 // CHECK2-NEXT: [[_TMP11:%.*]] = alloca i32, align 4
562 // CHECK2-NEXT: [[DOTCAPTURE_EXPR_12:%.*]] = alloca i32, align 4
563 // CHECK2-NEXT: [[DOTCAPTURE_EXPR_13:%.*]] = alloca i32, align 4
564 // CHECK2-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
565 // CHECK2-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
566 // CHECK2-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
567 // CHECK2-NEXT: store i32 [[DIV]], i32* [[TE]], align 4
568 // CHECK2-NEXT: store i32 128, i32* [[TH]], align 4
569 // CHECK2-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
570 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_CASTED]] to i32*
571 // CHECK2-NEXT: store i32 [[TMP1]], i32* [[CONV]], align 4
572 // CHECK2-NEXT: [[TMP2:%.*]] = load i64, i64* [[TE_CASTED]], align 8
573 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH]], align 4
574 // CHECK2-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_CASTED]] to i32*
575 // CHECK2-NEXT: store i32 [[TMP3]], i32* [[CONV1]], align 4
576 // CHECK2-NEXT: [[TMP4:%.*]] = load i64, i64* [[TH_CASTED]], align 8
577 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[N_ADDR]], align 4
578 // CHECK2-NEXT: [[CONV2:%.*]] = bitcast i64* [[N_CASTED]] to i32*
579 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[CONV2]], align 4
580 // CHECK2-NEXT: [[TMP6:%.*]] = load i64, i64* [[N_CASTED]], align 8
581 // CHECK2-NEXT: [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
582 // CHECK2-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
583 // CHECK2-NEXT: store i64 [[TMP2]], i64* [[TMP8]], align 8
584 // CHECK2-NEXT: [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
585 // CHECK2-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
586 // CHECK2-NEXT: store i64 [[TMP2]], i64* [[TMP10]], align 8
587 // CHECK2-NEXT: [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
588 // CHECK2-NEXT: store i8* null, i8** [[TMP11]], align 8
589 // CHECK2-NEXT: [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
590 // CHECK2-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
591 // CHECK2-NEXT: store i64 [[TMP4]], i64* [[TMP13]], align 8
592 // CHECK2-NEXT: [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
593 // CHECK2-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
594 // CHECK2-NEXT: store i64 [[TMP4]], i64* [[TMP15]], align 8
595 // CHECK2-NEXT: [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
596 // CHECK2-NEXT: store i8* null, i8** [[TMP16]], align 8
597 // CHECK2-NEXT: [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
598 // CHECK2-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i64*
599 // CHECK2-NEXT: store i64 [[TMP6]], i64* [[TMP18]], align 8
600 // CHECK2-NEXT: [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
601 // CHECK2-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i64*
602 // CHECK2-NEXT: store i64 [[TMP6]], i64* [[TMP20]], align 8
603 // CHECK2-NEXT: [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
604 // CHECK2-NEXT: store i8* null, i8** [[TMP21]], align 8
605 // CHECK2-NEXT: [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
606 // CHECK2-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [100 x i32]**
607 // CHECK2-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP23]], align 8
608 // CHECK2-NEXT: [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
609 // CHECK2-NEXT: [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [100 x i32]**
610 // CHECK2-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP25]], align 8
611 // CHECK2-NEXT: [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
612 // CHECK2-NEXT: store i8* null, i8** [[TMP26]], align 8
613 // CHECK2-NEXT: [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
614 // CHECK2-NEXT: [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
615 // CHECK2-NEXT: [[TMP29:%.*]] = load i32, i32* [[TE]], align 4
616 // CHECK2-NEXT: [[TMP30:%.*]] = load i32, i32* [[TH]], align 4
617 // CHECK2-NEXT: [[TMP31:%.*]] = load i32, i32* [[N_ADDR]], align 4
618 // CHECK2-NEXT: store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR_]], align 4
619 // CHECK2-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
620 // CHECK2-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP32]], 0
621 // CHECK2-NEXT: [[DIV4:%.*]] = sdiv i32 [[SUB]], 1
622 // CHECK2-NEXT: [[SUB5:%.*]] = sub nsw i32 [[DIV4]], 1
623 // CHECK2-NEXT: store i32 [[SUB5]], i32* [[DOTCAPTURE_EXPR_3]], align 4
624 // CHECK2-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_3]], align 4
625 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP33]], 1
626 // CHECK2-NEXT: [[TMP34:%.*]] = zext i32 [[ADD]] to i64
627 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 [[TMP34]])
628 // CHECK2-NEXT: [[TMP35:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP29]], i32 [[TMP30]])
629 // CHECK2-NEXT: [[TMP36:%.*]] = icmp ne i32 [[TMP35]], 0
630 // CHECK2-NEXT: br i1 [[TMP36]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
631 // CHECK2: omp_offload.failed:
632 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29(i64 [[TMP2]], i64 [[TMP4]], i64 [[TMP6]], [100 x i32]* @a) #[[ATTR2:[0-9]+]]
633 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT]]
634 // CHECK2: omp_offload.cont:
635 // CHECK2-NEXT: [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
636 // CHECK2-NEXT: [[CONV7:%.*]] = bitcast i64* [[N_CASTED6]] to i32*
637 // CHECK2-NEXT: store i32 [[TMP37]], i32* [[CONV7]], align 4
638 // CHECK2-NEXT: [[TMP38:%.*]] = load i64, i64* [[N_CASTED6]], align 8
639 // CHECK2-NEXT: [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
640 // CHECK2-NEXT: [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i64*
641 // CHECK2-NEXT: store i64 [[TMP38]], i64* [[TMP40]], align 8
642 // CHECK2-NEXT: [[TMP41:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
643 // CHECK2-NEXT: [[TMP42:%.*]] = bitcast i8** [[TMP41]] to i64*
644 // CHECK2-NEXT: store i64 [[TMP38]], i64* [[TMP42]], align 8
645 // CHECK2-NEXT: [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
646 // CHECK2-NEXT: store i8* null, i8** [[TMP43]], align 8
647 // CHECK2-NEXT: [[TMP44:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
648 // CHECK2-NEXT: [[TMP45:%.*]] = bitcast i8** [[TMP44]] to [100 x i32]**
649 // CHECK2-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP45]], align 8
650 // CHECK2-NEXT: [[TMP46:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
651 // CHECK2-NEXT: [[TMP47:%.*]] = bitcast i8** [[TMP46]] to [100 x i32]**
652 // CHECK2-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP47]], align 8
653 // CHECK2-NEXT: [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 1
654 // CHECK2-NEXT: store i8* null, i8** [[TMP48]], align 8
655 // CHECK2-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
656 // CHECK2-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
657 // CHECK2-NEXT: [[TMP51:%.*]] = load i32, i32* [[N_ADDR]], align 4
658 // CHECK2-NEXT: store i32 [[TMP51]], i32* [[DOTCAPTURE_EXPR_12]], align 4
659 // CHECK2-NEXT: [[TMP52:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
660 // CHECK2-NEXT: [[SUB14:%.*]] = sub nsw i32 [[TMP52]], 0
661 // CHECK2-NEXT: [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1
662 // CHECK2-NEXT: [[SUB16:%.*]] = sub nsw i32 [[DIV15]], 1
663 // CHECK2-NEXT: store i32 [[SUB16]], i32* [[DOTCAPTURE_EXPR_13]], align 4
664 // CHECK2-NEXT: [[TMP53:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_13]], align 4
665 // CHECK2-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP53]], 1
666 // CHECK2-NEXT: [[TMP54:%.*]] = zext i32 [[ADD17]] to i64
667 // CHECK2-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 [[TMP54]])
668 // CHECK2-NEXT: [[TMP55:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35.region_id, i32 2, i8** [[TMP49]], i8** [[TMP50]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
669 // CHECK2-NEXT: [[TMP56:%.*]] = icmp ne i32 [[TMP55]], 0
670 // CHECK2-NEXT: br i1 [[TMP56]], label [[OMP_OFFLOAD_FAILED18:%.*]], label [[OMP_OFFLOAD_CONT19:%.*]]
671 // CHECK2: omp_offload.failed18:
672 // CHECK2-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35(i64 [[TMP38]], [100 x i32]* @a) #[[ATTR2]]
673 // CHECK2-NEXT: br label [[OMP_OFFLOAD_CONT19]]
674 // CHECK2: omp_offload.cont19:
675 // CHECK2-NEXT: [[TMP57:%.*]] = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i64 0, i64 0), align 4
676 // CHECK2-NEXT: ret i32 [[TMP57]]
677 //
678 //
679 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29
680 // CHECK2-SAME: (i64 [[TE:%.*]], i64 [[TH:%.*]], i64 [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
681 // CHECK2-NEXT: entry:
682 // CHECK2-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8
683 // CHECK2-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8
684 // CHECK2-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
685 // CHECK2-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
686 // CHECK2-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
687 // CHECK2-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8
688 // CHECK2-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8
689 // CHECK2-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
690 // CHECK2-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
691 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_ADDR]] to i32*
692 // CHECK2-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_ADDR]] to i32*
693 // CHECK2-NEXT: [[CONV2:%.*]] = bitcast i64* [[N_ADDR]] to i32*
694 // CHECK2-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
695 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
696 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 8
697 // CHECK2-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
698 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV2]], [100 x i32]* [[TMP1]])
699 // CHECK2-NEXT: ret void
700 //
701 //
702 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
703 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
704 // CHECK2-NEXT: entry:
705 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
706 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
707 // CHECK2-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
708 // CHECK2-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
709 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
710 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
711 // CHECK2-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
712 // CHECK2-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
713 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
714 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
715 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
716 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
717 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
718 // CHECK2-NEXT: [[I3:%.*]] = alloca i32, align 4
719 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
720 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
721 // CHECK2-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
722 // CHECK2-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
723 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
724 // CHECK2-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
725 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
726 // CHECK2-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
727 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
728 // CHECK2-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
729 // CHECK2-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
730 // CHECK2-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
731 // CHECK2-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
732 // CHECK2-NEXT: store i32 0, i32* [[I]], align 4
733 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
734 // CHECK2-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
735 // CHECK2-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
736 // CHECK2: omp.precond.then:
737 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
738 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
739 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
740 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
741 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
742 // CHECK2-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
743 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
744 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
745 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
746 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
747 // CHECK2-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
748 // CHECK2-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
749 // CHECK2: cond.true:
750 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
751 // CHECK2-NEXT: br label [[COND_END:%.*]]
752 // CHECK2: cond.false:
753 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
754 // CHECK2-NEXT: br label [[COND_END]]
755 // CHECK2: cond.end:
756 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
757 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
758 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
759 // CHECK2-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
760 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
761 // CHECK2: omp.inner.for.cond:
762 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
763 // CHECK2-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
764 // CHECK2-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
765 // CHECK2-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
766 // CHECK2: omp.inner.for.body:
767 // CHECK2-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
768 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
769 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
770 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
771 // CHECK2-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
772 // CHECK2-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
773 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]]
774 // CHECK2-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
775 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
776 // CHECK2: omp.body.continue:
777 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
778 // CHECK2: omp.inner.for.inc:
779 // CHECK2-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
780 // CHECK2-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
781 // CHECK2-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
782 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
783 // CHECK2: omp.inner.for.end:
784 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
785 // CHECK2: omp.loop.exit:
786 // CHECK2-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
787 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
788 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
789 // CHECK2-NEXT: br label [[OMP_PRECOND_END]]
790 // CHECK2: omp.precond.end:
791 // CHECK2-NEXT: ret void
792 //
793 //
794 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35
795 // CHECK2-SAME: (i64 [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
796 // CHECK2-NEXT: entry:
797 // CHECK2-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
798 // CHECK2-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
799 // CHECK2-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
800 // CHECK2-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
801 // CHECK2-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
802 // CHECK2-NEXT: [[TMP0:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
803 // CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32* [[CONV]], [100 x i32]* [[TMP0]])
804 // CHECK2-NEXT: ret void
805 //
806 //
807 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
808 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
809 // CHECK2-NEXT: entry:
810 // CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
811 // CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
812 // CHECK2-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
813 // CHECK2-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
814 // CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
815 // CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
816 // CHECK2-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
817 // CHECK2-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
818 // CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
819 // CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
820 // CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
821 // CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
822 // CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
823 // CHECK2-NEXT: [[I3:%.*]] = alloca i32, align 4
824 // CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
825 // CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
826 // CHECK2-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
827 // CHECK2-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
828 // CHECK2-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
829 // CHECK2-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
830 // CHECK2-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
831 // CHECK2-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
832 // CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
833 // CHECK2-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
834 // CHECK2-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
835 // CHECK2-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
836 // CHECK2-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
837 // CHECK2-NEXT: store i32 0, i32* [[I]], align 4
838 // CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
839 // CHECK2-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
840 // CHECK2-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
841 // CHECK2: omp.precond.then:
842 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
843 // CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
844 // CHECK2-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
845 // CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
846 // CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
847 // CHECK2-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
848 // CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
849 // CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
850 // CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
851 // CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
852 // CHECK2-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
853 // CHECK2-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
854 // CHECK2: cond.true:
855 // CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
856 // CHECK2-NEXT: br label [[COND_END:%.*]]
857 // CHECK2: cond.false:
858 // CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
859 // CHECK2-NEXT: br label [[COND_END]]
860 // CHECK2: cond.end:
861 // CHECK2-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
862 // CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
863 // CHECK2-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
864 // CHECK2-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
865 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
866 // CHECK2: omp.inner.for.cond:
867 // CHECK2-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
868 // CHECK2-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
869 // CHECK2-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
870 // CHECK2-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
871 // CHECK2: omp.inner.for.body:
872 // CHECK2-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
873 // CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
874 // CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
875 // CHECK2-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
876 // CHECK2-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
877 // CHECK2-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
878 // CHECK2-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]]
879 // CHECK2-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
880 // CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
881 // CHECK2: omp.body.continue:
882 // CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
883 // CHECK2: omp.inner.for.inc:
884 // CHECK2-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
885 // CHECK2-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
886 // CHECK2-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
887 // CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]]
888 // CHECK2: omp.inner.for.end:
889 // CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
890 // CHECK2: omp.loop.exit:
891 // CHECK2-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
892 // CHECK2-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
893 // CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
894 // CHECK2-NEXT: br label [[OMP_PRECOND_END]]
895 // CHECK2: omp.precond.end:
896 // CHECK2-NEXT: ret void
897 //
898 //
899 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
900 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] {
901 // CHECK2-NEXT: entry:
902 // CHECK2-NEXT: call void @__tgt_register_requires(i64 1)
903 // CHECK2-NEXT: ret void
904 //
905 //
906 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_argument_globali
907 // CHECK3-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
908 // CHECK3-NEXT: entry:
909 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
910 // CHECK3-NEXT: [[TE:%.*]] = alloca i32, align 4
911 // CHECK3-NEXT: [[TH:%.*]] = alloca i32, align 4
912 // CHECK3-NEXT: [[TE_CASTED:%.*]] = alloca i32, align 4
913 // CHECK3-NEXT: [[TH_CASTED:%.*]] = alloca i32, align 4
914 // CHECK3-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
915 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
916 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
917 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
918 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
919 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
920 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
921 // CHECK3-NEXT: [[N_CASTED4:%.*]] = alloca i32, align 4
922 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4
923 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4
924 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4
925 // CHECK3-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
926 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
927 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
928 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
929 // CHECK3-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
930 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
931 // CHECK3-NEXT: store i32 [[DIV]], i32* [[TE]], align 4
932 // CHECK3-NEXT: store i32 128, i32* [[TH]], align 4
933 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
934 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[TE_CASTED]], align 4
935 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_CASTED]], align 4
936 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH]], align 4
937 // CHECK3-NEXT: store i32 [[TMP3]], i32* [[TH_CASTED]], align 4
938 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TH_CASTED]], align 4
939 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[N_ADDR]], align 4
940 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[N_CASTED]], align 4
941 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[N_CASTED]], align 4
942 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
943 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
944 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[TMP8]], align 4
945 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
946 // CHECK3-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
947 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[TMP10]], align 4
948 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
949 // CHECK3-NEXT: store i8* null, i8** [[TMP11]], align 4
950 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
951 // CHECK3-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
952 // CHECK3-NEXT: store i32 [[TMP4]], i32* [[TMP13]], align 4
953 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
954 // CHECK3-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
955 // CHECK3-NEXT: store i32 [[TMP4]], i32* [[TMP15]], align 4
956 // CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
957 // CHECK3-NEXT: store i8* null, i8** [[TMP16]], align 4
958 // CHECK3-NEXT: [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
959 // CHECK3-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
960 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[TMP18]], align 4
961 // CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
962 // CHECK3-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
963 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[TMP20]], align 4
964 // CHECK3-NEXT: [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
965 // CHECK3-NEXT: store i8* null, i8** [[TMP21]], align 4
966 // CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
967 // CHECK3-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [100 x i32]**
968 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP23]], align 4
969 // CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
970 // CHECK3-NEXT: [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [100 x i32]**
971 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP25]], align 4
972 // CHECK3-NEXT: [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
973 // CHECK3-NEXT: store i8* null, i8** [[TMP26]], align 4
974 // CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
975 // CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
976 // CHECK3-NEXT: [[TMP29:%.*]] = load i32, i32* [[TE]], align 4
977 // CHECK3-NEXT: [[TMP30:%.*]] = load i32, i32* [[TH]], align 4
978 // CHECK3-NEXT: [[TMP31:%.*]] = load i32, i32* [[N_ADDR]], align 4
979 // CHECK3-NEXT: store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR_]], align 4
980 // CHECK3-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
981 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP32]], 0
982 // CHECK3-NEXT: [[DIV2:%.*]] = sdiv i32 [[SUB]], 1
983 // CHECK3-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV2]], 1
984 // CHECK3-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
985 // CHECK3-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
986 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP33]], 1
987 // CHECK3-NEXT: [[TMP34:%.*]] = zext i32 [[ADD]] to i64
988 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 [[TMP34]])
989 // CHECK3-NEXT: [[TMP35:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP29]], i32 [[TMP30]])
990 // CHECK3-NEXT: [[TMP36:%.*]] = icmp ne i32 [[TMP35]], 0
991 // CHECK3-NEXT: br i1 [[TMP36]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
992 // CHECK3: omp_offload.failed:
993 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29(i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [100 x i32]* @a) #[[ATTR2:[0-9]+]]
994 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
995 // CHECK3: omp_offload.cont:
996 // CHECK3-NEXT: [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
997 // CHECK3-NEXT: store i32 [[TMP37]], i32* [[N_CASTED4]], align 4
998 // CHECK3-NEXT: [[TMP38:%.*]] = load i32, i32* [[N_CASTED4]], align 4
999 // CHECK3-NEXT: [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
