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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK1
9 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK3
12
13 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK9
57 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK11
60
61 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK17
90 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK19
93
94 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK25
133 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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=CHECK27
136
137 // RUN: %clang_cc1 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 -no-opaque-pointers -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 noundef 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: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
264 // CHECK1-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
265 // CHECK1-NEXT: store i32 1, i32* [[TMP35]], align 4
266 // CHECK1-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
267 // CHECK1-NEXT: store i32 4, i32* [[TMP36]], align 4
268 // CHECK1-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
269 // CHECK1-NEXT: store i8** [[TMP27]], i8*** [[TMP37]], align 8
270 // CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
271 // CHECK1-NEXT: store i8** [[TMP28]], i8*** [[TMP38]], align 8
272 // CHECK1-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
273 // CHECK1-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP39]], align 8
274 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
275 // CHECK1-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP40]], align 8
276 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
277 // CHECK1-NEXT: store i8** null, i8*** [[TMP41]], align 8
278 // CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
279 // CHECK1-NEXT: store i8** null, i8*** [[TMP42]], align 8
280 // CHECK1-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
281 // CHECK1-NEXT: store i64 [[TMP34]], i64* [[TMP43]], align 8
282 // CHECK1-NEXT: [[TMP44:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 [[TMP29]], i32 [[TMP30]], i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
283 // CHECK1-NEXT: [[TMP45:%.*]] = icmp ne i32 [[TMP44]], 0
284 // CHECK1-NEXT: br i1 [[TMP45]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
285 // CHECK1: omp_offload.failed:
286 // 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]+]]
287 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
288 // CHECK1: omp_offload.cont:
289 // CHECK1-NEXT: [[TMP46:%.*]] = load i32, i32* [[N_ADDR]], align 4
290 // CHECK1-NEXT: [[CONV7:%.*]] = bitcast i64* [[N_CASTED6]] to i32*
291 // CHECK1-NEXT: store i32 [[TMP46]], i32* [[CONV7]], align 4
292 // CHECK1-NEXT: [[TMP47:%.*]] = load i64, i64* [[N_CASTED6]], align 8
293 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
294 // CHECK1-NEXT: [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i64*
295 // CHECK1-NEXT: store i64 [[TMP47]], i64* [[TMP49]], align 8
296 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
297 // CHECK1-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
298 // CHECK1-NEXT: store i64 [[TMP47]], i64* [[TMP51]], align 8
299 // CHECK1-NEXT: [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
300 // CHECK1-NEXT: store i8* null, i8** [[TMP52]], align 8
301 // CHECK1-NEXT: [[TMP53:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
302 // CHECK1-NEXT: [[TMP54:%.*]] = bitcast i8** [[TMP53]] to [100 x i32]**
303 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP54]], align 8
304 // CHECK1-NEXT: [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
305 // CHECK1-NEXT: [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [100 x i32]**
306 // CHECK1-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP56]], align 8
307 // CHECK1-NEXT: [[TMP57:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 1
308 // CHECK1-NEXT: store i8* null, i8** [[TMP57]], align 8
309 // CHECK1-NEXT: [[TMP58:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
310 // CHECK1-NEXT: [[TMP59:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
311 // CHECK1-NEXT: [[TMP60:%.*]] = load i32, i32* [[N_ADDR]], align 4
312 // CHECK1-NEXT: store i32 [[TMP60]], i32* [[DOTCAPTURE_EXPR_12]], align 4
313 // CHECK1-NEXT: [[TMP61:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_12]], align 4
314 // CHECK1-NEXT: [[SUB14:%.*]] = sub nsw i32 [[TMP61]], 0
315 // CHECK1-NEXT: [[DIV15:%.*]] = sdiv i32 [[SUB14]], 1
316 // CHECK1-NEXT: [[SUB16:%.*]] = sub nsw i32 [[DIV15]], 1
317 // CHECK1-NEXT: store i32 [[SUB16]], i32* [[DOTCAPTURE_EXPR_13]], align 4
318 // CHECK1-NEXT: [[TMP62:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_13]], align 4
319 // CHECK1-NEXT: [[ADD17:%.*]] = add nsw i32 [[TMP62]], 1
320 // CHECK1-NEXT: [[TMP63:%.*]] = zext i32 [[ADD17]] to i64
321 // CHECK1-NEXT: [[KERNEL_ARGS18:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
322 // CHECK1-NEXT: [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 0
323 // CHECK1-NEXT: store i32 1, i32* [[TMP64]], align 4
324 // CHECK1-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 1
325 // CHECK1-NEXT: store i32 2, i32* [[TMP65]], align 4
326 // CHECK1-NEXT: [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 2
327 // CHECK1-NEXT: store i8** [[TMP58]], i8*** [[TMP66]], align 8
328 // CHECK1-NEXT: [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 3
329 // CHECK1-NEXT: store i8** [[TMP59]], i8*** [[TMP67]], align 8
330 // CHECK1-NEXT: [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 4
331 // CHECK1-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP68]], align 8
332 // CHECK1-NEXT: [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 5
333 // CHECK1-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP69]], align 8
334 // CHECK1-NEXT: [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 6
335 // CHECK1-NEXT: store i8** null, i8*** [[TMP70]], align 8
336 // CHECK1-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 7
337 // CHECK1-NEXT: store i8** null, i8*** [[TMP71]], align 8
338 // CHECK1-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]], i32 0, i32 8
339 // CHECK1-NEXT: store i64 [[TMP63]], i64* [[TMP72]], align 8
340 // CHECK1-NEXT: [[TMP73:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS18]])
341 // CHECK1-NEXT: [[TMP74:%.*]] = icmp ne i32 [[TMP73]], 0
342 // CHECK1-NEXT: br i1 [[TMP74]], label [[OMP_OFFLOAD_FAILED19:%.*]], label [[OMP_OFFLOAD_CONT20:%.*]]
343 // CHECK1: omp_offload.failed19:
344 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35(i64 [[TMP47]], [100 x i32]* @a) #[[ATTR2]]
345 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT20]]
346 // CHECK1: omp_offload.cont20:
347 // CHECK1-NEXT: [[TMP75:%.*]] = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i64 0, i64 0), align 4
348 // CHECK1-NEXT: ret i32 [[TMP75]]
349 //
350 //
351 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29
352 // CHECK1-SAME: (i64 noundef [[TE:%.*]], i64 noundef [[TH:%.*]], i64 noundef [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
353 // CHECK1-NEXT: entry:
354 // CHECK1-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8
355 // CHECK1-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8
356 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
357 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
358 // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
359 // CHECK1-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8
360 // CHECK1-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8
361 // CHECK1-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
362 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
363 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_ADDR]] to i32*
364 // CHECK1-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_ADDR]] to i32*
365 // CHECK1-NEXT: [[CONV2:%.*]] = bitcast i64* [[N_ADDR]] to i32*
366 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
367 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
368 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