1000 // CHECK3-NEXT: [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
1001 // CHECK3-NEXT: store i32 [[TMP38]], i32* [[TMP40]], align 4
1002 // CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
1003 // CHECK3-NEXT: [[TMP42:%.*]] = bitcast i8** [[TMP41]] to i32*
1004 // CHECK3-NEXT: store i32 [[TMP38]], i32* [[TMP42]], align 4
1005 // CHECK3-NEXT: [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
1006 // CHECK3-NEXT: store i8* null, i8** [[TMP43]], align 4
1007 // CHECK3-NEXT: [[TMP44:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
1008 // CHECK3-NEXT: [[TMP45:%.*]] = bitcast i8** [[TMP44]] to [100 x i32]**
1009 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP45]], align 4
1010 // CHECK3-NEXT: [[TMP46:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
1011 // CHECK3-NEXT: [[TMP47:%.*]] = bitcast i8** [[TMP46]] to [100 x i32]**
1012 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP47]], align 4
1013 // CHECK3-NEXT: [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
1014 // CHECK3-NEXT: store i8* null, i8** [[TMP48]], align 4
1015 // CHECK3-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
1016 // CHECK3-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
1017 // CHECK3-NEXT: [[TMP51:%.*]] = load i32, i32* [[N_ADDR]], align 4
1018 // CHECK3-NEXT: store i32 [[TMP51]], i32* [[DOTCAPTURE_EXPR_9]], align 4
1019 // CHECK3-NEXT: [[TMP52:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
1020 // CHECK3-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP52]], 0
1021 // CHECK3-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
1022 // CHECK3-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
1023 // CHECK3-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
1024 // CHECK3-NEXT: [[TMP53:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
1025 // CHECK3-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP53]], 1
1026 // CHECK3-NEXT: [[TMP54:%.*]] = zext i32 [[ADD14]] to i64
1027 // CHECK3-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 [[TMP54]])
1028 // CHECK3-NEXT: [[TMP55:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35.region_id, i32 2, i8** [[TMP49]], i8** [[TMP50]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1029 // CHECK3-NEXT: [[TMP56:%.*]] = icmp ne i32 [[TMP55]], 0
1030 // CHECK3-NEXT: br i1 [[TMP56]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
1031 // CHECK3: omp_offload.failed15:
1032 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35(i32 [[TMP38]], [100 x i32]* @a) #[[ATTR2]]
1033 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT16]]
1034 // CHECK3: omp_offload.cont16:
1035 // CHECK3-NEXT: [[TMP57:%.*]] = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i32 0, i32 0), align 4
1036 // CHECK3-NEXT: ret i32 [[TMP57]]
1037 //
1038 //
1039 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29
1040 // CHECK3-SAME: (i32 [[TE:%.*]], i32 [[TH:%.*]], i32 [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
1041 // CHECK3-NEXT: entry:
1042 // CHECK3-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4
1043 // CHECK3-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4
1044 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
1045 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
1046 // CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
1047 // CHECK3-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4
1048 // CHECK3-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4
1049 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
1050 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
1051 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
1052 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4
1053 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4
1054 // CHECK3-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
1055 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], [100 x i32]* [[TMP1]])
1056 // CHECK3-NEXT: ret void
1057 //
1058 //
1059 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
1060 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
1061 // CHECK3-NEXT: entry:
1062 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1063 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1064 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
1065 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
1066 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1067 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1068 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1069 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1070 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1071 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1072 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1073 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1074 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1075 // CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4
1076 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1077 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1078 // CHECK3-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
1079 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
1080 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
1081 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
1082 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
1083 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
1084 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1085 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1086 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1087 // CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1088 // CHECK3-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1089 // CHECK3-NEXT: store i32 0, i32* [[I]], align 4
1090 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1091 // CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1092 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1093 // CHECK3: omp.precond.then:
1094 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1095 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1096 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
1097 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1098 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1099 // CHECK3-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1100 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
1101 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1102 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1103 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1104 // CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
1105 // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1106 // CHECK3: cond.true:
1107 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1108 // CHECK3-NEXT: br label [[COND_END:%.*]]
1109 // CHECK3: cond.false:
1110 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1111 // CHECK3-NEXT: br label [[COND_END]]
1112 // CHECK3: cond.end:
1113 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1114 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1115 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1116 // CHECK3-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
1117 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1118 // CHECK3: omp.inner.for.cond:
1119 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1120 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1121 // CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1122 // CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1123 // CHECK3: omp.inner.for.body:
1124 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1125 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
1126 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1127 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
1128 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
1129 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i32 0, i32 [[TMP16]]
1130 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1131 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1132 // CHECK3: omp.body.continue:
1133 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1134 // CHECK3: omp.inner.for.inc:
1135 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1136 // CHECK3-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
1137 // CHECK3-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1138 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
1139 // CHECK3: omp.inner.for.end:
1140 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1141 // CHECK3: omp.loop.exit:
1142 // CHECK3-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1143 // CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
1144 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
1145 // CHECK3-NEXT: br label [[OMP_PRECOND_END]]
1146 // CHECK3: omp.precond.end:
1147 // CHECK3-NEXT: ret void
1148 //
1149 //
1150 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35
1151 // CHECK3-SAME: (i32 [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
1152 // CHECK3-NEXT: entry:
1153 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
1154 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
1155 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
1156 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
1157 // CHECK3-NEXT: [[TMP0:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
1158 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], [100 x i32]* [[TMP0]])
1159 // CHECK3-NEXT: ret void
1160 //
1161 //
1162 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
1163 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
1164 // CHECK3-NEXT: entry:
1165 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1166 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1167 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
1168 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
1169 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1170 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1171 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1172 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1173 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1174 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1175 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1176 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1177 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1178 // CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4
1179 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1180 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1181 // CHECK3-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
1182 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
1183 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
1184 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
1185 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
1186 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
1187 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1188 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1189 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1190 // CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1191 // CHECK3-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1192 // CHECK3-NEXT: store i32 0, i32* [[I]], align 4
1193 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1194 // CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1195 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1196 // CHECK3: omp.precond.then:
1197 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1198 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1199 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
1200 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1201 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1202 // CHECK3-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1203 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
1204 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1205 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1206 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1207 // CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
1208 // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1209 // CHECK3: cond.true:
1210 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1211 // CHECK3-NEXT: br label [[COND_END:%.*]]
1212 // CHECK3: cond.false:
1213 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1214 // CHECK3-NEXT: br label [[COND_END]]
1215 // CHECK3: cond.end:
1216 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1217 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1218 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1219 // CHECK3-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
1220 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1221 // CHECK3: omp.inner.for.cond:
1222 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1223 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1224 // CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1225 // CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1226 // CHECK3: omp.inner.for.body:
1227 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1228 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
1229 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1230 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
1231 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
1232 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i32 0, i32 [[TMP16]]
1233 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1234 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1235 // CHECK3: omp.body.continue:
1236 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1237 // CHECK3: omp.inner.for.inc:
1238 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1239 // CHECK3-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
1240 // CHECK3-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1241 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
1242 // CHECK3: omp.inner.for.end:
1243 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1244 // CHECK3: omp.loop.exit:
1245 // CHECK3-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1246 // CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
1247 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
1248 // CHECK3-NEXT: br label [[OMP_PRECOND_END]]
1249 // CHECK3: omp.precond.end:
1250 // CHECK3-NEXT: ret void
1251 //
1252 //
1253 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1254 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
1255 // CHECK3-NEXT: entry:
1256 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
1257 // CHECK3-NEXT: ret void
1258 //
1259 //
1260 // CHECK4-LABEL: define {{[^@]+}}@_Z21teams_argument_globali
1261 // CHECK4-SAME: (i32 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
1262 // CHECK4-NEXT: entry:
1263 // CHECK4-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
1264 // CHECK4-NEXT: [[TE:%.*]] = alloca i32, align 4
1265 // CHECK4-NEXT: [[TH:%.*]] = alloca i32, align 4
1266 // CHECK4-NEXT: [[TE_CASTED:%.*]] = alloca i32, align 4
1267 // CHECK4-NEXT: [[TH_CASTED:%.*]] = alloca i32, align 4
1268 // CHECK4-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
1269 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
1270 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
1271 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
1272 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
1273 // CHECK4-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1274 // CHECK4-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1275 // CHECK4-NEXT: [[N_CASTED4:%.*]] = alloca i32, align 4
1276 // CHECK4-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4
1277 // CHECK4-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4
1278 // CHECK4-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4
1279 // CHECK4-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
1280 // CHECK4-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
1281 // CHECK4-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
1282 // CHECK4-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
1283 // CHECK4-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1284 // CHECK4-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
1285 // CHECK4-NEXT: store i32 [[DIV]], i32* [[TE]], align 4
1286 // CHECK4-NEXT: store i32 128, i32* [[TH]], align 4
1287 // CHECK4-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
1288 // CHECK4-NEXT: store i32 [[TMP1]], i32* [[TE_CASTED]], align 4
1289 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_CASTED]], align 4
1290 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH]], align 4
1291 // CHECK4-NEXT: store i32 [[TMP3]], i32* [[TH_CASTED]], align 4
1292 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[TH_CASTED]], align 4
1293 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[N_ADDR]], align 4
1294 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[N_CASTED]], align 4
1295 // CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[N_CASTED]], align 4
1296 // CHECK4-NEXT: [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1297 // CHECK4-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
1298 // CHECK4-NEXT: store i32 [[TMP2]], i32* [[TMP8]], align 4
1299 // CHECK4-NEXT: [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1300 // CHECK4-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
1301 // CHECK4-NEXT: store i32 [[TMP2]], i32* [[TMP10]], align 4
1302 // CHECK4-NEXT: [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1303 // CHECK4-NEXT: store i8* null, i8** [[TMP11]], align 4
1304 // CHECK4-NEXT: [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1305 // CHECK4-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
1306 // CHECK4-NEXT: store i32 [[TMP4]], i32* [[TMP13]], align 4
1307 // CHECK4-NEXT: [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1308 // CHECK4-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
1309 // CHECK4-NEXT: store i32 [[TMP4]], i32* [[TMP15]], align 4
1310 // CHECK4-NEXT: [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1311 // CHECK4-NEXT: store i8* null, i8** [[TMP16]], align 4
1312 // CHECK4-NEXT: [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1313 // CHECK4-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
1314 // CHECK4-NEXT: store i32 [[TMP6]], i32* [[TMP18]], align 4
1315 // CHECK4-NEXT: [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1316 // CHECK4-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
1317 // CHECK4-NEXT: store i32 [[TMP6]], i32* [[TMP20]], align 4
1318 // CHECK4-NEXT: [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1319 // CHECK4-NEXT: store i8* null, i8** [[TMP21]], align 4
1320 // CHECK4-NEXT: [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1321 // CHECK4-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [100 x i32]**
1322 // CHECK4-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP23]], align 4
1323 // CHECK4-NEXT: [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1324 // CHECK4-NEXT: [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [100 x i32]**
1325 // CHECK4-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP25]], align 4
1326 // CHECK4-NEXT: [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
1327 // CHECK4-NEXT: store i8* null, i8** [[TMP26]], align 4
1328 // CHECK4-NEXT: [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1329 // CHECK4-NEXT: [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1330 // CHECK4-NEXT: [[TMP29:%.*]] = load i32, i32* [[TE]], align 4
1331 // CHECK4-NEXT: [[TMP30:%.*]] = load i32, i32* [[TH]], align 4
1332 // CHECK4-NEXT: [[TMP31:%.*]] = load i32, i32* [[N_ADDR]], align 4
1333 // CHECK4-NEXT: store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR_]], align 4
1334 // CHECK4-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1335 // CHECK4-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP32]], 0
1336 // CHECK4-NEXT: [[DIV2:%.*]] = sdiv i32 [[SUB]], 1
1337 // CHECK4-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV2]], 1
1338 // CHECK4-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1339 // CHECK4-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1340 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP33]], 1
1341 // CHECK4-NEXT: [[TMP34:%.*]] = zext i32 [[ADD]] to i64
1342 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 [[TMP34]])
1343 // CHECK4-NEXT: [[TMP35:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29.region_id, i32 4, i8** [[TMP27]], i8** [[TMP28]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 [[TMP29]], i32 [[TMP30]])
1344 // CHECK4-NEXT: [[TMP36:%.*]] = icmp ne i32 [[TMP35]], 0
1345 // CHECK4-NEXT: br i1 [[TMP36]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1346 // CHECK4: omp_offload.failed:
1347 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29(i32 [[TMP2]], i32 [[TMP4]], i32 [[TMP6]], [100 x i32]* @a) #[[ATTR2:[0-9]+]]
1348 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT]]
1349 // CHECK4: omp_offload.cont:
1350 // CHECK4-NEXT: [[TMP37:%.*]] = load i32, i32* [[N_ADDR]], align 4
1351 // CHECK4-NEXT: store i32 [[TMP37]], i32* [[N_CASTED4]], align 4
1352 // CHECK4-NEXT: [[TMP38:%.*]] = load i32, i32* [[N_CASTED4]], align 4
1353 // CHECK4-NEXT: [[TMP39:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
1354 // CHECK4-NEXT: [[TMP40:%.*]] = bitcast i8** [[TMP39]] to i32*
1355 // CHECK4-NEXT: store i32 [[TMP38]], i32* [[TMP40]], align 4
1356 // CHECK4-NEXT: [[TMP41:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
1357 // CHECK4-NEXT: [[TMP42:%.*]] = bitcast i8** [[TMP41]] to i32*
1358 // CHECK4-NEXT: store i32 [[TMP38]], i32* [[TMP42]], align 4
1359 // CHECK4-NEXT: [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
1360 // CHECK4-NEXT: store i8* null, i8** [[TMP43]], align 4
1361 // CHECK4-NEXT: [[TMP44:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
1362 // CHECK4-NEXT: [[TMP45:%.*]] = bitcast i8** [[TMP44]] to [100 x i32]**
1363 // CHECK4-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP45]], align 4
1364 // CHECK4-NEXT: [[TMP46:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
1365 // CHECK4-NEXT: [[TMP47:%.*]] = bitcast i8** [[TMP46]] to [100 x i32]**
1366 // CHECK4-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP47]], align 4
1367 // CHECK4-NEXT: [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
1368 // CHECK4-NEXT: store i8* null, i8** [[TMP48]], align 4
1369 // CHECK4-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
1370 // CHECK4-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
1371 // CHECK4-NEXT: [[TMP51:%.*]] = load i32, i32* [[N_ADDR]], align 4
1372 // CHECK4-NEXT: store i32 [[TMP51]], i32* [[DOTCAPTURE_EXPR_9]], align 4
1373 // CHECK4-NEXT: [[TMP52:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
1374 // CHECK4-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP52]], 0
1375 // CHECK4-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
1376 // CHECK4-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
1377 // CHECK4-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
1378 // CHECK4-NEXT: [[TMP53:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
1379 // CHECK4-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP53]], 1
1380 // CHECK4-NEXT: [[TMP54:%.*]] = zext i32 [[ADD14]] to i64
1381 // CHECK4-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 [[TMP54]])
1382 // CHECK4-NEXT: [[TMP55:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35.region_id, i32 2, i8** [[TMP49]], i8** [[TMP50]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.2, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.3, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
1383 // CHECK4-NEXT: [[TMP56:%.*]] = icmp ne i32 [[TMP55]], 0
1384 // CHECK4-NEXT: br i1 [[TMP56]], label [[OMP_OFFLOAD_FAILED15:%.*]], label [[OMP_OFFLOAD_CONT16:%.*]]
1385 // CHECK4: omp_offload.failed15:
1386 // CHECK4-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35(i32 [[TMP38]], [100 x i32]* @a) #[[ATTR2]]
1387 // CHECK4-NEXT: br label [[OMP_OFFLOAD_CONT16]]
1388 // CHECK4: omp_offload.cont16:
1389 // CHECK4-NEXT: [[TMP57:%.*]] = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i32 0, i32 0), align 4
1390 // CHECK4-NEXT: ret i32 [[TMP57]]
1391 //
1392 //
1393 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29
1394 // CHECK4-SAME: (i32 [[TE:%.*]], i32 [[TH:%.*]], i32 [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
1395 // CHECK4-NEXT: entry:
1396 // CHECK4-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4
1397 // CHECK4-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4
1398 // CHECK4-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
1399 // CHECK4-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
1400 // CHECK4-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
1401 // CHECK4-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4
1402 // CHECK4-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4
1403 // CHECK4-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
1404 // CHECK4-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
1405 // CHECK4-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
1406 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4
1407 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4
1408 // CHECK4-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
1409 // CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], [100 x i32]* [[TMP1]])
1410 // CHECK4-NEXT: ret void
1411 //
1412 //
1413 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
1414 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
1415 // CHECK4-NEXT: entry:
1416 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1417 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1418 // CHECK4-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
1419 // CHECK4-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
1420 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1421 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
1422 // CHECK4-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1423 // CHECK4-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1424 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
1425 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1426 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1427 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1428 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1429 // CHECK4-NEXT: [[I3:%.*]] = alloca i32, align 4
1430 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1431 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1432 // CHECK4-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
1433 // CHECK4-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
1434 // CHECK4-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
1435 // CHECK4-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
1436 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
1437 // CHECK4-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
1438 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1439 // CHECK4-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1440 // CHECK4-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1441 // CHECK4-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1442 // CHECK4-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1443 // CHECK4-NEXT: store i32 0, i32* [[I]], align 4
1444 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1445 // CHECK4-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1446 // CHECK4-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1447 // CHECK4: omp.precond.then:
1448 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1449 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1450 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
1451 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1452 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1453 // CHECK4-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1454 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
1455 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1456 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1457 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1458 // CHECK4-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
1459 // CHECK4-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1460 // CHECK4: cond.true:
1461 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1462 // CHECK4-NEXT: br label [[COND_END:%.*]]
1463 // CHECK4: cond.false:
1464 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1465 // CHECK4-NEXT: br label [[COND_END]]
1466 // CHECK4: cond.end:
1467 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1468 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1469 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1470 // CHECK4-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
1471 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1472 // CHECK4: omp.inner.for.cond:
1473 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1474 // CHECK4-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1475 // CHECK4-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1476 // CHECK4-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1477 // CHECK4: omp.inner.for.body:
1478 // CHECK4-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1479 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
1480 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1481 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
1482 // CHECK4-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
1483 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i32 0, i32 [[TMP16]]
1484 // CHECK4-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1485 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1486 // CHECK4: omp.body.continue:
1487 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1488 // CHECK4: omp.inner.for.inc:
1489 // CHECK4-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1490 // CHECK4-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
1491 // CHECK4-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1492 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]]
1493 // CHECK4: omp.inner.for.end:
1494 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1495 // CHECK4: omp.loop.exit:
1496 // CHECK4-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1497 // CHECK4-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
1498 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
1499 // CHECK4-NEXT: br label [[OMP_PRECOND_END]]
1500 // CHECK4: omp.precond.end:
1501 // CHECK4-NEXT: ret void
1502 //
1503 //
1504 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35
1505 // CHECK4-SAME: (i32 [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
1506 // CHECK4-NEXT: entry:
1507 // CHECK4-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
1508 // CHECK4-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
1509 // CHECK4-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
1510 // CHECK4-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
1511 // CHECK4-NEXT: [[TMP0:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
1512 // CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, [100 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], [100 x i32]* [[TMP0]])
1513 // CHECK4-NEXT: ret void
1514 //
1515 //
1516 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
1517 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i32* nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
1518 // CHECK4-NEXT: entry:
1519 // CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1520 // CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1521 // CHECK4-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
1522 // CHECK4-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
1523 // CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1524 // CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
1525 // CHECK4-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1526 // CHECK4-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1527 // CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
1528 // CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1529 // CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1530 // CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1531 // CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1532 // CHECK4-NEXT: [[I3:%.*]] = alloca i32, align 4
1533 // CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1534 // CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1535 // CHECK4-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
1536 // CHECK4-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
1537 // CHECK4-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
1538 // CHECK4-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
1539 // CHECK4-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
1540 // CHECK4-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
1541 // CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1542 // CHECK4-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
1543 // CHECK4-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1544 // CHECK4-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1545 // CHECK4-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1546 // CHECK4-NEXT: store i32 0, i32* [[I]], align 4
1547 // CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1548 // CHECK4-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
1549 // CHECK4-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1550 // CHECK4: omp.precond.then:
1551 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1552 // CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1553 // CHECK4-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
1554 // CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1555 // CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1556 // CHECK4-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1557 // CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
1558 // CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP7]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1559 // CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1560 // CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1561 // CHECK4-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
1562 // CHECK4-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1563 // CHECK4: cond.true:
1564 // CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1565 // CHECK4-NEXT: br label [[COND_END:%.*]]
1566 // CHECK4: cond.false:
1567 // CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1568 // CHECK4-NEXT: br label [[COND_END]]
1569 // CHECK4: cond.end:
1570 // CHECK4-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1571 // CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1572 // CHECK4-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1573 // CHECK4-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
1574 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1575 // CHECK4: omp.inner.for.cond:
1576 // CHECK4-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1577 // CHECK4-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1578 // CHECK4-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1579 // CHECK4-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1580 // CHECK4: omp.inner.for.body:
1581 // CHECK4-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1582 // CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
1583 // CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1584 // CHECK4-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
1585 // CHECK4-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
1586 // CHECK4-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i32 0, i32 [[TMP16]]
1587 // CHECK4-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1588 // CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1589 // CHECK4: omp.body.continue:
1590 // CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1591 // CHECK4: omp.inner.for.inc:
1592 // CHECK4-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1593 // CHECK4-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
1594 // CHECK4-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1595 // CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]]
1596 // CHECK4: omp.inner.for.end:
1597 // CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1598 // CHECK4: omp.loop.exit:
1599 // CHECK4-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1600 // CHECK4-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
1601 // CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
1602 // CHECK4-NEXT: br label [[OMP_PRECOND_END]]
1603 // CHECK4: omp.precond.end:
1604 // CHECK4-NEXT: ret void
1605 //
1606 //
1607 // CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1608 // CHECK4-SAME: () #[[ATTR3:[0-9]+]] {
1609 // CHECK4-NEXT: entry:
1610 // CHECK4-NEXT: call void @__tgt_register_requires(i64 1)
1611 // CHECK4-NEXT: ret void
1612 //
1613 //
1614 // CHECK9-LABEL: define {{[^@]+}}@_Z15teams_local_argv
1615 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
1616 // CHECK9-NEXT: entry:
1617 // CHECK9-NEXT: [[N:%.*]] = alloca i32, align 4
1618 // CHECK9-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
1619 // CHECK9-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1620 // CHECK9-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
1621 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1622 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1623 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1624 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
1625 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1626 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1627 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1628 // CHECK9-NEXT: store i32 100, i32* [[N]], align 4
1629 // CHECK9-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1630 // CHECK9-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
1631 // CHECK9-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
1632 // CHECK9-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
1633 // CHECK9-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
1634 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
1635 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
1636 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
1637 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
1638 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
1639 // CHECK9-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
1640 // CHECK9-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1641 // CHECK9-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
1642 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP7]], align 8
1643 // CHECK9-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1644 // CHECK9-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
1645 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP9]], align 8
1646 // CHECK9-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1647 // CHECK9-NEXT: store i64 4, i64* [[TMP10]], align 8
1648 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1649 // CHECK9-NEXT: store i8* null, i8** [[TMP11]], align 8
1650 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1651 // CHECK9-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1652 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
1653 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1654 // CHECK9-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
1655 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
1656 // CHECK9-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
1657 // CHECK9-NEXT: store i64 8, i64* [[TMP16]], align 8
1658 // CHECK9-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1659 // CHECK9-NEXT: store i8* null, i8** [[TMP17]], align 8
1660 // CHECK9-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1661 // CHECK9-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
1662 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 8
1663 // CHECK9-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1664 // CHECK9-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
1665 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 8
1666 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1667 // CHECK9-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 8
1668 // CHECK9-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1669 // CHECK9-NEXT: store i8* null, i8** [[TMP23]], align 8
1670 // CHECK9-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1671 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1672 // CHECK9-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1673 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
1674 // CHECK9-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
1675 // CHECK9-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1676 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
1677 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1678 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1679 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1680 // CHECK9-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1681 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
1682 // CHECK9-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
1683 // CHECK9-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 [[TMP30]])
1684 // CHECK9-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73.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)
1685 // CHECK9-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
1686 // CHECK9-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1687 // CHECK9: omp_offload.failed:
1688 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
1689 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
1690 // CHECK9: omp_offload.cont:
1691 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 0
1692 // CHECK9-NEXT: [[TMP33:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1693 // CHECK9-NEXT: [[TMP34:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1694 // CHECK9-NEXT: call void @llvm.stackrestore(i8* [[TMP34]])
1695 // CHECK9-NEXT: ret i32 [[TMP33]]
1696 //
1697 //
1698 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73
1699 // CHECK9-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1700 // CHECK9-NEXT: entry:
1701 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
1702 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1703 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1704 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
1705 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1706 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1707 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
1708 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1709 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1710 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
1711 // CHECK9-NEXT: ret void
1712 //
1713 //
1714 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
1715 // 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]] {
1716 // CHECK9-NEXT: entry:
1717 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1718 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1719 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
1720 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1721 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1722 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1723 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1724 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1725 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1726 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1727 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1728 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1729 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1730 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1731 // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4
1732 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1733 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1734 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
1735 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1736 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1737 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
1738 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1739 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1740 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1741 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1742 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1743 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1744 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1745 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1746 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1747 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
1748 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1749 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1750 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1751 // CHECK9: omp.precond.then:
1752 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1753 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1754 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
1755 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1756 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1757 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1758 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
1759 // 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_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1760 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1761 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1762 // CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1763 // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1764 // CHECK9: cond.true:
1765 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1766 // CHECK9-NEXT: br label [[COND_END:%.*]]
1767 // CHECK9: cond.false:
1768 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1769 // CHECK9-NEXT: br label [[COND_END]]
1770 // CHECK9: cond.end:
1771 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1772 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1773 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1774 // CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1775 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1776 // CHECK9: omp.inner.for.cond:
1777 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1778 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1779 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1780 // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1781 // CHECK9: omp.inner.for.body:
1782 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1783 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
1784 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1785 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
1786 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
1787 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
1788 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
1789 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1790 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1791 // CHECK9: omp.body.continue:
1792 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1793 // CHECK9: omp.inner.for.inc:
1794 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1795 // CHECK9-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
1796 // CHECK9-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1797 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
1798 // CHECK9: omp.inner.for.end:
1799 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1800 // CHECK9: omp.loop.exit:
1801 // CHECK9-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1802 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
1803 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
1804 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
1805 // CHECK9: omp.precond.end:
1806 // CHECK9-NEXT: ret void
1807 //
1808 //
1809 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1810 // CHECK9-SAME: () #[[ATTR4:[0-9]+]] {
1811 // CHECK9-NEXT: entry:
1812 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
1813 // CHECK9-NEXT: ret void
1814 //
1815 //
1816 // CHECK10-LABEL: define {{[^@]+}}@_Z15teams_local_argv
1817 // CHECK10-SAME: () #[[ATTR0:[0-9]+]] {
1818 // CHECK10-NEXT: entry:
1819 // CHECK10-NEXT: [[N:%.*]] = alloca i32, align 4
1820 // CHECK10-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
1821 // CHECK10-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1822 // CHECK10-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
1823 // CHECK10-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1824 // CHECK10-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1825 // CHECK10-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1826 // CHECK10-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
1827 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
1828 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1829 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1830 // CHECK10-NEXT: store i32 100, i32* [[N]], align 4
1831 // CHECK10-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1832 // CHECK10-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
1833 // CHECK10-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
1834 // CHECK10-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
1835 // CHECK10-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
1836 // CHECK10-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
1837 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
1838 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
1839 // CHECK10-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
1840 // CHECK10-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
1841 // CHECK10-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
1842 // CHECK10-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1843 // CHECK10-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
1844 // CHECK10-NEXT: store i64 [[TMP4]], i64* [[TMP7]], align 8
1845 // CHECK10-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1846 // CHECK10-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
1847 // CHECK10-NEXT: store i64 [[TMP4]], i64* [[TMP9]], align 8
1848 // CHECK10-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1849 // CHECK10-NEXT: store i64 4, i64* [[TMP10]], align 8
1850 // CHECK10-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1851 // CHECK10-NEXT: store i8* null, i8** [[TMP11]], align 8
1852 // CHECK10-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1853 // CHECK10-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1854 // CHECK10-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
1855 // CHECK10-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1856 // CHECK10-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
1857 // CHECK10-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
1858 // CHECK10-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
1859 // CHECK10-NEXT: store i64 8, i64* [[TMP16]], align 8
1860 // CHECK10-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1861 // CHECK10-NEXT: store i8* null, i8** [[TMP17]], align 8
1862 // CHECK10-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1863 // CHECK10-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
1864 // CHECK10-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 8
1865 // CHECK10-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1866 // CHECK10-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
1867 // CHECK10-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 8
1868 // CHECK10-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1869 // CHECK10-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 8
1870 // CHECK10-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1871 // CHECK10-NEXT: store i8* null, i8** [[TMP23]], align 8
1872 // CHECK10-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1873 // CHECK10-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1874 // CHECK10-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1875 // CHECK10-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
1876 // CHECK10-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
1877 // CHECK10-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1878 // CHECK10-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
1879 // CHECK10-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1880 // CHECK10-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1881 // CHECK10-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1882 // CHECK10-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1883 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
1884 // CHECK10-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
1885 // CHECK10-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 [[TMP30]])
1886 // CHECK10-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73.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)
1887 // CHECK10-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
1888 // CHECK10-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1889 // CHECK10: omp_offload.failed:
1890 // CHECK10-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
1891 // CHECK10-NEXT: br label [[OMP_OFFLOAD_CONT]]
1892 // CHECK10: omp_offload.cont:
1893 // CHECK10-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 0
1894 // CHECK10-NEXT: [[TMP33:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1895 // CHECK10-NEXT: [[TMP34:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1896 // CHECK10-NEXT: call void @llvm.stackrestore(i8* [[TMP34]])
1897 // CHECK10-NEXT: ret i32 [[TMP33]]
1898 //
1899 //
1900 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73
1901 // CHECK10-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1902 // CHECK10-NEXT: entry:
1903 // CHECK10-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
1904 // CHECK10-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1905 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1906 // CHECK10-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
1907 // CHECK10-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1908 // CHECK10-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1909 // CHECK10-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
1910 // CHECK10-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1911 // CHECK10-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1912 // CHECK10-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
1913 // CHECK10-NEXT: ret void
1914 //
1915 //
1916 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined.