369 // CHECK1-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
370 // 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]])
371 // CHECK1-NEXT: ret void
372 //
373 //
374 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
375 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
376 // CHECK1-NEXT: entry:
377 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
378 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
379 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
380 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
381 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
382 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
383 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
384 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
385 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
386 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
387 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
388 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
389 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
390 // CHECK1-NEXT: [[I3:%.*]] = alloca i32, align 4
391 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
392 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
393 // CHECK1-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
394 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
395 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
396 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
397 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
398 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
399 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
400 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
401 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
402 // CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
403 // CHECK1-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
404 // CHECK1-NEXT: store i32 0, i32* [[I]], align 4
405 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
406 // CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
407 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
408 // CHECK1: omp.precond.then:
409 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
410 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
411 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
412 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
413 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
414 // CHECK1-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
415 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
416 // 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)
417 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
418 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
419 // CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
420 // CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
421 // CHECK1: cond.true:
422 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
423 // CHECK1-NEXT: br label [[COND_END:%.*]]
424 // CHECK1: cond.false:
425 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
426 // CHECK1-NEXT: br label [[COND_END]]
427 // CHECK1: cond.end:
428 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
429 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
430 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
431 // CHECK1-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
432 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
433 // CHECK1: omp.inner.for.cond:
434 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
435 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
436 // CHECK1-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
437 // CHECK1-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
438 // CHECK1: omp.inner.for.body:
439 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
440 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
441 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
442 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
443 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
444 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
445 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]]
446 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
447 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
448 // CHECK1: omp.body.continue:
449 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
450 // CHECK1: omp.inner.for.inc:
451 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
452 // CHECK1-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
453 // CHECK1-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
454 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
455 // CHECK1: omp.inner.for.end:
456 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
457 // CHECK1: omp.loop.exit:
458 // CHECK1-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
459 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
460 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
461 // CHECK1-NEXT: br label [[OMP_PRECOND_END]]
462 // CHECK1: omp.precond.end:
463 // CHECK1-NEXT: ret void
464 //
465 //
466 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35
467 // CHECK1-SAME: (i64 noundef [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
468 // CHECK1-NEXT: entry:
469 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
470 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
471 // CHECK1-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
472 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
473 // CHECK1-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
474 // CHECK1-NEXT: [[TMP0:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
475 // 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]])
476 // CHECK1-NEXT: ret void
477 //
478 //
479 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
480 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
481 // CHECK1-NEXT: entry:
482 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
483 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
484 // CHECK1-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
485 // CHECK1-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 8
486 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
487 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
488 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
489 // CHECK1-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
490 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
491 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
492 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
493 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
494 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
495 // CHECK1-NEXT: [[I3:%.*]] = alloca i32, align 4
496 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
497 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
498 // CHECK1-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
499 // CHECK1-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 8
500 // CHECK1-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
501 // CHECK1-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 8
502 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
503 // CHECK1-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
504 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
505 // CHECK1-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
506 // CHECK1-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
507 // CHECK1-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
508 // CHECK1-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
509 // CHECK1-NEXT: store i32 0, i32* [[I]], align 4
510 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
511 // CHECK1-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
512 // CHECK1-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
513 // CHECK1: omp.precond.then:
514 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
515 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
516 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