1917 // 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]] {
1918 // CHECK10-NEXT: entry:
1919 // CHECK10-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1920 // CHECK10-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1921 // CHECK10-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
1922 // CHECK10-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1923 // CHECK10-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1924 // CHECK10-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1925 // CHECK10-NEXT: [[TMP:%.*]] = alloca i32, align 4
1926 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1927 // CHECK10-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1928 // CHECK10-NEXT: [[I:%.*]] = alloca i32, align 4
1929 // CHECK10-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1930 // CHECK10-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1931 // CHECK10-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1932 // CHECK10-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1933 // CHECK10-NEXT: [[I3:%.*]] = alloca i32, align 4
1934 // CHECK10-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1935 // CHECK10-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1936 // CHECK10-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
1937 // CHECK10-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1938 // CHECK10-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1939 // CHECK10-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
1940 // CHECK10-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1941 // CHECK10-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1942 // CHECK10-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1943 // CHECK10-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1944 // CHECK10-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1945 // CHECK10-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1946 // CHECK10-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1947 // CHECK10-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1948 // CHECK10-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1949 // CHECK10-NEXT: store i32 0, i32* [[I]], align 4
1950 // CHECK10-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1951 // CHECK10-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1952 // CHECK10-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1953 // CHECK10: omp.precond.then:
1954 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1955 // CHECK10-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1956 // CHECK10-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
1957 // CHECK10-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1958 // CHECK10-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1959 // CHECK10-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1960 // CHECK10-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
1961 // 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_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1962 // CHECK10-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1963 // CHECK10-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1964 // CHECK10-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1965 // CHECK10-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1966 // CHECK10: cond.true:
1967 // CHECK10-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1968 // CHECK10-NEXT: br label [[COND_END:%.*]]
1969 // CHECK10: cond.false:
1970 // CHECK10-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1971 // CHECK10-NEXT: br label [[COND_END]]
1972 // CHECK10: cond.end:
1973 // CHECK10-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1974 // CHECK10-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1975 // CHECK10-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1976 // CHECK10-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1977 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1978 // CHECK10: omp.inner.for.cond:
1979 // CHECK10-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1980 // CHECK10-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1981 // CHECK10-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1982 // CHECK10-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1983 // CHECK10: omp.inner.for.body:
1984 // CHECK10-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1985 // CHECK10-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
1986 // CHECK10-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1987 // CHECK10-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
1988 // CHECK10-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
1989 // CHECK10-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
1990 // CHECK10-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
1991 // CHECK10-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1992 // CHECK10-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1993 // CHECK10: omp.body.continue:
1994 // CHECK10-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1995 // CHECK10: omp.inner.for.inc:
1996 // CHECK10-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1997 // CHECK10-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
1998 // CHECK10-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1999 // CHECK10-NEXT: br label [[OMP_INNER_FOR_COND]]
2000 // CHECK10: omp.inner.for.end:
2001 // CHECK10-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2002 // CHECK10: omp.loop.exit:
2003 // CHECK10-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2004 // CHECK10-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
2005 // CHECK10-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
2006 // CHECK10-NEXT: br label [[OMP_PRECOND_END]]
2007 // CHECK10: omp.precond.end:
2008 // CHECK10-NEXT: ret void
2009 //
2010 //
2011 // CHECK10-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2012 // CHECK10-SAME: () #[[ATTR4:[0-9]+]] {
2013 // CHECK10-NEXT: entry:
2014 // CHECK10-NEXT: call void @__tgt_register_requires(i64 1)
2015 // CHECK10-NEXT: ret void
2016 //
2017 //
2018 // CHECK11-LABEL: define {{[^@]+}}@_Z15teams_local_argv
2019 // CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
2020 // CHECK11-NEXT: entry:
2021 // CHECK11-NEXT: [[N:%.*]] = alloca i32, align 4
2022 // CHECK11-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
2023 // CHECK11-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
2024 // CHECK11-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
2025 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
2026 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
2027 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
2028 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
2029 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
2030 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2031 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2032 // CHECK11-NEXT: store i32 100, i32* [[N]], align 4
2033 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
2034 // CHECK11-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
2035 // CHECK11-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
2036 // CHECK11-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
2037 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
2038 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
2039 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
2040 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
2041 // CHECK11-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
2042 // CHECK11-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
2043 // CHECK11-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2044 // CHECK11-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
2045 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP7]], align 4
2046 // CHECK11-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2047 // CHECK11-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
2048 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP9]], align 4
2049 // CHECK11-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2050 // CHECK11-NEXT: store i64 4, i64* [[TMP10]], align 4
2051 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2052 // CHECK11-NEXT: store i8* null, i8** [[TMP11]], align 4
2053 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2054 // CHECK11-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
2055 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
2056 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2057 // CHECK11-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
2058 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
2059 // CHECK11-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
2060 // CHECK11-NEXT: store i64 4, i64* [[TMP16]], align 4
2061 // CHECK11-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
2062 // CHECK11-NEXT: store i8* null, i8** [[TMP17]], align 4
2063 // CHECK11-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2064 // CHECK11-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
2065 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 4
2066 // CHECK11-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2067 // CHECK11-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
2068 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 4
2069 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
2070 // CHECK11-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 4
2071 // CHECK11-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
2072 // CHECK11-NEXT: store i8* null, i8** [[TMP23]], align 4
2073 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2074 // CHECK11-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2075 // CHECK11-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2076 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
2077 // CHECK11-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
2078 // CHECK11-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2079 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
2080 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2081 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2082 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2083 // CHECK11-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2084 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
2085 // CHECK11-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
2086 // CHECK11-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 [[TMP30]])
2087 // CHECK11-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73.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)
2088 // CHECK11-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
2089 // CHECK11-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2090 // CHECK11: omp_offload.failed:
2091 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
2092 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
2093 // CHECK11: omp_offload.cont:
2094 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 0
2095 // CHECK11-NEXT: [[TMP33:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2096 // CHECK11-NEXT: [[TMP34:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
2097 // CHECK11-NEXT: call void @llvm.stackrestore(i8* [[TMP34]])
2098 // CHECK11-NEXT: ret i32 [[TMP33]]
2099 //
2100 //
2101 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73
2102 // CHECK11-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
2103 // CHECK11-NEXT: entry:
2104 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
2105 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
2106 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
2107 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
2108 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2109 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
2110 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2111 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
2112 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], 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]])
2113 // CHECK11-NEXT: ret void
2114 //
2115 //
2116 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
2117 // 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]] {
2118 // CHECK11-NEXT: entry:
2119 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2120 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2121 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
2122 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
2123 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
2124 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2125 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
2126 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2127 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2128 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
2129 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2130 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2131 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2132 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2133 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
2134 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2135 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2136 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
2137 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2138 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
2139 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
2140 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2141 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
2142 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2143 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
2144 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2145 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2146 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2147 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2148 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2149 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
2150 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2151 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2152 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2153 // CHECK11: omp.precond.then:
2154 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2155 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2156 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
2157 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2158 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2159 // CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2160 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
2161 // 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_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2162 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2163 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2164 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
2165 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2166 // CHECK11: cond.true:
2167 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2168 // CHECK11-NEXT: br label [[COND_END:%.*]]
2169 // CHECK11: cond.false:
2170 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2171 // CHECK11-NEXT: br label [[COND_END]]
2172 // CHECK11: cond.end:
2173 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2174 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2175 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2176 // CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
2177 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2178 // CHECK11: omp.inner.for.cond:
2179 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2180 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2181 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2182 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2183 // CHECK11: omp.inner.for.body:
2184 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2185 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
2186 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2187 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
2188 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
2189 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP17]]
2190 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2191 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2192 // CHECK11: omp.body.continue:
2193 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2194 // CHECK11: omp.inner.for.inc:
2195 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2196 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
2197 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
2198 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
2199 // CHECK11: omp.inner.for.end:
2200 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2201 // CHECK11: omp.loop.exit:
2202 // CHECK11-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2203 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
2204 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
2205 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
2206 // CHECK11: omp.precond.end:
2207 // CHECK11-NEXT: ret void
2208 //
2209 //
2210 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2211 // CHECK11-SAME: () #[[ATTR4:[0-9]+]] {
2212 // CHECK11-NEXT: entry:
2213 // CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
2214 // CHECK11-NEXT: ret void
2215 //
2216 //
2217 // CHECK12-LABEL: define {{[^@]+}}@_Z15teams_local_argv
2218 // CHECK12-SAME: () #[[ATTR0:[0-9]+]] {
2219 // CHECK12-NEXT: entry:
2220 // CHECK12-NEXT: [[N:%.*]] = alloca i32, align 4
2221 // CHECK12-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
2222 // CHECK12-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
2223 // CHECK12-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
2224 // CHECK12-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
2225 // CHECK12-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
2226 // CHECK12-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
2227 // CHECK12-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
2228 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
2229 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2230 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2231 // CHECK12-NEXT: store i32 100, i32* [[N]], align 4
2232 // CHECK12-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
2233 // CHECK12-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
2234 // CHECK12-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
2235 // CHECK12-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
2236 // CHECK12-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
2237 // CHECK12-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
2238 // CHECK12-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
2239 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
2240 // CHECK12-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
2241 // CHECK12-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
2242 // CHECK12-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2243 // CHECK12-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
2244 // CHECK12-NEXT: store i32 [[TMP3]], i32* [[TMP7]], align 4
2245 // CHECK12-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2246 // CHECK12-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
2247 // CHECK12-NEXT: store i32 [[TMP3]], i32* [[TMP9]], align 4
2248 // CHECK12-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2249 // CHECK12-NEXT: store i64 4, i64* [[TMP10]], align 4
2250 // CHECK12-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2251 // CHECK12-NEXT: store i8* null, i8** [[TMP11]], align 4
2252 // CHECK12-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2253 // CHECK12-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
2254 // CHECK12-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
2255 // CHECK12-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2256 // CHECK12-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
2257 // CHECK12-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
2258 // CHECK12-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
2259 // CHECK12-NEXT: store i64 4, i64* [[TMP16]], align 4
2260 // CHECK12-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
2261 // CHECK12-NEXT: store i8* null, i8** [[TMP17]], align 4
2262 // CHECK12-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2263 // CHECK12-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
2264 // CHECK12-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 4
2265 // CHECK12-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2266 // CHECK12-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
2267 // CHECK12-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 4
2268 // CHECK12-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
2269 // CHECK12-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 4
2270 // CHECK12-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
2271 // CHECK12-NEXT: store i8* null, i8** [[TMP23]], align 4
2272 // CHECK12-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2273 // CHECK12-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2274 // CHECK12-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2275 // CHECK12-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
2276 // CHECK12-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
2277 // CHECK12-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2278 // CHECK12-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
2279 // CHECK12-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2280 // CHECK12-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2281 // CHECK12-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2282 // CHECK12-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2283 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
2284 // CHECK12-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
2285 // CHECK12-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 [[TMP30]])
2286 // CHECK12-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73.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)
2287 // CHECK12-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
2288 // CHECK12-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2289 // CHECK12: omp_offload.failed:
2290 // CHECK12-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
2291 // CHECK12-NEXT: br label [[OMP_OFFLOAD_CONT]]
2292 // CHECK12: omp_offload.cont:
2293 // CHECK12-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 0
2294 // CHECK12-NEXT: [[TMP33:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2295 // CHECK12-NEXT: [[TMP34:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
2296 // CHECK12-NEXT: call void @llvm.stackrestore(i8* [[TMP34]])
2297 // CHECK12-NEXT: ret i32 [[TMP33]]
2298 //
2299 //
2300 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73
2301 // CHECK12-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
2302 // CHECK12-NEXT: entry:
2303 // CHECK12-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
2304 // CHECK12-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
2305 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
2306 // CHECK12-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
2307 // CHECK12-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2308 // CHECK12-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
2309 // CHECK12-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2310 // CHECK12-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
2311 // CHECK12-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], 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]])
2312 // CHECK12-NEXT: ret void
2313 //
2314 //
2315 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined.
2316 // 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]] {
2317 // CHECK12-NEXT: entry:
2318 // CHECK12-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2319 // CHECK12-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2320 // CHECK12-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
2321 // CHECK12-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
2322 // CHECK12-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
2323 // CHECK12-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2324 // CHECK12-NEXT: [[TMP:%.*]] = alloca i32, align 4
2325 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2326 // CHECK12-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2327 // CHECK12-NEXT: [[I:%.*]] = alloca i32, align 4
2328 // CHECK12-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2329 // CHECK12-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2330 // CHECK12-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2331 // CHECK12-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2332 // CHECK12-NEXT: [[I3:%.*]] = alloca i32, align 4
2333 // CHECK12-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2334 // CHECK12-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2335 // CHECK12-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
2336 // CHECK12-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2337 // CHECK12-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
2338 // CHECK12-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
2339 // CHECK12-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2340 // CHECK12-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
2341 // CHECK12-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2342 // CHECK12-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
2343 // CHECK12-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2344 // CHECK12-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2345 // CHECK12-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2346 // CHECK12-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2347 // CHECK12-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2348 // CHECK12-NEXT: store i32 0, i32* [[I]], align 4
2349 // CHECK12-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2350 // CHECK12-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2351 // CHECK12-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2352 // CHECK12: omp.precond.then:
2353 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2354 // CHECK12-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2355 // CHECK12-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
2356 // CHECK12-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2357 // CHECK12-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2358 // CHECK12-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2359 // CHECK12-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
2360 // 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_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2361 // CHECK12-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2362 // CHECK12-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2363 // CHECK12-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
2364 // CHECK12-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2365 // CHECK12: cond.true:
2366 // CHECK12-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2367 // CHECK12-NEXT: br label [[COND_END:%.*]]
2368 // CHECK12: cond.false:
2369 // CHECK12-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2370 // CHECK12-NEXT: br label [[COND_END]]
2371 // CHECK12: cond.end:
2372 // CHECK12-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2373 // CHECK12-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2374 // CHECK12-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2375 // CHECK12-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
2376 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2377 // CHECK12: omp.inner.for.cond:
2378 // CHECK12-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2379 // CHECK12-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2380 // CHECK12-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2381 // CHECK12-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2382 // CHECK12: omp.inner.for.body:
2383 // CHECK12-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2384 // CHECK12-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
2385 // CHECK12-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2386 // CHECK12-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
2387 // CHECK12-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
2388 // CHECK12-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP17]]
2389 // CHECK12-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2390 // CHECK12-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2391 // CHECK12: omp.body.continue:
2392 // CHECK12-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2393 // CHECK12: omp.inner.for.inc:
2394 // CHECK12-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2395 // CHECK12-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
2396 // CHECK12-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
2397 // CHECK12-NEXT: br label [[OMP_INNER_FOR_COND]]
2398 // CHECK12: omp.inner.for.end:
2399 // CHECK12-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2400 // CHECK12: omp.loop.exit:
2401 // CHECK12-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2402 // CHECK12-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
2403 // CHECK12-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
2404 // CHECK12-NEXT: br label [[OMP_PRECOND_END]]
2405 // CHECK12: omp.precond.end:
2406 // CHECK12-NEXT: ret void
2407 //
2408 //
2409 // CHECK12-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2410 // CHECK12-SAME: () #[[ATTR4:[0-9]+]] {
2411 // CHECK12-NEXT: entry:
2412 // CHECK12-NEXT: call void @__tgt_register_requires(i64 1)
2413 // CHECK12-NEXT: ret void
2414 //
2415 //
2416 // CHECK17-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2417 // CHECK17-SAME: () #[[ATTR0:[0-9]+]] {
2418 // CHECK17-NEXT: entry:
2419 // CHECK17-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2420 // CHECK17-NEXT: [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
2421 // CHECK17-NEXT: ret i32 [[CALL]]
2422 //
2423 //
2424 // CHECK17-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2425 // CHECK17-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2426 // CHECK17-NEXT: entry:
2427 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2428 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
2429 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
2430 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
2431 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
2432 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2433 // CHECK17-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2434 // CHECK17-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
2435 // CHECK17-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2436 // CHECK17-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
2437 // CHECK17-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
2438 // CHECK17-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2439 // CHECK17-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
2440 // CHECK17-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
2441 // CHECK17-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2442 // CHECK17-NEXT: store i8* null, i8** [[TMP4]], align 8
2443 // CHECK17-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2444 // CHECK17-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2445 // CHECK17-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 123)
2446 // CHECK17-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109.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)
2447 // CHECK17-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
2448 // CHECK17-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2449 // CHECK17: omp_offload.failed:
2450 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
2451 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT]]
2452 // CHECK17: omp_offload.cont:
2453 // CHECK17-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2454 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A2]], i64 0, i64 0
2455 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2456 // CHECK17-NEXT: ret i32 [[TMP9]]
2457 //
2458 //
2459 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109
2460 // CHECK17-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
2461 // CHECK17-NEXT: entry:
2462 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2463 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2464 // CHECK17-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2465 // CHECK17-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
2466 // CHECK17-NEXT: ret void
2467 //
2468 //
2469 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
2470 // CHECK17-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2471 // CHECK17-NEXT: entry:
2472 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2473 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2474 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2475 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2476 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
2477 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2478 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2479 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2480 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2481 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
2482 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2483 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2484 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2485 // CHECK17-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2486 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2487 // CHECK17-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2488 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2489 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2490 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2491 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2492 // CHECK17-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2493 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2494 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2495 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2496 // CHECK17: cond.true:
2497 // CHECK17-NEXT: br label [[COND_END:%.*]]
2498 // CHECK17: cond.false:
2499 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2500 // CHECK17-NEXT: br label [[COND_END]]
2501 // CHECK17: cond.end:
2502 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2503 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2504 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2505 // CHECK17-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2506 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2507 // CHECK17: omp.inner.for.cond:
2508 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2509 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2510 // CHECK17-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2511 // CHECK17-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2512 // CHECK17: omp.inner.for.body:
2513 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2514 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2515 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2516 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2517 // CHECK17-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2518 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
2519 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
2520 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
2521 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2522 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2523 // CHECK17: omp.body.continue:
2524 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2525 // CHECK17: omp.inner.for.inc:
2526 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2527 // CHECK17-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
2528 // CHECK17-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2529 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
2530 // CHECK17: omp.inner.for.end:
2531 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2532 // CHECK17: omp.loop.exit:
2533 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2534 // CHECK17-NEXT: ret void
2535 //
2536 //
2537 // CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2538 // CHECK17-SAME: () #[[ATTR3:[0-9]+]] {
2539 // CHECK17-NEXT: entry:
2540 // CHECK17-NEXT: call void @__tgt_register_requires(i64 1)
2541 // CHECK17-NEXT: ret void
2542 //
2543 //
2544 // CHECK18-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2545 // CHECK18-SAME: () #[[ATTR0:[0-9]+]] {
2546 // CHECK18-NEXT: entry:
2547 // CHECK18-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2548 // CHECK18-NEXT: [[CALL:%.*]] = call signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
2549 // CHECK18-NEXT: ret i32 [[CALL]]
2550 //
2551 //
2552 // CHECK18-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2553 // CHECK18-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2554 // CHECK18-NEXT: entry:
2555 // CHECK18-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2556 // CHECK18-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
2557 // CHECK18-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
2558 // CHECK18-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
2559 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
2560 // CHECK18-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2561 // CHECK18-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2562 // CHECK18-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
2563 // CHECK18-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2564 // CHECK18-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
2565 // CHECK18-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
2566 // CHECK18-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2567 // CHECK18-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
2568 // CHECK18-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
2569 // CHECK18-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2570 // CHECK18-NEXT: store i8* null, i8** [[TMP4]], align 8
2571 // CHECK18-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2572 // CHECK18-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2573 // CHECK18-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 123)
2574 // CHECK18-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109.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)
2575 // CHECK18-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
2576 // CHECK18-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2577 // CHECK18: omp_offload.failed:
2578 // CHECK18-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
2579 // CHECK18-NEXT: br label [[OMP_OFFLOAD_CONT]]
2580 // CHECK18: omp_offload.cont:
2581 // CHECK18-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2582 // CHECK18-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A2]], i64 0, i64 0
2583 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2584 // CHECK18-NEXT: ret i32 [[TMP9]]
2585 //
2586 //
2587 // CHECK18-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109
2588 // CHECK18-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
2589 // CHECK18-NEXT: entry:
2590 // CHECK18-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2591 // CHECK18-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2592 // CHECK18-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2593 // CHECK18-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
2594 // CHECK18-NEXT: ret void
2595 //
2596 //
2597 // CHECK18-LABEL: define {{[^@]+}}@.omp_outlined.
2598 // CHECK18-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2599 // CHECK18-NEXT: entry:
2600 // CHECK18-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2601 // CHECK18-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2602 // CHECK18-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
2603 // CHECK18-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2604 // CHECK18-NEXT: [[TMP:%.*]] = alloca i32, align 4
2605 // CHECK18-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2606 // CHECK18-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2607 // CHECK18-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2608 // CHECK18-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2609 // CHECK18-NEXT: [[I:%.*]] = alloca i32, align 4
2610 // CHECK18-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2611 // CHECK18-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2612 // CHECK18-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
2613 // CHECK18-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
2614 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2615 // CHECK18-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2616 // CHECK18-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2617 // CHECK18-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2618 // CHECK18-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2619 // CHECK18-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2620 // CHECK18-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2621 // CHECK18-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2622 // CHECK18-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2623 // CHECK18-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2624 // CHECK18: cond.true:
2625 // CHECK18-NEXT: br label [[COND_END:%.*]]
2626 // CHECK18: cond.false:
2627 // CHECK18-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2628 // CHECK18-NEXT: br label [[COND_END]]
2629 // CHECK18: cond.end:
2630 // CHECK18-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2631 // CHECK18-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2632 // CHECK18-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2633 // CHECK18-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2634 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2635 // CHECK18: omp.inner.for.cond:
2636 // CHECK18-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2637 // CHECK18-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2638 // CHECK18-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2639 // CHECK18-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2640 // CHECK18: omp.inner.for.body:
2641 // CHECK18-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2642 // CHECK18-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2643 // CHECK18-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2644 // CHECK18-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2645 // CHECK18-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2646 // CHECK18-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
2647 // CHECK18-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
2648 // CHECK18-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
2649 // CHECK18-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2650 // CHECK18-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2651 // CHECK18: omp.body.continue:
2652 // CHECK18-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2653 // CHECK18: omp.inner.for.inc:
2654 // CHECK18-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2655 // CHECK18-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
2656 // CHECK18-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2657 // CHECK18-NEXT: br label [[OMP_INNER_FOR_COND]]
2658 // CHECK18: omp.inner.for.end:
2659 // CHECK18-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2660 // CHECK18: omp.loop.exit:
2661 // CHECK18-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2662 // CHECK18-NEXT: ret void
2663 //
2664 //
2665 // CHECK18-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2666 // CHECK18-SAME: () #[[ATTR3:[0-9]+]] {
2667 // CHECK18-NEXT: entry:
2668 // CHECK18-NEXT: call void @__tgt_register_requires(i64 1)
2669 // CHECK18-NEXT: ret void
2670 //
2671 //
2672 // CHECK19-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2673 // CHECK19-SAME: () #[[ATTR0:[0-9]+]] {
2674 // CHECK19-NEXT: entry:
2675 // CHECK19-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2676 // CHECK19-NEXT: [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
2677 // CHECK19-NEXT: ret i32 [[CALL]]
2678 //
2679 //
2680 // CHECK19-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2681 // CHECK19-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2682 // CHECK19-NEXT: entry:
2683 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2684 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
2685 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
2686 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
2687 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
2688 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2689 // CHECK19-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2690 // CHECK19-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
2691 // CHECK19-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2692 // CHECK19-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
2693 // CHECK19-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
2694 // CHECK19-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2695 // CHECK19-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
2696 // CHECK19-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
2697 // CHECK19-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2698 // CHECK19-NEXT: store i8* null, i8** [[TMP4]], align 4
2699 // CHECK19-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2700 // CHECK19-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2701 // CHECK19-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 123)
2702 // CHECK19-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109.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)
2703 // CHECK19-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
2704 // CHECK19-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2705 // CHECK19: omp_offload.failed:
2706 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
2707 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT]]
2708 // CHECK19: omp_offload.cont:
2709 // CHECK19-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2710 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A2]], i32 0, i32 0
2711 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2712 // CHECK19-NEXT: ret i32 [[TMP9]]
2713 //
2714 //
2715 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109
2716 // CHECK19-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
2717 // CHECK19-NEXT: entry:
2718 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2719 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2720 // CHECK19-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2721 // CHECK19-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
2722 // CHECK19-NEXT: ret void
2723 //
2724 //
2725 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined.
2726 // CHECK19-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2727 // CHECK19-NEXT: entry:
2728 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2729 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2730 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2731 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2732 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
2733 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2734 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2735 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2736 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2737 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
2738 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2739 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2740 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2741 // CHECK19-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2742 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2743 // CHECK19-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2744 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2745 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2746 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2747 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2748 // CHECK19-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2749 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2750 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2751 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2752 // CHECK19: cond.true:
2753 // CHECK19-NEXT: br label [[COND_END:%.*]]
2754 // CHECK19: cond.false:
2755 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2756 // CHECK19-NEXT: br label [[COND_END]]
2757 // CHECK19: cond.end:
2758 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2759 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2760 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2761 // CHECK19-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2762 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2763 // CHECK19: omp.inner.for.cond:
2764 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2765 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2766 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2767 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2768 // CHECK19: omp.inner.for.body:
2769 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2770 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2771 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2772 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2773 // CHECK19-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2774 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
2775 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP9]]
2776 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2777 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2778 // CHECK19: omp.body.continue:
2779 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2780 // CHECK19: omp.inner.for.inc:
2781 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2782 // CHECK19-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
2783 // CHECK19-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2784 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
2785 // CHECK19: omp.inner.for.end:
2786 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2787 // CHECK19: omp.loop.exit:
2788 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2789 // CHECK19-NEXT: ret void
2790 //
2791 //
2792 // CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2793 // CHECK19-SAME: () #[[ATTR3:[0-9]+]] {
2794 // CHECK19-NEXT: entry:
2795 // CHECK19-NEXT: call void @__tgt_register_requires(i64 1)
2796 // CHECK19-NEXT: ret void
2797 //
2798 //
2799 // CHECK20-LABEL: define {{[^@]+}}@_Z21teams_template_structv
2800 // CHECK20-SAME: () #[[ATTR0:[0-9]+]] {
2801 // CHECK20-NEXT: entry:
2802 // CHECK20-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
2803 // CHECK20-NEXT: [[CALL:%.*]] = call i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* nonnull align 4 dereferenceable(496) [[V]])
2804 // CHECK20-NEXT: ret i32 [[CALL]]
2805 //
2806 //
2807 // CHECK20-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
2808 // CHECK20-SAME: (%struct.SS* nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
2809 // CHECK20-NEXT: entry:
2810 // CHECK20-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2811 // CHECK20-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
2812 // CHECK20-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
2813 // CHECK20-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
2814 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
2815 // CHECK20-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2816 // CHECK20-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2817 // CHECK20-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
2818 // CHECK20-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2819 // CHECK20-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
2820 // CHECK20-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
2821 // CHECK20-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2822 // CHECK20-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
2823 // CHECK20-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
2824 // CHECK20-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2825 // CHECK20-NEXT: store i8* null, i8** [[TMP4]], align 4
2826 // CHECK20-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2827 // CHECK20-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2828 // CHECK20-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 123)
2829 // CHECK20-NEXT: [[TMP7:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109.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)
2830 // CHECK20-NEXT: [[TMP8:%.*]] = icmp ne i32 [[TMP7]], 0
2831 // CHECK20-NEXT: br i1 [[TMP8]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2832 // CHECK20: omp_offload.failed:
2833 // CHECK20-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
2834 // CHECK20-NEXT: br label [[OMP_OFFLOAD_CONT]]
2835 // CHECK20: omp_offload.cont:
2836 // CHECK20-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
2837 // CHECK20-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A2]], i32 0, i32 0
2838 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
2839 // CHECK20-NEXT: ret i32 [[TMP9]]
2840 //
2841 //
2842 // CHECK20-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109
2843 // CHECK20-SAME: (%struct.SS* [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
2844 // CHECK20-NEXT: entry:
2845 // CHECK20-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2846 // CHECK20-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2847 // CHECK20-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2848 // CHECK20-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
2849 // CHECK20-NEXT: ret void
2850 //
2851 //
2852 // CHECK20-LABEL: define {{[^@]+}}@.omp_outlined.
2853 // CHECK20-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.SS* [[THIS:%.*]]) #[[ATTR1]] {
2854 // CHECK20-NEXT: entry:
2855 // CHECK20-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2856 // CHECK20-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2857 // CHECK20-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
2858 // CHECK20-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2859 // CHECK20-NEXT: [[TMP:%.*]] = alloca i32, align 4
2860 // CHECK20-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2861 // CHECK20-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2862 // CHECK20-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2863 // CHECK20-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2864 // CHECK20-NEXT: [[I:%.*]] = alloca i32, align 4
2865 // CHECK20-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2866 // CHECK20-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2867 // CHECK20-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
2868 // CHECK20-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
2869 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2870 // CHECK20-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
2871 // CHECK20-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2872 // CHECK20-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2873 // CHECK20-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2874 // CHECK20-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2875 // CHECK20-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2876 // CHECK20-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2877 // CHECK20-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
2878 // CHECK20-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2879 // CHECK20: cond.true:
2880 // CHECK20-NEXT: br label [[COND_END:%.*]]
2881 // CHECK20: cond.false:
2882 // CHECK20-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2883 // CHECK20-NEXT: br label [[COND_END]]
2884 // CHECK20: cond.end:
2885 // CHECK20-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2886 // CHECK20-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2887 // CHECK20-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2888 // CHECK20-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2889 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2890 // CHECK20: omp.inner.for.cond:
2891 // CHECK20-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2892 // CHECK20-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2893 // CHECK20-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2894 // CHECK20-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2895 // CHECK20: omp.inner.for.body:
2896 // CHECK20-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2897 // CHECK20-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2898 // CHECK20-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2899 // CHECK20-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2900 // CHECK20-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
2901 // CHECK20-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
2902 // CHECK20-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP9]]
2903 // CHECK20-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2904 // CHECK20-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2905 // CHECK20: omp.body.continue:
2906 // CHECK20-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2907 // CHECK20: omp.inner.for.inc:
2908 // CHECK20-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2909 // CHECK20-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
2910 // CHECK20-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2911 // CHECK20-NEXT: br label [[OMP_INNER_FOR_COND]]
2912 // CHECK20: omp.inner.for.end:
2913 // CHECK20-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2914 // CHECK20: omp.loop.exit:
2915 // CHECK20-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2916 // CHECK20-NEXT: ret void
2917 //
2918 //
2919 // CHECK20-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2920 // CHECK20-SAME: () #[[ATTR3:[0-9]+]] {
2921 // CHECK20-NEXT: entry:
2922 // CHECK20-NEXT: call void @__tgt_register_requires(i64 1)
2923 // CHECK20-NEXT: ret void
2924 //
2925 //
2926 // CHECK25-LABEL: define {{[^@]+}}@main
2927 // CHECK25-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
2928 // CHECK25-NEXT: entry:
2929 // CHECK25-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
2930 // CHECK25-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
2931 // CHECK25-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
2932 // CHECK25-NEXT: [[N:%.*]] = alloca i32, align 4
2933 // CHECK25-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
2934 // CHECK25-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
2935 // CHECK25-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
2936 // CHECK25-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
2937 // CHECK25-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
2938 // CHECK25-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
2939 // CHECK25-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
2940 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4
2941 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2942 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2943 // CHECK25-NEXT: store i32 0, i32* [[RETVAL]], align 4
2944 // CHECK25-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
2945 // CHECK25-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
2946 // CHECK25-NEXT: store i32 100, i32* [[N]], align 4
2947 // CHECK25-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
2948 // CHECK25-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
2949 // CHECK25-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
2950 // CHECK25-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
2951 // CHECK25-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
2952 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
2953 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
2954 // CHECK25-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
2955 // CHECK25-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
2956 // CHECK25-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
2957 // CHECK25-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
2958 // CHECK25-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2959 // CHECK25-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
2960 // CHECK25-NEXT: store i64 [[TMP4]], i64* [[TMP7]], align 8
2961 // CHECK25-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2962 // CHECK25-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
2963 // CHECK25-NEXT: store i64 [[TMP4]], i64* [[TMP9]], align 8
2964 // CHECK25-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2965 // CHECK25-NEXT: store i64 4, i64* [[TMP10]], align 8
2966 // CHECK25-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2967 // CHECK25-NEXT: store i8* null, i8** [[TMP11]], align 8
2968 // CHECK25-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2969 // CHECK25-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
2970 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
2971 // CHECK25-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2972 // CHECK25-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
2973 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
2974 // CHECK25-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
2975 // CHECK25-NEXT: store i64 8, i64* [[TMP16]], align 8
2976 // CHECK25-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
2977 // CHECK25-NEXT: store i8* null, i8** [[TMP17]], align 8
2978 // CHECK25-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2979 // CHECK25-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
2980 // CHECK25-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 8
2981 // CHECK25-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2982 // CHECK25-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
2983 // CHECK25-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 8
2984 // CHECK25-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
2985 // CHECK25-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 8
2986 // CHECK25-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
2987 // CHECK25-NEXT: store i8* null, i8** [[TMP23]], align 8
2988 // CHECK25-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2989 // CHECK25-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2990 // CHECK25-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2991 // CHECK25-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
2992 // CHECK25-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
2993 // CHECK25-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2994 // CHECK25-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
2995 // CHECK25-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2996 // CHECK25-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2997 // CHECK25-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2998 // CHECK25-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2999 // CHECK25-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
3000 // CHECK25-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
3001 // CHECK25-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 [[TMP30]])
3002 // CHECK25-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162.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)
3003 // CHECK25-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
3004 // CHECK25-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3005 // CHECK25: omp_offload.failed:
3006 // CHECK25-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
3007 // CHECK25-NEXT: br label [[OMP_OFFLOAD_CONT]]
3008 // CHECK25: omp_offload.cont:
3009 // CHECK25-NEXT: [[TMP33:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3010 // CHECK25-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP33]])
3011 // CHECK25-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
3012 // CHECK25-NEXT: [[TMP34:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
3013 // CHECK25-NEXT: call void @llvm.stackrestore(i8* [[TMP34]])
3014 // CHECK25-NEXT: [[TMP35:%.*]] = load i32, i32* [[RETVAL]], align 4
3015 // CHECK25-NEXT: ret i32 [[TMP35]]
3016 //
3017 //
3018 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162
3019 // CHECK25-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
3020 // CHECK25-NEXT: entry:
3021 // CHECK25-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
3022 // CHECK25-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3023 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3024 // CHECK25-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
3025 // CHECK25-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3026 // CHECK25-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3027 // CHECK25-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
3028 // CHECK25-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3029 // CHECK25-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3030 // CHECK25-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
3031 // CHECK25-NEXT: ret void
3032 //
3033 //
3034 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined.