517 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
518 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
519 // CHECK1-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
520 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
521 // 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)
522 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
523 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
524 // CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
525 // CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
526 // CHECK1: cond.true:
527 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
528 // CHECK1-NEXT: br label [[COND_END:%.*]]
529 // CHECK1: cond.false:
530 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
531 // CHECK1-NEXT: br label [[COND_END]]
532 // CHECK1: cond.end:
533 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
534 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
535 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
536 // CHECK1-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
537 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
538 // CHECK1: omp.inner.for.cond:
539 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
540 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
541 // CHECK1-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
542 // CHECK1-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
543 // CHECK1: omp.inner.for.body:
544 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
545 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
546 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
547 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
548 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
549 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
550 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i64 0, i64 [[IDXPROM]]
551 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
552 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
553 // CHECK1: omp.body.continue:
554 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
555 // CHECK1: omp.inner.for.inc:
556 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
557 // CHECK1-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
558 // CHECK1-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
559 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
560 // CHECK1: omp.inner.for.end:
561 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
562 // CHECK1: omp.loop.exit:
563 // CHECK1-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
564 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
565 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
566 // CHECK1-NEXT: br label [[OMP_PRECOND_END]]
567 // CHECK1: omp.precond.end:
568 // CHECK1-NEXT: ret void
569 //
570 //
571 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
572 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
573 // CHECK1-NEXT: entry:
574 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
575 // CHECK1-NEXT: ret void
576 //
577 //
578 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_argument_globali
579 // CHECK3-SAME: (i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
580 // CHECK3-NEXT: entry:
581 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
582 // CHECK3-NEXT: [[TE:%.*]] = alloca i32, align 4
583 // CHECK3-NEXT: [[TH:%.*]] = alloca i32, align 4
584 // CHECK3-NEXT: [[TE_CASTED:%.*]] = alloca i32, align 4
585 // CHECK3-NEXT: [[TH_CASTED:%.*]] = alloca i32, align 4
586 // CHECK3-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
587 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x i8*], align 4
588 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x i8*], align 4
589 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x i8*], align 4
590 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
591 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
592 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
593 // CHECK3-NEXT: [[N_CASTED4:%.*]] = alloca i32, align 4
594 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [2 x i8*], align 4
595 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [2 x i8*], align 4
596 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [2 x i8*], align 4
597 // CHECK3-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
598 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
599 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
600 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
601 // CHECK3-NEXT: [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
602 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[TMP0]], 128
603 // CHECK3-NEXT: store i32 [[DIV]], i32* [[TE]], align 4
604 // CHECK3-NEXT: store i32 128, i32* [[TH]], align 4
605 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE]], align 4
606 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[TE_CASTED]], align 4
607 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_CASTED]], align 4
608 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH]], align 4
609 // CHECK3-NEXT: store i32 [[TMP3]], i32* [[TH_CASTED]], align 4
610 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TH_CASTED]], align 4
611 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[N_ADDR]], align 4
612 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[N_CASTED]], align 4
613 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[N_CASTED]], align 4
614 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
615 // CHECK3-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
616 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[TMP8]], align 4
617 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
618 // CHECK3-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
619 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[TMP10]], align 4
620 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
621 // CHECK3-NEXT: store i8* null, i8** [[TMP11]], align 4
622 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
623 // CHECK3-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
624 // CHECK3-NEXT: store i32 [[TMP4]], i32* [[TMP13]], align 4
625 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
626 // CHECK3-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
627 // CHECK3-NEXT: store i32 [[TMP4]], i32* [[TMP15]], align 4
628 // CHECK3-NEXT: [[TMP16:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
629 // CHECK3-NEXT: store i8* null, i8** [[TMP16]], align 4
630 // CHECK3-NEXT: [[TMP17:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
631 // CHECK3-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32*
632 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[TMP18]], align 4
633 // CHECK3-NEXT: [[TMP19:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
634 // CHECK3-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32*
635 // CHECK3-NEXT: store i32 [[TMP6]], i32* [[TMP20]], align 4
636 // CHECK3-NEXT: [[TMP21:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
637 // CHECK3-NEXT: store i8* null, i8** [[TMP21]], align 4
638 // CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
639 // CHECK3-NEXT: [[TMP23:%.*]] = bitcast i8** [[TMP22]] to [100 x i32]**
640 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP23]], align 4
641 // CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 3
642 // CHECK3-NEXT: [[TMP25:%.*]] = bitcast i8** [[TMP24]] to [100 x i32]**
643 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP25]], align 4
644 // CHECK3-NEXT: [[TMP26:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
645 // CHECK3-NEXT: store i8* null, i8** [[TMP26]], align 4
646 // CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
647 // CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
648 // CHECK3-NEXT: [[TMP29:%.*]] = load i32, i32* [[TE]], align 4
649 // CHECK3-NEXT: [[TMP30:%.*]] = load i32, i32* [[TH]], align 4
650 // CHECK3-NEXT: [[TMP31:%.*]] = load i32, i32* [[N_ADDR]], align 4
651 // CHECK3-NEXT: store i32 [[TMP31]], i32* [[DOTCAPTURE_EXPR_]], align 4
652 // CHECK3-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