3035 // CHECK25-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]] {
3036 // CHECK25-NEXT: entry:
3037 // CHECK25-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3038 // CHECK25-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3039 // CHECK25-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
3040 // CHECK25-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3041 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3042 // CHECK25-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3043 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4
3044 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3045 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3046 // CHECK25-NEXT: [[I:%.*]] = alloca i32, align 4
3047 // CHECK25-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3048 // CHECK25-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3049 // CHECK25-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3050 // CHECK25-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3051 // CHECK25-NEXT: [[I3:%.*]] = alloca i32, align 4
3052 // CHECK25-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3053 // CHECK25-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3054 // CHECK25-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
3055 // CHECK25-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3056 // CHECK25-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3057 // CHECK25-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
3058 // CHECK25-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3059 // CHECK25-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3060 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3061 // CHECK25-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3062 // CHECK25-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3063 // CHECK25-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3064 // CHECK25-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3065 // CHECK25-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3066 // CHECK25-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3067 // CHECK25-NEXT: store i32 0, i32* [[I]], align 4
3068 // CHECK25-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3069 // CHECK25-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3070 // CHECK25-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3071 // CHECK25: omp.precond.then:
3072 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3073 // CHECK25-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3074 // CHECK25-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
3075 // CHECK25-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3076 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3077 // CHECK25-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3078 // CHECK25-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3079 // CHECK25-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3080 // CHECK25-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3081 // CHECK25-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3082 // CHECK25-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
3083 // CHECK25-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3084 // CHECK25: cond.true:
3085 // CHECK25-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3086 // CHECK25-NEXT: br label [[COND_END:%.*]]
3087 // CHECK25: cond.false:
3088 // CHECK25-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3089 // CHECK25-NEXT: br label [[COND_END]]
3090 // CHECK25: cond.end:
3091 // CHECK25-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3092 // CHECK25-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3093 // CHECK25-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3094 // CHECK25-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
3095 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3096 // CHECK25: omp.inner.for.cond:
3097 // CHECK25-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3098 // CHECK25-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3099 // CHECK25-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
3100 // CHECK25-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3101 // CHECK25: omp.inner.for.body:
3102 // CHECK25-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3103 // CHECK25-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
3104 // CHECK25-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3105 // CHECK25-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
3106 // CHECK25-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
3107 // CHECK25-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
3108 // CHECK25-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
3109 // CHECK25-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3110 // CHECK25-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3111 // CHECK25: omp.body.continue:
3112 // CHECK25-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3113 // CHECK25: omp.inner.for.inc:
3114 // CHECK25-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3115 // CHECK25-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
3116 // CHECK25-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
3117 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND]]
3118 // CHECK25: omp.inner.for.end:
3119 // CHECK25-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3120 // CHECK25: omp.loop.exit:
3121 // CHECK25-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3122 // CHECK25-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
3123 // CHECK25-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
3124 // CHECK25-NEXT: br label [[OMP_PRECOND_END]]
3125 // CHECK25: omp.precond.end:
3126 // CHECK25-NEXT: ret void
3127 //
3128 //
3129 // CHECK25-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
3130 // CHECK25-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
3131 // CHECK25-NEXT: entry:
3132 // CHECK25-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3133 // CHECK25-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
3134 // CHECK25-NEXT: [[TE:%.*]] = alloca i32, align 4
3135 // CHECK25-NEXT: [[TH:%.*]] = alloca i32, align 4
3136 // CHECK25-NEXT: [[TE_CASTED:%.*]] = alloca i64, align 8
3137 // CHECK25-NEXT: [[TH_CASTED:%.*]] = alloca i64, align 8
3138 // CHECK25-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
3139 // CHECK25-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
3140 // CHECK25-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
3141 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4
3142 // CHECK25-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3143 // CHECK25-NEXT: store i32 0, i32* [[TE]], align 4
3144 // CHECK25-NEXT: store i32 128, i32* [[TH]], align 4
3145 // CHECK25-NEXT: [[TMP0:%.*]] = load i32, i32* [[TE]], align 4
3146 // CHECK25-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_CASTED]] to i32*
3147 // CHECK25-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4
3148 // CHECK25-NEXT: [[TMP1:%.*]] = load i64, i64* [[TE_CASTED]], align 8
3149 // CHECK25-NEXT: [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
3150 // CHECK25-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_CASTED]] to i32*
3151 // CHECK25-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
3152 // CHECK25-NEXT: [[TMP3:%.*]] = load i64, i64* [[TH_CASTED]], align 8
3153 // CHECK25-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3154 // CHECK25-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
3155 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[TMP5]], align 8
3156 // CHECK25-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3157 // CHECK25-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
3158 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[TMP7]], align 8
3159 // CHECK25-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3160 // CHECK25-NEXT: store i8* null, i8** [[TMP8]], align 8
3161 // CHECK25-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3162 // CHECK25-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
3163 // CHECK25-NEXT: store i64 [[TMP3]], i64* [[TMP10]], align 8
3164 // CHECK25-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3165 // CHECK25-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
3166 // CHECK25-NEXT: store i64 [[TMP3]], i64* [[TMP12]], align 8
3167 // CHECK25-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3168 // CHECK25-NEXT: store i8* null, i8** [[TMP13]], align 8
3169 // CHECK25-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3170 // CHECK25-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [10 x i32]**
3171 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP15]], align 8
3172 // CHECK25-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3173 // CHECK25-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [10 x i32]**
3174 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP17]], align 8
3175 // CHECK25-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3176 // CHECK25-NEXT: store i8* null, i8** [[TMP18]], align 8
3177 // CHECK25-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3178 // CHECK25-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3179 // CHECK25-NEXT: [[TMP21:%.*]] = load i32, i32* [[TE]], align 4
3180 // CHECK25-NEXT: [[TMP22:%.*]] = load i32, i32* [[TH]], align 4
3181 // CHECK25-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
3182 // CHECK25-NEXT: [[TMP23:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151.region_id, i32 3, i8** [[TMP19]], i8** [[TMP20]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.2, i32 0, i32 0), i8** null, i8** null, i32 [[TMP21]], i32 [[TMP22]])
3183 // CHECK25-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
3184 // CHECK25-NEXT: br i1 [[TMP24]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3185 // CHECK25: omp_offload.failed:
3186 // CHECK25-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[A]]) #[[ATTR3]]
3187 // CHECK25-NEXT: br label [[OMP_OFFLOAD_CONT]]
3188 // CHECK25: omp_offload.cont:
3189 // CHECK25-NEXT: ret i32 0
3190 //
3191 //
3192 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151
3193 // CHECK25-SAME: (i64 [[TE:%.*]], i64 [[TH:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3194 // CHECK25-NEXT: entry:
3195 // CHECK25-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8
3196 // CHECK25-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8
3197 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
3198 // CHECK25-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
3199 // CHECK25-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8
3200 // CHECK25-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8
3201 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
3202 // CHECK25-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_ADDR]] to i32*
3203 // CHECK25-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_ADDR]] to i32*
3204 // CHECK25-NEXT: [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
3205 // CHECK25-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
3206 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 8
3207 // CHECK25-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
3208 // CHECK25-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]])
3209 // CHECK25-NEXT: ret void
3210 //
3211 //
3212 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..1
3213 // CHECK25-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3214 // CHECK25-NEXT: entry:
3215 // CHECK25-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3216 // CHECK25-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3217 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
3218 // CHECK25-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3219 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4
3220 // CHECK25-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3221 // CHECK25-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3222 // CHECK25-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3223 // CHECK25-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3224 // CHECK25-NEXT: [[I:%.*]] = alloca i32, align 4
3225 // CHECK25-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3226 // CHECK25-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3227 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
3228 // CHECK25-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
3229 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3230 // CHECK25-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
3231 // CHECK25-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3232 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3233 // CHECK25-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3234 // CHECK25-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3235 // CHECK25-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3236 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3237 // CHECK25-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
3238 // CHECK25-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3239 // CHECK25: cond.true:
3240 // CHECK25-NEXT: br label [[COND_END:%.*]]
3241 // CHECK25: cond.false:
3242 // CHECK25-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3243 // CHECK25-NEXT: br label [[COND_END]]
3244 // CHECK25: cond.end:
3245 // CHECK25-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3246 // CHECK25-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3247 // CHECK25-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3248 // CHECK25-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3249 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3250 // CHECK25: omp.inner.for.cond:
3251 // CHECK25-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3252 // CHECK25-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3253 // CHECK25-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3254 // CHECK25-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3255 // CHECK25: omp.inner.for.body:
3256 // CHECK25-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3257 // CHECK25-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
3258 // CHECK25-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3259 // CHECK25-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3260 // CHECK25-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
3261 // CHECK25-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
3262 // CHECK25-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
3263 // CHECK25-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3264 // CHECK25-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3265 // CHECK25: omp.body.continue:
3266 // CHECK25-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3267 // CHECK25: omp.inner.for.inc:
3268 // CHECK25-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3269 // CHECK25-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
3270 // CHECK25-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
3271 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND]]
3272 // CHECK25: omp.inner.for.end:
3273 // CHECK25-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3274 // CHECK25: omp.loop.exit:
3275 // CHECK25-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3276 // CHECK25-NEXT: ret void
3277 //
3278 //
3279 // CHECK25-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3280 // CHECK25-SAME: () #[[ATTR5:[0-9]+]] {
3281 // CHECK25-NEXT: entry:
3282 // CHECK25-NEXT: call void @__tgt_register_requires(i64 1)
3283 // CHECK25-NEXT: ret void
3284 //
3285 //
3286 // CHECK26-LABEL: define {{[^@]+}}@main
3287 // CHECK26-SAME: (i32 signext [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
3288 // CHECK26-NEXT: entry:
3289 // CHECK26-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
3290 // CHECK26-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3291 // CHECK26-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
3292 // CHECK26-NEXT: [[N:%.*]] = alloca i32, align 4
3293 // CHECK26-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
3294 // CHECK26-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
3295 // CHECK26-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
3296 // CHECK26-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
3297 // CHECK26-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
3298 // CHECK26-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
3299 // CHECK26-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
3300 // CHECK26-NEXT: [[TMP:%.*]] = alloca i32, align 4
3301 // CHECK26-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3302 // CHECK26-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3303 // CHECK26-NEXT: store i32 0, i32* [[RETVAL]], align 4
3304 // CHECK26-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3305 // CHECK26-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
3306 // CHECK26-NEXT: store i32 100, i32* [[N]], align 4
3307 // CHECK26-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
3308 // CHECK26-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
3309 // CHECK26-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
3310 // CHECK26-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
3311 // CHECK26-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
3312 // CHECK26-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
3313 // CHECK26-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
3314 // CHECK26-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
3315 // CHECK26-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
3316 // CHECK26-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
3317 // CHECK26-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
3318 // CHECK26-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3319 // CHECK26-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
3320 // CHECK26-NEXT: store i64 [[TMP4]], i64* [[TMP7]], align 8
3321 // CHECK26-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3322 // CHECK26-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
3323 // CHECK26-NEXT: store i64 [[TMP4]], i64* [[TMP9]], align 8
3324 // CHECK26-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3325 // CHECK26-NEXT: store i64 4, i64* [[TMP10]], align 8
3326 // CHECK26-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3327 // CHECK26-NEXT: store i8* null, i8** [[TMP11]], align 8
3328 // CHECK26-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3329 // CHECK26-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
3330 // CHECK26-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
3331 // CHECK26-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3332 // CHECK26-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
3333 // CHECK26-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
3334 // CHECK26-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
3335 // CHECK26-NEXT: store i64 8, i64* [[TMP16]], align 8
3336 // CHECK26-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3337 // CHECK26-NEXT: store i8* null, i8** [[TMP17]], align 8
3338 // CHECK26-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3339 // CHECK26-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
3340 // CHECK26-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 8
3341 // CHECK26-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3342 // CHECK26-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
3343 // CHECK26-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 8
3344 // CHECK26-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
3345 // CHECK26-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 8
3346 // CHECK26-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3347 // CHECK26-NEXT: store i8* null, i8** [[TMP23]], align 8
3348 // CHECK26-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3349 // CHECK26-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3350 // CHECK26-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3351 // CHECK26-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
3352 // CHECK26-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
3353 // CHECK26-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3354 // CHECK26-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
3355 // CHECK26-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3356 // CHECK26-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3357 // CHECK26-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3358 // CHECK26-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3359 // CHECK26-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
3360 // CHECK26-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
3361 // CHECK26-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 [[TMP30]])
3362 // CHECK26-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162.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)
3363 // CHECK26-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
3364 // CHECK26-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3365 // CHECK26: omp_offload.failed:
3366 // CHECK26-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
3367 // CHECK26-NEXT: br label [[OMP_OFFLOAD_CONT]]
3368 // CHECK26: omp_offload.cont:
3369 // CHECK26-NEXT: [[TMP33:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3370 // CHECK26-NEXT: [[CALL:%.*]] = call signext i32 @_Z5tmainIiLi10EEiT_(i32 signext [[TMP33]])
3371 // CHECK26-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
3372 // CHECK26-NEXT: [[TMP34:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
3373 // CHECK26-NEXT: call void @llvm.stackrestore(i8* [[TMP34]])
3374 // CHECK26-NEXT: [[TMP35:%.*]] = load i32, i32* [[RETVAL]], align 4
3375 // CHECK26-NEXT: ret i32 [[TMP35]]
3376 //
3377 //
3378 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162
3379 // CHECK26-SAME: (i64 [[N:%.*]], i64 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
3380 // CHECK26-NEXT: entry:
3381 // CHECK26-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
3382 // CHECK26-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3383 // CHECK26-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3384 // CHECK26-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
3385 // CHECK26-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3386 // CHECK26-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3387 // CHECK26-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
3388 // CHECK26-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3389 // CHECK26-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3390 // CHECK26-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
3391 // CHECK26-NEXT: ret void
3392 //
3393 //
3394 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined.