653 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP32]], 0
654 // CHECK3-NEXT: [[DIV2:%.*]] = sdiv i32 [[SUB]], 1
655 // CHECK3-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV2]], 1
656 // CHECK3-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
657 // CHECK3-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
658 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP33]], 1
659 // CHECK3-NEXT: [[TMP34:%.*]] = zext i32 [[ADD]] to i64
660 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
661 // CHECK3-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
662 // CHECK3-NEXT: store i32 1, i32* [[TMP35]], align 4
663 // CHECK3-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
664 // CHECK3-NEXT: store i32 4, i32* [[TMP36]], align 4
665 // CHECK3-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
666 // CHECK3-NEXT: store i8** [[TMP27]], i8*** [[TMP37]], align 4
667 // CHECK3-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
668 // CHECK3-NEXT: store i8** [[TMP28]], i8*** [[TMP38]], align 4
669 // CHECK3-NEXT: [[TMP39:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
670 // CHECK3-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP39]], align 4
671 // CHECK3-NEXT: [[TMP40:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
672 // CHECK3-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP40]], align 4
673 // CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
674 // CHECK3-NEXT: store i8** null, i8*** [[TMP41]], align 4
675 // CHECK3-NEXT: [[TMP42:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
676 // CHECK3-NEXT: store i8** null, i8*** [[TMP42]], align 4
677 // CHECK3-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
678 // CHECK3-NEXT: store i64 [[TMP34]], i64* [[TMP43]], align 8
679 // CHECK3-NEXT: [[TMP44:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 [[TMP29]], i32 [[TMP30]], i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
680 // CHECK3-NEXT: [[TMP45:%.*]] = icmp ne i32 [[TMP44]], 0
681 // CHECK3-NEXT: br i1 [[TMP45]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
682 // CHECK3: omp_offload.failed:
683 // 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]+]]
684 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
685 // CHECK3: omp_offload.cont:
686 // CHECK3-NEXT: [[TMP46:%.*]] = load i32, i32* [[N_ADDR]], align 4
687 // CHECK3-NEXT: store i32 [[TMP46]], i32* [[N_CASTED4]], align 4
688 // CHECK3-NEXT: [[TMP47:%.*]] = load i32, i32* [[N_CASTED4]], align 4
689 // CHECK3-NEXT: [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
690 // CHECK3-NEXT: [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32*
691 // CHECK3-NEXT: store i32 [[TMP47]], i32* [[TMP49]], align 4
692 // CHECK3-NEXT: [[TMP50:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
693 // CHECK3-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i32*
694 // CHECK3-NEXT: store i32 [[TMP47]], i32* [[TMP51]], align 4
695 // CHECK3-NEXT: [[TMP52:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 0
696 // CHECK3-NEXT: store i8* null, i8** [[TMP52]], align 4
697 // CHECK3-NEXT: [[TMP53:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
698 // CHECK3-NEXT: [[TMP54:%.*]] = bitcast i8** [[TMP53]] to [100 x i32]**
699 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP54]], align 4
700 // CHECK3-NEXT: [[TMP55:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
701 // CHECK3-NEXT: [[TMP56:%.*]] = bitcast i8** [[TMP55]] to [100 x i32]**
702 // CHECK3-NEXT: store [100 x i32]* @a, [100 x i32]** [[TMP56]], align 4
703 // CHECK3-NEXT: [[TMP57:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i32 0, i32 1
704 // CHECK3-NEXT: store i8* null, i8** [[TMP57]], align 4
705 // CHECK3-NEXT: [[TMP58:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
706 // CHECK3-NEXT: [[TMP59:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
707 // CHECK3-NEXT: [[TMP60:%.*]] = load i32, i32* [[N_ADDR]], align 4
708 // CHECK3-NEXT: store i32 [[TMP60]], i32* [[DOTCAPTURE_EXPR_9]], align 4
709 // CHECK3-NEXT: [[TMP61:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
710 // CHECK3-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP61]], 0
711 // CHECK3-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
712 // CHECK3-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
713 // CHECK3-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
714 // CHECK3-NEXT: [[TMP62:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
715 // CHECK3-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP62]], 1
716 // CHECK3-NEXT: [[TMP63:%.*]] = zext i32 [[ADD14]] to i64
717 // CHECK3-NEXT: [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
718 // CHECK3-NEXT: [[TMP64:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
719 // CHECK3-NEXT: store i32 1, i32* [[TMP64]], align 4
720 // CHECK3-NEXT: [[TMP65:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
721 // CHECK3-NEXT: store i32 2, i32* [[TMP65]], align 4
722 // CHECK3-NEXT: [[TMP66:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
723 // CHECK3-NEXT: store i8** [[TMP58]], i8*** [[TMP66]], align 4
724 // CHECK3-NEXT: [[TMP67:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
725 // CHECK3-NEXT: store i8** [[TMP59]], i8*** [[TMP67]], align 4
726 // CHECK3-NEXT: [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
727 // CHECK3-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP68]], align 4
728 // CHECK3-NEXT: [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
729 // CHECK3-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP69]], align 4
730 // CHECK3-NEXT: [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
731 // CHECK3-NEXT: store i8** null, i8*** [[TMP70]], align 4
732 // CHECK3-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
733 // CHECK3-NEXT: store i8** null, i8*** [[TMP71]], align 4
734 // CHECK3-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
735 // CHECK3-NEXT: store i64 [[TMP63]], i64* [[TMP72]], align 8
736 // CHECK3-NEXT: [[TMP73:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
737 // CHECK3-NEXT: [[TMP74:%.*]] = icmp ne i32 [[TMP73]], 0
738 // CHECK3-NEXT: br i1 [[TMP74]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
739 // CHECK3: omp_offload.failed16:
740 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35(i32 [[TMP47]], [100 x i32]* @a) #[[ATTR2]]
741 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT17]]
742 // CHECK3: omp_offload.cont17:
743 // CHECK3-NEXT: [[TMP75:%.*]] = load i32, i32* getelementptr inbounds ([100 x i32], [100 x i32]* @a, i32 0, i32 0), align 4
744 // CHECK3-NEXT: ret i32 [[TMP75]]
745 //
746 //
747 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l29
748 // CHECK3-SAME: (i32 noundef [[TE:%.*]], i32 noundef [[TH:%.*]], i32 noundef [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
749 // CHECK3-NEXT: entry:
750 // CHECK3-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4
751 // CHECK3-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4
752 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
753 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
754 // CHECK3-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
755 // CHECK3-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4
756 // CHECK3-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4
757 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
758 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
759 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
760 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4
761 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4
762 // CHECK3-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
763 // 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]])
764 // CHECK3-NEXT: ret void
765 //
766 //
767 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
768 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
769 // CHECK3-NEXT: entry:
770 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
771 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
772 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
773 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
774 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
775 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
776 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
777 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
778 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
779 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
780 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
781 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
782 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
783 // CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4
784 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
785 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
786 // CHECK3-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
787 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
788 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
789 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
790 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
791 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
792 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
793 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
794 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
795 // CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
796 // CHECK3-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
797 // CHECK3-NEXT: store i32 0, i32* [[I]], align 4
798 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
799 // CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
800 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
801 // CHECK3: omp.precond.then:
802 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
803 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
804 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
805 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
806 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
807 // CHECK3-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
808 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
809 // 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)
810 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
811 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
812 // CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
813 // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
814 // CHECK3: cond.true:
815 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
816 // CHECK3-NEXT: br label [[COND_END:%.*]]
817 // CHECK3: cond.false:
818 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
819 // CHECK3-NEXT: br label [[COND_END]]
820 // CHECK3: cond.end:
821 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
822 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
823 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
824 // CHECK3-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
825 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
826 // CHECK3: omp.inner.for.cond:
827 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
828 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
829 // CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
830 // CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
831 // CHECK3: omp.inner.for.body:
832 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
833 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
834 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
835 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
836 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
837 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i32 0, i32 [[TMP16]]
838 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
839 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
840 // CHECK3: omp.body.continue:
841 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
842 // CHECK3: omp.inner.for.inc:
843 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
844 // CHECK3-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
845 // CHECK3-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
846 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
847 // CHECK3: omp.inner.for.end:
848 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
849 // CHECK3: omp.loop.exit:
850 // CHECK3-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
851 // CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
852 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
853 // CHECK3-NEXT: br label [[OMP_PRECOND_END]]
854 // CHECK3: omp.precond.end:
855 // CHECK3-NEXT: ret void
856 //
857 //
858 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z21teams_argument_globali_l35
859 // CHECK3-SAME: (i32 noundef [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
860 // CHECK3-NEXT: entry:
861 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
862 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
863 // CHECK3-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
864 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
865 // CHECK3-NEXT: [[TMP0:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
866 // 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]])
867 // CHECK3-NEXT: ret void
868 //
869 //
870 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
871 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], [100 x i32]* noundef nonnull align 4 dereferenceable(400) [[A:%.*]]) #[[ATTR1]] {
872 // CHECK3-NEXT: entry:
873 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
874 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
875 // CHECK3-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
876 // CHECK3-NEXT: [[A_ADDR:%.*]] = alloca [100 x i32]*, align 4
877 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
878 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
879 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
880 // CHECK3-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
881 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
882 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
883 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
884 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
885 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
886 // CHECK3-NEXT: [[I3:%.*]] = alloca i32, align 4
887 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
888 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
889 // CHECK3-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
890 // CHECK3-NEXT: store [100 x i32]* [[A]], [100 x i32]** [[A_ADDR]], align 4
891 // CHECK3-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
892 // CHECK3-NEXT: [[TMP1:%.*]] = load [100 x i32]*, [100 x i32]** [[A_ADDR]], align 4
893 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP0]], align 4
894 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
895 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
896 // CHECK3-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP3]], 0
897 // CHECK3-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
898 // CHECK3-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
899 // CHECK3-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
900 // CHECK3-NEXT: store i32 0, i32* [[I]], align 4
901 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
902 // CHECK3-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP4]]
903 // CHECK3-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
904 // CHECK3: omp.precond.then:
905 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
906 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
907 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_UB]], align 4
908 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
909 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
910 // CHECK3-NEXT: [[TMP6:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
911 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[TMP6]], align 4
912 // 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)
913 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
914 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