3395 // CHECK26-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]] {
3396 // CHECK26-NEXT: entry:
3397 // CHECK26-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3398 // CHECK26-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3399 // CHECK26-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
3400 // CHECK26-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
3401 // CHECK26-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
3402 // CHECK26-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3403 // CHECK26-NEXT: [[TMP:%.*]] = alloca i32, align 4
3404 // CHECK26-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3405 // CHECK26-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3406 // CHECK26-NEXT: [[I:%.*]] = alloca i32, align 4
3407 // CHECK26-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3408 // CHECK26-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3409 // CHECK26-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3410 // CHECK26-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3411 // CHECK26-NEXT: [[I3:%.*]] = alloca i32, align 4
3412 // CHECK26-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3413 // CHECK26-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3414 // CHECK26-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
3415 // CHECK26-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
3416 // CHECK26-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
3417 // CHECK26-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
3418 // CHECK26-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
3419 // CHECK26-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
3420 // CHECK26-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3421 // CHECK26-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3422 // CHECK26-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3423 // CHECK26-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3424 // CHECK26-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3425 // CHECK26-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3426 // CHECK26-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3427 // CHECK26-NEXT: store i32 0, i32* [[I]], align 4
3428 // CHECK26-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3429 // CHECK26-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3430 // CHECK26-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3431 // CHECK26: omp.precond.then:
3432 // CHECK26-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3433 // CHECK26-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3434 // CHECK26-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
3435 // CHECK26-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3436 // CHECK26-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3437 // CHECK26-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3438 // CHECK26-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3439 // CHECK26-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3440 // CHECK26-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3441 // CHECK26-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3442 // CHECK26-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
3443 // CHECK26-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3444 // CHECK26: cond.true:
3445 // CHECK26-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3446 // CHECK26-NEXT: br label [[COND_END:%.*]]
3447 // CHECK26: cond.false:
3448 // CHECK26-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3449 // CHECK26-NEXT: br label [[COND_END]]
3450 // CHECK26: cond.end:
3451 // CHECK26-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3452 // CHECK26-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3453 // CHECK26-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3454 // CHECK26-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
3455 // CHECK26-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3456 // CHECK26: omp.inner.for.cond:
3457 // CHECK26-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3458 // CHECK26-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3459 // CHECK26-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
3460 // CHECK26-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3461 // CHECK26: omp.inner.for.body:
3462 // CHECK26-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3463 // CHECK26-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
3464 // CHECK26-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3465 // CHECK26-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
3466 // CHECK26-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
3467 // CHECK26-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
3468 // CHECK26-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
3469 // CHECK26-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3470 // CHECK26-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3471 // CHECK26: omp.body.continue:
3472 // CHECK26-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3473 // CHECK26: omp.inner.for.inc:
3474 // CHECK26-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3475 // CHECK26-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
3476 // CHECK26-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
3477 // CHECK26-NEXT: br label [[OMP_INNER_FOR_COND]]
3478 // CHECK26: omp.inner.for.end:
3479 // CHECK26-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3480 // CHECK26: omp.loop.exit:
3481 // CHECK26-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3482 // CHECK26-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
3483 // CHECK26-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
3484 // CHECK26-NEXT: br label [[OMP_PRECOND_END]]
3485 // CHECK26: omp.precond.end:
3486 // CHECK26-NEXT: ret void
3487 //
3488 //
3489 // CHECK26-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
3490 // CHECK26-SAME: (i32 signext [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
3491 // CHECK26-NEXT: entry:
3492 // CHECK26-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3493 // CHECK26-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
3494 // CHECK26-NEXT: [[TE:%.*]] = alloca i32, align 4
3495 // CHECK26-NEXT: [[TH:%.*]] = alloca i32, align 4
3496 // CHECK26-NEXT: [[TE_CASTED:%.*]] = alloca i64, align 8
3497 // CHECK26-NEXT: [[TH_CASTED:%.*]] = alloca i64, align 8
3498 // CHECK26-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
3499 // CHECK26-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
3500 // CHECK26-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
3501 // CHECK26-NEXT: [[TMP:%.*]] = alloca i32, align 4
3502 // CHECK26-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3503 // CHECK26-NEXT: store i32 0, i32* [[TE]], align 4
3504 // CHECK26-NEXT: store i32 128, i32* [[TH]], align 4
3505 // CHECK26-NEXT: [[TMP0:%.*]] = load i32, i32* [[TE]], align 4
3506 // CHECK26-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_CASTED]] to i32*
3507 // CHECK26-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4
3508 // CHECK26-NEXT: [[TMP1:%.*]] = load i64, i64* [[TE_CASTED]], align 8
3509 // CHECK26-NEXT: [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
3510 // CHECK26-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_CASTED]] to i32*
3511 // CHECK26-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
3512 // CHECK26-NEXT: [[TMP3:%.*]] = load i64, i64* [[TH_CASTED]], align 8
3513 // CHECK26-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3514 // CHECK26-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
3515 // CHECK26-NEXT: store i64 [[TMP1]], i64* [[TMP5]], align 8
3516 // CHECK26-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3517 // CHECK26-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
3518 // CHECK26-NEXT: store i64 [[TMP1]], i64* [[TMP7]], align 8
3519 // CHECK26-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3520 // CHECK26-NEXT: store i8* null, i8** [[TMP8]], align 8
3521 // CHECK26-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3522 // CHECK26-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
3523 // CHECK26-NEXT: store i64 [[TMP3]], i64* [[TMP10]], align 8
3524 // CHECK26-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3525 // CHECK26-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
3526 // CHECK26-NEXT: store i64 [[TMP3]], i64* [[TMP12]], align 8
3527 // CHECK26-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
3528 // CHECK26-NEXT: store i8* null, i8** [[TMP13]], align 8
3529 // CHECK26-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3530 // CHECK26-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [10 x i32]**
3531 // CHECK26-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP15]], align 8
3532 // CHECK26-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3533 // CHECK26-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [10 x i32]**
3534 // CHECK26-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP17]], align 8
3535 // CHECK26-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
3536 // CHECK26-NEXT: store i8* null, i8** [[TMP18]], align 8
3537 // CHECK26-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3538 // CHECK26-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3539 // CHECK26-NEXT: [[TMP21:%.*]] = load i32, i32* [[TE]], align 4
3540 // CHECK26-NEXT: [[TMP22:%.*]] = load i32, i32* [[TH]], align 4
3541 // CHECK26-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
3542 // CHECK26-NEXT: [[TMP23:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151.region_id, i32 3, i8** [[TMP19]], i8** [[TMP20]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.2, i32 0, i32 0), i8** null, i8** null, i32 [[TMP21]], i32 [[TMP22]])
3543 // CHECK26-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
3544 // CHECK26-NEXT: br i1 [[TMP24]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3545 // CHECK26: omp_offload.failed:
3546 // CHECK26-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[A]]) #[[ATTR3]]
3547 // CHECK26-NEXT: br label [[OMP_OFFLOAD_CONT]]
3548 // CHECK26: omp_offload.cont:
3549 // CHECK26-NEXT: ret i32 0
3550 //
3551 //
3552 // CHECK26-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151
3553 // CHECK26-SAME: (i64 [[TE:%.*]], i64 [[TH:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3554 // CHECK26-NEXT: entry:
3555 // CHECK26-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8
3556 // CHECK26-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8
3557 // CHECK26-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
3558 // CHECK26-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
3559 // CHECK26-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8
3560 // CHECK26-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8
3561 // CHECK26-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
3562 // CHECK26-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_ADDR]] to i32*
3563 // CHECK26-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_ADDR]] to i32*
3564 // CHECK26-NEXT: [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
3565 // CHECK26-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 8
3566 // CHECK26-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 8
3567 // CHECK26-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
3568 // CHECK26-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]])
3569 // CHECK26-NEXT: ret void
3570 //
3571 //
3572 // CHECK26-LABEL: define {{[^@]+}}@.omp_outlined..1
3573 // CHECK26-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3574 // CHECK26-NEXT: entry:
3575 // CHECK26-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3576 // CHECK26-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3577 // CHECK26-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
3578 // CHECK26-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3579 // CHECK26-NEXT: [[TMP:%.*]] = alloca i32, align 4
3580 // CHECK26-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3581 // CHECK26-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3582 // CHECK26-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3583 // CHECK26-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3584 // CHECK26-NEXT: [[I:%.*]] = alloca i32, align 4
3585 // CHECK26-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3586 // CHECK26-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3587 // CHECK26-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
3588 // CHECK26-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
3589 // CHECK26-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3590 // CHECK26-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
3591 // CHECK26-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3592 // CHECK26-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3593 // CHECK26-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3594 // CHECK26-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3595 // CHECK26-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3596 // CHECK26-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3597 // CHECK26-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
3598 // CHECK26-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3599 // CHECK26: cond.true:
3600 // CHECK26-NEXT: br label [[COND_END:%.*]]
3601 // CHECK26: cond.false:
3602 // CHECK26-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3603 // CHECK26-NEXT: br label [[COND_END]]
3604 // CHECK26: cond.end:
3605 // CHECK26-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3606 // CHECK26-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3607 // CHECK26-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3608 // CHECK26-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3609 // CHECK26-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3610 // CHECK26: omp.inner.for.cond:
3611 // CHECK26-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3612 // CHECK26-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3613 // CHECK26-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3614 // CHECK26-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3615 // CHECK26: omp.inner.for.body:
3616 // CHECK26-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3617 // CHECK26-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
3618 // CHECK26-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3619 // CHECK26-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3620 // CHECK26-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
3621 // CHECK26-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
3622 // CHECK26-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
3623 // CHECK26-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3624 // CHECK26-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3625 // CHECK26: omp.body.continue:
3626 // CHECK26-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3627 // CHECK26: omp.inner.for.inc:
3628 // CHECK26-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3629 // CHECK26-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
3630 // CHECK26-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
3631 // CHECK26-NEXT: br label [[OMP_INNER_FOR_COND]]
3632 // CHECK26: omp.inner.for.end:
3633 // CHECK26-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3634 // CHECK26: omp.loop.exit:
3635 // CHECK26-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3636 // CHECK26-NEXT: ret void
3637 //
3638 //
3639 // CHECK26-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3640 // CHECK26-SAME: () #[[ATTR5:[0-9]+]] {
3641 // CHECK26-NEXT: entry:
3642 // CHECK26-NEXT: call void @__tgt_register_requires(i64 1)
3643 // CHECK26-NEXT: ret void
3644 //
3645 //
3646 // CHECK27-LABEL: define {{[^@]+}}@main
3647 // CHECK27-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
3648 // CHECK27-NEXT: entry:
3649 // CHECK27-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
3650 // CHECK27-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3651 // CHECK27-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
3652 // CHECK27-NEXT: [[N:%.*]] = alloca i32, align 4
3653 // CHECK27-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
3654 // CHECK27-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
3655 // CHECK27-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
3656 // CHECK27-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
3657 // CHECK27-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
3658 // CHECK27-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
3659 // CHECK27-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
3660 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4
3661 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3662 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3663 // CHECK27-NEXT: store i32 0, i32* [[RETVAL]], align 4
3664 // CHECK27-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3665 // CHECK27-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
3666 // CHECK27-NEXT: store i32 100, i32* [[N]], align 4
3667 // CHECK27-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
3668 // CHECK27-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
3669 // CHECK27-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
3670 // CHECK27-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
3671 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
3672 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
3673 // CHECK27-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
3674 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
3675 // CHECK27-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
3676 // CHECK27-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
3677 // CHECK27-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3678 // CHECK27-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
3679 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[TMP7]], align 4
3680 // CHECK27-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3681 // CHECK27-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
3682 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[TMP9]], align 4
3683 // CHECK27-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3684 // CHECK27-NEXT: store i64 4, i64* [[TMP10]], align 4
3685 // CHECK27-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3686 // CHECK27-NEXT: store i8* null, i8** [[TMP11]], align 4
3687 // CHECK27-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3688 // CHECK27-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
3689 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
3690 // CHECK27-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3691 // CHECK27-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
3692 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
3693 // CHECK27-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
3694 // CHECK27-NEXT: store i64 4, i64* [[TMP16]], align 4
3695 // CHECK27-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3696 // CHECK27-NEXT: store i8* null, i8** [[TMP17]], align 4
3697 // CHECK27-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3698 // CHECK27-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
3699 // CHECK27-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 4
3700 // CHECK27-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3701 // CHECK27-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
3702 // CHECK27-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 4
3703 // CHECK27-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
3704 // CHECK27-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 4
3705 // CHECK27-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3706 // CHECK27-NEXT: store i8* null, i8** [[TMP23]], align 4
3707 // CHECK27-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3708 // CHECK27-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3709 // CHECK27-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3710 // CHECK27-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
3711 // CHECK27-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
3712 // CHECK27-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3713 // CHECK27-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
3714 // CHECK27-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3715 // CHECK27-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3716 // CHECK27-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3717 // CHECK27-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3718 // CHECK27-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
3719 // CHECK27-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
3720 // CHECK27-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 [[TMP30]])
3721 // CHECK27-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162.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)
3722 // CHECK27-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
3723 // CHECK27-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3724 // CHECK27: omp_offload.failed:
3725 // CHECK27-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
3726 // CHECK27-NEXT: br label [[OMP_OFFLOAD_CONT]]
3727 // CHECK27: omp_offload.cont:
3728 // CHECK27-NEXT: [[TMP33:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3729 // CHECK27-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP33]])
3730 // CHECK27-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
3731 // CHECK27-NEXT: [[TMP34:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
3732 // CHECK27-NEXT: call void @llvm.stackrestore(i8* [[TMP34]])
3733 // CHECK27-NEXT: [[TMP35:%.*]] = load i32, i32* [[RETVAL]], align 4
3734 // CHECK27-NEXT: ret i32 [[TMP35]]
3735 //
3736 //
3737 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162
3738 // CHECK27-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
3739 // CHECK27-NEXT: entry:
3740 // CHECK27-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
3741 // CHECK27-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3742 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3743 // CHECK27-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
3744 // CHECK27-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3745 // CHECK27-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3746 // CHECK27-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3747 // CHECK27-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3748 // CHECK27-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], 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]])
3749 // CHECK27-NEXT: ret void
3750 //
3751 //
3752 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined.
3753 // CHECK27-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]] {
3754 // CHECK27-NEXT: entry:
3755 // CHECK27-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3756 // CHECK27-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3757 // CHECK27-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
3758 // CHECK27-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3759 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3760 // CHECK27-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3761 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4
3762 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3763 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3764 // CHECK27-NEXT: [[I:%.*]] = alloca i32, align 4
3765 // CHECK27-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3766 // CHECK27-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3767 // CHECK27-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3768 // CHECK27-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3769 // CHECK27-NEXT: [[I3:%.*]] = alloca i32, align 4
3770 // CHECK27-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3771 // CHECK27-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3772 // CHECK27-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
3773 // CHECK27-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3774 // CHECK27-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3775 // CHECK27-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
3776 // CHECK27-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3777 // CHECK27-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3778 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3779 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3780 // CHECK27-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3781 // CHECK27-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3782 // CHECK27-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3783 // CHECK27-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3784 // CHECK27-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3785 // CHECK27-NEXT: store i32 0, i32* [[I]], align 4
3786 // CHECK27-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3787 // CHECK27-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3788 // CHECK27-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3789 // CHECK27: omp.precond.then:
3790 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3791 // CHECK27-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3792 // CHECK27-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
3793 // CHECK27-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3794 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3795 // CHECK27-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3796 // CHECK27-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3797 // CHECK27-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3798 // CHECK27-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3799 // CHECK27-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3800 // CHECK27-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
3801 // CHECK27-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3802 // CHECK27: cond.true:
3803 // CHECK27-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3804 // CHECK27-NEXT: br label [[COND_END:%.*]]
3805 // CHECK27: cond.false:
3806 // CHECK27-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3807 // CHECK27-NEXT: br label [[COND_END]]
3808 // CHECK27: cond.end:
3809 // CHECK27-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3810 // CHECK27-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3811 // CHECK27-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3812 // CHECK27-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
3813 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3814 // CHECK27: omp.inner.for.cond:
3815 // CHECK27-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3816 // CHECK27-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3817 // CHECK27-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
3818 // CHECK27-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3819 // CHECK27: omp.inner.for.body:
3820 // CHECK27-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3821 // CHECK27-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
3822 // CHECK27-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3823 // CHECK27-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
3824 // CHECK27-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
3825 // CHECK27-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP17]]
3826 // CHECK27-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3827 // CHECK27-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3828 // CHECK27: omp.body.continue:
3829 // CHECK27-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3830 // CHECK27: omp.inner.for.inc:
3831 // CHECK27-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3832 // CHECK27-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
3833 // CHECK27-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
3834 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND]]
3835 // CHECK27: omp.inner.for.end:
3836 // CHECK27-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3837 // CHECK27: omp.loop.exit:
3838 // CHECK27-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3839 // CHECK27-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
3840 // CHECK27-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
3841 // CHECK27-NEXT: br label [[OMP_PRECOND_END]]
3842 // CHECK27: omp.precond.end:
3843 // CHECK27-NEXT: ret void
3844 //
3845 //
3846 // CHECK27-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
3847 // CHECK27-SAME: (i32 [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
3848 // CHECK27-NEXT: entry:
3849 // CHECK27-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3850 // CHECK27-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
3851 // CHECK27-NEXT: [[TE:%.*]] = alloca i32, align 4
3852 // CHECK27-NEXT: [[TH:%.*]] = alloca i32, align 4
3853 // CHECK27-NEXT: [[TE_CASTED:%.*]] = alloca i32, align 4
3854 // CHECK27-NEXT: [[TH_CASTED:%.*]] = alloca i32, align 4
3855 // CHECK27-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
3856 // CHECK27-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
3857 // CHECK27-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
3858 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4
3859 // CHECK27-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3860 // CHECK27-NEXT: store i32 0, i32* [[TE]], align 4
3861 // CHECK27-NEXT: store i32 128, i32* [[TH]], align 4
3862 // CHECK27-NEXT: [[TMP0:%.*]] = load i32, i32* [[TE]], align 4
3863 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[TE_CASTED]], align 4
3864 // CHECK27-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE_CASTED]], align 4
3865 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
3866 // CHECK27-NEXT: store i32 [[TMP2]], i32* [[TH_CASTED]], align 4
3867 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_CASTED]], align 4
3868 // CHECK27-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3869 // CHECK27-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32*
3870 // CHECK27-NEXT: store i32 [[TMP1]], i32* [[TMP5]], align 4
3871 // CHECK27-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3872 // CHECK27-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