915 // CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP8]], [[TMP9]]
916 // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
917 // CHECK3: cond.true:
918 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
919 // CHECK3-NEXT: br label [[COND_END:%.*]]
920 // CHECK3: cond.false:
921 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
922 // CHECK3-NEXT: br label [[COND_END]]
923 // CHECK3: cond.end:
924 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ [[TMP10]], [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
925 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
926 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
927 // CHECK3-NEXT: store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
928 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
929 // CHECK3: omp.inner.for.cond:
930 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
931 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
932 // CHECK3-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
933 // CHECK3-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
934 // CHECK3: omp.inner.for.body:
935 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
936 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
937 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
938 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
939 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[I3]], align 4
940 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [100 x i32], [100 x i32]* [[TMP1]], i32 0, i32 [[TMP16]]
941 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
942 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
943 // CHECK3: omp.body.continue:
944 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
945 // CHECK3: omp.inner.for.inc:
946 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
947 // CHECK3-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP17]], 1
948 // CHECK3-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
949 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
950 // CHECK3: omp.inner.for.end:
951 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
952 // CHECK3: omp.loop.exit:
953 // CHECK3-NEXT: [[TMP18:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
954 // CHECK3-NEXT: [[TMP19:%.*]] = load i32, i32* [[TMP18]], align 4
955 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP19]])
956 // CHECK3-NEXT: br label [[OMP_PRECOND_END]]
957 // CHECK3: omp.precond.end:
958 // CHECK3-NEXT: ret void
959 //
960 //
961 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
962 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
963 // CHECK3-NEXT: entry:
964 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
965 // CHECK3-NEXT: ret void
966 //
967 //
968 // CHECK9-LABEL: define {{[^@]+}}@_Z15teams_local_argv
969 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
970 // CHECK9-NEXT: entry:
971 // CHECK9-NEXT: [[N:%.*]] = alloca i32, align 4
972 // CHECK9-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
973 // CHECK9-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
974 // CHECK9-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
975 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
976 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
977 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
978 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
979 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
980 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
981 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
982 // CHECK9-NEXT: store i32 100, i32* [[N]], align 4
983 // CHECK9-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
984 // CHECK9-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
985 // CHECK9-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
986 // CHECK9-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
987 // CHECK9-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
988 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
989 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
990 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
991 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
992 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
993 // CHECK9-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
994 // CHECK9-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
995 // CHECK9-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
996 // CHECK9-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
997 // CHECK9-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
998 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP8]], align 8
999 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1000 // CHECK9-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
1001 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP10]], align 8
1002 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1003 // CHECK9-NEXT: store i8* null, i8** [[TMP11]], align 8
1004 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1005 // CHECK9-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1006 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
1007 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1008 // CHECK9-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
1009 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
1010 // CHECK9-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1011 // CHECK9-NEXT: store i8* null, i8** [[TMP16]], align 8
1012 // CHECK9-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1013 // CHECK9-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
1014 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 8
1015 // CHECK9-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1016 // CHECK9-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
1017 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 8
1018 // CHECK9-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1019 // CHECK9-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 8
1020 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1021 // CHECK9-NEXT: store i8* null, i8** [[TMP22]], align 8
1022 // CHECK9-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1023 // CHECK9-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1024 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1025 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
1026 // CHECK9-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
1027 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1028 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
1029 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1030 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1031 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1032 // CHECK9-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1033 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
1034 // CHECK9-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
1035 // CHECK9-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1036 // CHECK9-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1037 // CHECK9-NEXT: store i32 1, i32* [[TMP30]], align 4
1038 // CHECK9-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1039 // CHECK9-NEXT: store i32 3, i32* [[TMP31]], align 4
1040 // CHECK9-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1041 // CHECK9-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 8
1042 // CHECK9-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1043 // CHECK9-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 8
1044 // CHECK9-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1045 // CHECK9-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 8
1046 // CHECK9-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1047 // CHECK9-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 8
1048 // CHECK9-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1049 // CHECK9-NEXT: store i8** null, i8*** [[TMP36]], align 8
1050 // CHECK9-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1051 // CHECK9-NEXT: store i8** null, i8*** [[TMP37]], align 8
1052 // CHECK9-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1053 // CHECK9-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
1054 // CHECK9-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1055 // CHECK9-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
1056 // CHECK9-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1057 // CHECK9: omp_offload.failed:
1058 // 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]+]]
1059 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
1060 // CHECK9: omp_offload.cont:
1061 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i64 0
1062 // CHECK9-NEXT: [[TMP41:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1063 // CHECK9-NEXT: [[TMP42:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1064 // CHECK9-NEXT: call void @llvm.stackrestore(i8* [[TMP42]])
1065 // CHECK9-NEXT: ret i32 [[TMP41]]
1066 //
1067 //
1068 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73
1069 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1070 // CHECK9-NEXT: entry:
1071 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
1072 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1073 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1074 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
1075 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1076 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1077 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
1078 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1079 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1080 // 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]])
1081 // CHECK9-NEXT: ret void
1082 //
1083 //
1084 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
1085 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1086 // CHECK9-NEXT: entry:
1087 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1088 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1089 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
1090 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1091 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1092 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1093 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1094 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1095 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1096 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1097 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1098 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1099 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1100 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1101 // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4
1102 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1103 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1104 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
1105 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1106 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1107 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
1108 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1109 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1110 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1111 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1112 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1113 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1114 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1115 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1116 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1117 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
1118 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1119 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1120 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1121 // CHECK9: omp.precond.then:
1122 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1123 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1124 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
1125 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1126 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1127 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1128 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
1129 // 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)
1130 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1131 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1132 // CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1133 // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1134 // CHECK9: cond.true:
1135 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1136 // CHECK9-NEXT: br label [[COND_END:%.*]]
1137 // CHECK9: cond.false:
1138 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1139 // CHECK9-NEXT: br label [[COND_END]]
1140 // CHECK9: cond.end:
1141 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1142 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1143 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1144 // CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1145 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1146 // CHECK9: omp.inner.for.cond:
1147 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1148 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1149 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1150 // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1151 // CHECK9: omp.inner.for.body:
1152 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1153 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
1154 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1155 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
1156 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
1157 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
1158 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
1159 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1160 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1161 // CHECK9: omp.body.continue:
1162 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1163 // CHECK9: omp.inner.for.inc:
1164 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1165 // CHECK9-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
1166 // CHECK9-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1167 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
1168 // CHECK9: omp.inner.for.end:
1169 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1170 // CHECK9: omp.loop.exit:
1171 // CHECK9-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1172 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
1173 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
1174 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
1175 // CHECK9: omp.precond.end:
1176 // CHECK9-NEXT: ret void
1177 //
1178 //
1179 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1180 // CHECK9-SAME: () #[[ATTR5:[0-9]+]] {
1181 // CHECK9-NEXT: entry:
1182 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
1183 // CHECK9-NEXT: ret void
1184 //
1185 //
1186 // CHECK11-LABEL: define {{[^@]+}}@_Z15teams_local_argv
1187 // CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
1188 // CHECK11-NEXT: entry:
1189 // CHECK11-NEXT: [[N:%.*]] = alloca i32, align 4
1190 // CHECK11-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
1191 // CHECK11-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
1192 // CHECK11-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
1193 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
1194 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
1195 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
1196 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
1197 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
1198 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1199 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1200 // CHECK11-NEXT: store i32 100, i32* [[N]], align 4
1201 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1202 // CHECK11-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
1203 // CHECK11-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
1204 // CHECK11-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