3873 // CHECK27-NEXT: store i32 [[TMP1]], i32* [[TMP7]], align 4
3874 // CHECK27-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3875 // CHECK27-NEXT: store i8* null, i8** [[TMP8]], align 4
3876 // CHECK27-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3877 // CHECK27-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
3878 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
3879 // CHECK27-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3880 // CHECK27-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
3881 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[TMP12]], align 4
3882 // CHECK27-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3883 // CHECK27-NEXT: store i8* null, i8** [[TMP13]], align 4
3884 // CHECK27-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3885 // CHECK27-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [10 x i32]**
3886 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP15]], align 4
3887 // CHECK27-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3888 // CHECK27-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [10 x i32]**
3889 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP17]], align 4
3890 // CHECK27-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3891 // CHECK27-NEXT: store i8* null, i8** [[TMP18]], align 4
3892 // CHECK27-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3893 // CHECK27-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3894 // CHECK27-NEXT: [[TMP21:%.*]] = load i32, i32* [[TE]], align 4
3895 // CHECK27-NEXT: [[TMP22:%.*]] = load i32, i32* [[TH]], align 4
3896 // CHECK27-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
3897 // CHECK27-NEXT: [[TMP23:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151.region_id, i32 3, i8** [[TMP19]], i8** [[TMP20]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.2, i32 0, i32 0), i8** null, i8** null, i32 [[TMP21]], i32 [[TMP22]])
3898 // CHECK27-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
3899 // CHECK27-NEXT: br i1 [[TMP24]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3900 // CHECK27: omp_offload.failed:
3901 // CHECK27-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[A]]) #[[ATTR3]]
3902 // CHECK27-NEXT: br label [[OMP_OFFLOAD_CONT]]
3903 // CHECK27: omp_offload.cont:
3904 // CHECK27-NEXT: ret i32 0
3905 //
3906 //
3907 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151
3908 // CHECK27-SAME: (i32 [[TE:%.*]], i32 [[TH:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3909 // CHECK27-NEXT: entry:
3910 // CHECK27-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4
3911 // CHECK27-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4
3912 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
3913 // CHECK27-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
3914 // CHECK27-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4
3915 // CHECK27-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4
3916 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
3917 // CHECK27-NEXT: [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
3918 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4
3919 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4
3920 // CHECK27-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
3921 // CHECK27-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]])
3922 // CHECK27-NEXT: ret void
3923 //
3924 //
3925 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..1
3926 // CHECK27-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
3927 // CHECK27-NEXT: entry:
3928 // CHECK27-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3929 // CHECK27-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3930 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
3931 // CHECK27-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3932 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4
3933 // CHECK27-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3934 // CHECK27-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3935 // CHECK27-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3936 // CHECK27-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3937 // CHECK27-NEXT: [[I:%.*]] = alloca i32, align 4
3938 // CHECK27-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3939 // CHECK27-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3940 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
3941 // CHECK27-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
3942 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3943 // CHECK27-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
3944 // CHECK27-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3945 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3946 // CHECK27-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3947 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3948 // CHECK27-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3949 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3950 // CHECK27-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
3951 // CHECK27-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3952 // CHECK27: cond.true:
3953 // CHECK27-NEXT: br label [[COND_END:%.*]]
3954 // CHECK27: cond.false:
3955 // CHECK27-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3956 // CHECK27-NEXT: br label [[COND_END]]
3957 // CHECK27: cond.end:
3958 // CHECK27-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3959 // CHECK27-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3960 // CHECK27-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3961 // CHECK27-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3962 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3963 // CHECK27: omp.inner.for.cond:
3964 // CHECK27-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3965 // CHECK27-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3966 // CHECK27-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3967 // CHECK27-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3968 // CHECK27: omp.inner.for.body:
3969 // CHECK27-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3970 // CHECK27-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
3971 // CHECK27-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3972 // CHECK27-NEXT: store i32 [[ADD]], i32* [[I]], align 4
3973 // CHECK27-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
3974 // CHECK27-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP9]]
3975 // CHECK27-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3976 // CHECK27-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3977 // CHECK27: omp.body.continue:
3978 // CHECK27-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3979 // CHECK27: omp.inner.for.inc:
3980 // CHECK27-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3981 // CHECK27-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
3982 // CHECK27-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
3983 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND]]
3984 // CHECK27: omp.inner.for.end:
3985 // CHECK27-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3986 // CHECK27: omp.loop.exit:
3987 // CHECK27-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3988 // CHECK27-NEXT: ret void
3989 //
3990 //
3991 // CHECK27-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3992 // CHECK27-SAME: () #[[ATTR5:[0-9]+]] {
3993 // CHECK27-NEXT: entry:
3994 // CHECK27-NEXT: call void @__tgt_register_requires(i64 1)
3995 // CHECK27-NEXT: ret void
3996 //
3997 //
3998 // CHECK28-LABEL: define {{[^@]+}}@main
3999 // CHECK28-SAME: (i32 [[ARGC:%.*]], i8** [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
4000 // CHECK28-NEXT: entry:
4001 // CHECK28-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
4002 // CHECK28-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
4003 // CHECK28-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
4004 // CHECK28-NEXT: [[N:%.*]] = alloca i32, align 4
4005 // CHECK28-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
4006 // CHECK28-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
4007 // CHECK28-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
4008 // CHECK28-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
4009 // CHECK28-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
4010 // CHECK28-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
4011 // CHECK28-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
4012 // CHECK28-NEXT: [[TMP:%.*]] = alloca i32, align 4
4013 // CHECK28-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4014 // CHECK28-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4015 // CHECK28-NEXT: store i32 0, i32* [[RETVAL]], align 4
4016 // CHECK28-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4017 // CHECK28-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
4018 // CHECK28-NEXT: store i32 100, i32* [[N]], align 4
4019 // CHECK28-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
4020 // CHECK28-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
4021 // CHECK28-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
4022 // CHECK28-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
4023 // CHECK28-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
4024 // CHECK28-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
4025 // CHECK28-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
4026 // CHECK28-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
4027 // CHECK28-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
4028 // CHECK28-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
4029 // CHECK28-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4030 // CHECK28-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
4031 // CHECK28-NEXT: store i32 [[TMP3]], i32* [[TMP7]], align 4
4032 // CHECK28-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4033 // CHECK28-NEXT: [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i32*
4034 // CHECK28-NEXT: store i32 [[TMP3]], i32* [[TMP9]], align 4
4035 // CHECK28-NEXT: [[TMP10:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
4036 // CHECK28-NEXT: store i64 4, i64* [[TMP10]], align 4
4037 // CHECK28-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4038 // CHECK28-NEXT: store i8* null, i8** [[TMP11]], align 4
4039 // CHECK28-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4040 // CHECK28-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
4041 // CHECK28-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
4042 // CHECK28-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4043 // CHECK28-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
4044 // CHECK28-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
4045 // CHECK28-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 1
4046 // CHECK28-NEXT: store i64 4, i64* [[TMP16]], align 4
4047 // CHECK28-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4048 // CHECK28-NEXT: store i8* null, i8** [[TMP17]], align 4
4049 // CHECK28-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4050 // CHECK28-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to i32**
4051 // CHECK28-NEXT: store i32* [[VLA]], i32** [[TMP19]], align 4
4052 // CHECK28-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4053 // CHECK28-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to i32**
4054 // CHECK28-NEXT: store i32* [[VLA]], i32** [[TMP21]], align 4
4055 // CHECK28-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
4056 // CHECK28-NEXT: store i64 [[TMP5]], i64* [[TMP22]], align 4
4057 // CHECK28-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4058 // CHECK28-NEXT: store i8* null, i8** [[TMP23]], align 4
4059 // CHECK28-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4060 // CHECK28-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4061 // CHECK28-NEXT: [[TMP26:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
4062 // CHECK28-NEXT: [[TMP27:%.*]] = load i32, i32* [[N]], align 4
4063 // CHECK28-NEXT: store i32 [[TMP27]], i32* [[DOTCAPTURE_EXPR_]], align 4
4064 // CHECK28-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4065 // CHECK28-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP28]], 0
4066 // CHECK28-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4067 // CHECK28-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
4068 // CHECK28-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4069 // CHECK28-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4070 // CHECK28-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP29]], 1
4071 // CHECK28-NEXT: [[TMP30:%.*]] = zext i32 [[ADD]] to i64
4072 // CHECK28-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i64 [[TMP30]])
4073 // CHECK28-NEXT: [[TMP31:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162.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)
4074 // CHECK28-NEXT: [[TMP32:%.*]] = icmp ne i32 [[TMP31]], 0
4075 // CHECK28-NEXT: br i1 [[TMP32]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4076 // CHECK28: omp_offload.failed:
4077 // CHECK28-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
4078 // CHECK28-NEXT: br label [[OMP_OFFLOAD_CONT]]
4079 // CHECK28: omp_offload.cont:
4080 // CHECK28-NEXT: [[TMP33:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
4081 // CHECK28-NEXT: [[CALL:%.*]] = call i32 @_Z5tmainIiLi10EEiT_(i32 [[TMP33]])
4082 // CHECK28-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
4083 // CHECK28-NEXT: [[TMP34:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
4084 // CHECK28-NEXT: call void @llvm.stackrestore(i8* [[TMP34]])
4085 // CHECK28-NEXT: [[TMP35:%.*]] = load i32, i32* [[RETVAL]], align 4
4086 // CHECK28-NEXT: ret i32 [[TMP35]]
4087 //
4088 //
4089 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162
4090 // CHECK28-SAME: (i32 [[N:%.*]], i32 [[VLA:%.*]], i32* nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
4091 // CHECK28-NEXT: entry:
4092 // CHECK28-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
4093 // CHECK28-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
4094 // CHECK28-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
4095 // CHECK28-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
4096 // CHECK28-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4097 // CHECK28-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
4098 // CHECK28-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4099 // CHECK28-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
4100 // CHECK28-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], 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]])
4101 // CHECK28-NEXT: ret void
4102 //
4103 //
4104 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined.
4105 // CHECK28-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]] {
4106 // CHECK28-NEXT: entry:
4107 // CHECK28-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4108 // CHECK28-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4109 // CHECK28-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
4110 // CHECK28-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
4111 // CHECK28-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
4112 // CHECK28-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4113 // CHECK28-NEXT: [[TMP:%.*]] = alloca i32, align 4
4114 // CHECK28-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4115 // CHECK28-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
4116 // CHECK28-NEXT: [[I:%.*]] = alloca i32, align 4
4117 // CHECK28-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4118 // CHECK28-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4119 // CHECK28-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4120 // CHECK28-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4121 // CHECK28-NEXT: [[I3:%.*]] = alloca i32, align 4
4122 // CHECK28-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4123 // CHECK28-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4124 // CHECK28-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
4125 // CHECK28-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
4126 // CHECK28-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
4127 // CHECK28-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
4128 // CHECK28-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
4129 // CHECK28-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
4130 // CHECK28-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
4131 // CHECK28-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
4132 // CHECK28-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4133 // CHECK28-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
4134 // CHECK28-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
4135 // CHECK28-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
4136 // CHECK28-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
4137 // CHECK28-NEXT: store i32 0, i32* [[I]], align 4
4138 // CHECK28-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4139 // CHECK28-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
4140 // CHECK28-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4141 // CHECK28: omp.precond.then:
4142 // CHECK28-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4143 // CHECK28-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4144 // CHECK28-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
4145 // CHECK28-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4146 // CHECK28-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4147 // CHECK28-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4148 // CHECK28-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
4149 // CHECK28-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4150 // CHECK28-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4151 // CHECK28-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4152 // CHECK28-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
4153 // CHECK28-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4154 // CHECK28: cond.true:
4155 // CHECK28-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
4156 // CHECK28-NEXT: br label [[COND_END:%.*]]
4157 // CHECK28: cond.false:
4158 // CHECK28-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4159 // CHECK28-NEXT: br label [[COND_END]]
4160 // CHECK28: cond.end:
4161 // CHECK28-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
4162 // CHECK28-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4163 // CHECK28-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4164 // CHECK28-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
4165 // CHECK28-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4166 // CHECK28: omp.inner.for.cond:
4167 // CHECK28-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4168 // CHECK28-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4169 // CHECK28-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
4170 // CHECK28-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4171 // CHECK28: omp.inner.for.body:
4172 // CHECK28-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4173 // CHECK28-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
4174 // CHECK28-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4175 // CHECK28-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
4176 // CHECK28-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
4177 // CHECK28-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP17]]
4178 // CHECK28-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4179 // CHECK28-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4180 // CHECK28: omp.body.continue:
4181 // CHECK28-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4182 // CHECK28: omp.inner.for.inc:
4183 // CHECK28-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4184 // CHECK28-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
4185 // CHECK28-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
4186 // CHECK28-NEXT: br label [[OMP_INNER_FOR_COND]]
4187 // CHECK28: omp.inner.for.end:
4188 // CHECK28-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4189 // CHECK28: omp.loop.exit:
4190 // CHECK28-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4191 // CHECK28-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
4192 // CHECK28-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
4193 // CHECK28-NEXT: br label [[OMP_PRECOND_END]]
4194 // CHECK28: omp.precond.end:
4195 // CHECK28-NEXT: ret void
4196 //
4197 //
4198 // CHECK28-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
4199 // CHECK28-SAME: (i32 [[ARGC:%.*]]) #[[ATTR4:[0-9]+]] comdat {
4200 // CHECK28-NEXT: entry:
4201 // CHECK28-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
4202 // CHECK28-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
4203 // CHECK28-NEXT: [[TE:%.*]] = alloca i32, align 4
4204 // CHECK28-NEXT: [[TH:%.*]] = alloca i32, align 4
4205 // CHECK28-NEXT: [[TE_CASTED:%.*]] = alloca i32, align 4
4206 // CHECK28-NEXT: [[TH_CASTED:%.*]] = alloca i32, align 4
4207 // CHECK28-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
4208 // CHECK28-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
4209 // CHECK28-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
4210 // CHECK28-NEXT: [[TMP:%.*]] = alloca i32, align 4
4211 // CHECK28-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
4212 // CHECK28-NEXT: store i32 0, i32* [[TE]], align 4
4213 // CHECK28-NEXT: store i32 128, i32* [[TH]], align 4
4214 // CHECK28-NEXT: [[TMP0:%.*]] = load i32, i32* [[TE]], align 4
4215 // CHECK28-NEXT: store i32 [[TMP0]], i32* [[TE_CASTED]], align 4
4216 // CHECK28-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE_CASTED]], align 4
4217 // CHECK28-NEXT: [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
4218 // CHECK28-NEXT: store i32 [[TMP2]], i32* [[TH_CASTED]], align 4
4219 // CHECK28-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_CASTED]], align 4
4220 // CHECK28-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4221 // CHECK28-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32*
4222 // CHECK28-NEXT: store i32 [[TMP1]], i32* [[TMP5]], align 4
4223 // CHECK28-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4224 // CHECK28-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
4225 // CHECK28-NEXT: store i32 [[TMP1]], i32* [[TMP7]], align 4
4226 // CHECK28-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
4227 // CHECK28-NEXT: store i8* null, i8** [[TMP8]], align 4
4228 // CHECK28-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
4229 // CHECK28-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
4230 // CHECK28-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
4231 // CHECK28-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
4232 // CHECK28-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
4233 // CHECK28-NEXT: store i32 [[TMP3]], i32* [[TMP12]], align 4
4234 // CHECK28-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
4235 // CHECK28-NEXT: store i8* null, i8** [[TMP13]], align 4
4236 // CHECK28-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
4237 // CHECK28-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [10 x i32]**
4238 // CHECK28-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP15]], align 4
4239 // CHECK28-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
4240 // CHECK28-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [10 x i32]**
4241 // CHECK28-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP17]], align 4
4242 // CHECK28-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
4243 // CHECK28-NEXT: store i8* null, i8** [[TMP18]], align 4
4244 // CHECK28-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4245 // CHECK28-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4246 // CHECK28-NEXT: [[TMP21:%.*]] = load i32, i32* [[TE]], align 4
4247 // CHECK28-NEXT: [[TMP22:%.*]] = load i32, i32* [[TH]], align 4
4248 // CHECK28-NEXT: call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i64 10)
4249 // CHECK28-NEXT: [[TMP23:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB2]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151.region_id, i32 3, i8** [[TMP19]], i8** [[TMP20]], i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.2, i32 0, i32 0), i8** null, i8** null, i32 [[TMP21]], i32 [[TMP22]])
4250 // CHECK28-NEXT: [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
4251 // CHECK28-NEXT: br i1 [[TMP24]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4252 // CHECK28: omp_offload.failed:
4253 // CHECK28-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[A]]) #[[ATTR3]]
4254 // CHECK28-NEXT: br label [[OMP_OFFLOAD_CONT]]
4255 // CHECK28: omp_offload.cont:
4256 // CHECK28-NEXT: ret i32 0
4257 //
4258 //
4259 // CHECK28-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151
4260 // CHECK28-SAME: (i32 [[TE:%.*]], i32 [[TH:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4261 // CHECK28-NEXT: entry:
4262 // CHECK28-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4
4263 // CHECK28-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4
4264 // CHECK28-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4265 // CHECK28-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
4266 // CHECK28-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4
4267 // CHECK28-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4
4268 // CHECK28-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4269 // CHECK28-NEXT: [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4270 // CHECK28-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4
4271 // CHECK28-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4
4272 // CHECK28-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
4273 // CHECK28-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP1]])
4274 // CHECK28-NEXT: ret void
4275 //
4276 //
4277 // CHECK28-LABEL: define {{[^@]+}}@.omp_outlined..1
4278 // CHECK28-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], [10 x i32]* nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4279 // CHECK28-NEXT: entry:
4280 // CHECK28-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4281 // CHECK28-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4282 // CHECK28-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4283 // CHECK28-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4284 // CHECK28-NEXT: [[TMP:%.*]] = alloca i32, align 4
4285 // CHECK28-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4286 // CHECK28-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4287 // CHECK28-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4288 // CHECK28-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4289 // CHECK28-NEXT: [[I:%.*]] = alloca i32, align 4
4290 // CHECK28-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4291 // CHECK28-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4292 // CHECK28-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4293 // CHECK28-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4294 // CHECK28-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4295 // CHECK28-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
4296 // CHECK28-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4297 // CHECK28-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4298 // CHECK28-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4299 // CHECK28-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4300 // CHECK28-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4301 // CHECK28-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4302 // CHECK28-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
4303 // CHECK28-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4304 // CHECK28: cond.true:
4305 // CHECK28-NEXT: br label [[COND_END:%.*]]
4306 // CHECK28: cond.false:
4307 // CHECK28-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4308 // CHECK28-NEXT: br label [[COND_END]]
4309 // CHECK28: cond.end:
4310 // CHECK28-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4311 // CHECK28-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4312 // CHECK28-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4313 // CHECK28-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4314 // CHECK28-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4315 // CHECK28: omp.inner.for.cond:
4316 // CHECK28-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4317 // CHECK28-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4318 // CHECK28-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4319 // CHECK28-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4320 // CHECK28: omp.inner.for.body:
4321 // CHECK28-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4322 // CHECK28-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
4323 // CHECK28-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4324 // CHECK28-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4325 // CHECK28-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
4326 // CHECK28-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP9]]
4327 // CHECK28-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4328 // CHECK28-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4329 // CHECK28: omp.body.continue:
4330 // CHECK28-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4331 // CHECK28: omp.inner.for.inc:
4332 // CHECK28-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4333 // CHECK28-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
4334 // CHECK28-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
4335 // CHECK28-NEXT: br label [[OMP_INNER_FOR_COND]]
4336 // CHECK28: omp.inner.for.end:
4337 // CHECK28-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4338 // CHECK28: omp.loop.exit:
4339 // CHECK28-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4340 // CHECK28-NEXT: ret void
4341 //
4342 //
4343 // CHECK28-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4344 // CHECK28-SAME: () #[[ATTR5:[0-9]+]] {
4345 // CHECK28-NEXT: entry:
4346 // CHECK28-NEXT: call void @__tgt_register_requires(i64 1)
4347 // CHECK28-NEXT: ret void
4348 //
4349 //