1205 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
1206 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
1207 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
1208 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
1209 // CHECK11-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
1210 // CHECK11-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
1211 // CHECK11-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1212 // CHECK11-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP6]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
1213 // CHECK11-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1214 // CHECK11-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
1215 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP8]], align 4
1216 // CHECK11-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1217 // CHECK11-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
1218 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
1219 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1220 // CHECK11-NEXT: store i8* null, i8** [[TMP11]], align 4
1221 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1222 // CHECK11-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
1223 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
1224 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1225 // CHECK11-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
1226 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
1227 // CHECK11-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1228 // CHECK11-NEXT: store i8* null, i8** [[TMP16]], align 4
1229 // CHECK11-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1230 // CHECK11-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
1231 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 4
1232 // CHECK11-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1233 // CHECK11-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
1234 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 4
1235 // CHECK11-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1236 // CHECK11-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 4
1237 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1238 // CHECK11-NEXT: store i8* null, i8** [[TMP22]], align 4
1239 // CHECK11-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1240 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1241 // CHECK11-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1242 // CHECK11-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
1243 // CHECK11-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
1244 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1245 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
1246 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1247 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1248 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1249 // CHECK11-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1250 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
1251 // CHECK11-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
1252 // CHECK11-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1253 // CHECK11-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1254 // CHECK11-NEXT: store i32 1, i32* [[TMP30]], align 4
1255 // CHECK11-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1256 // CHECK11-NEXT: store i32 3, i32* [[TMP31]], align 4
1257 // CHECK11-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1258 // CHECK11-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 4
1259 // CHECK11-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1260 // CHECK11-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 4
1261 // CHECK11-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1262 // CHECK11-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 4
1263 // CHECK11-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1264 // CHECK11-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 4
1265 // CHECK11-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1266 // CHECK11-NEXT: store i8** null, i8*** [[TMP36]], align 4
1267 // CHECK11-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1268 // CHECK11-NEXT: store i8** null, i8*** [[TMP37]], align 4
1269 // CHECK11-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1270 // CHECK11-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
1271 // CHECK11-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1272 // CHECK11-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
1273 // CHECK11-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1274 // CHECK11: omp_offload.failed:
1275 // 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]+]]
1276 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
1277 // CHECK11: omp_offload.cont:
1278 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[VLA]], i32 0
1279 // CHECK11-NEXT: [[TMP41:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1280 // CHECK11-NEXT: [[TMP42:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
1281 // CHECK11-NEXT: call void @llvm.stackrestore(i8* [[TMP42]])
1282 // CHECK11-NEXT: ret i32 [[TMP41]]
1283 //
1284 //
1285 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z15teams_local_argv_l73
1286 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1287 // CHECK11-NEXT: entry:
1288 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
1289 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
1290 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
1291 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
1292 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
1293 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
1294 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
1295 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
1296 // 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]])
1297 // CHECK11-NEXT: ret void
1298 //
1299 //
1300 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
1301 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1302 // CHECK11-NEXT: entry:
1303 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1304 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1305 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
1306 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
1307 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
1308 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1309 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
1310 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1311 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1312 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
1313 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1314 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1315 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1316 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1317 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
1318 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1319 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1320 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
1321 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
1322 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
1323 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
1324 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
1325 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
1326 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1327 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1328 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1329 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1330 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1331 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1332 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1333 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
1334 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1335 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1336 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1337 // CHECK11: omp.precond.then:
1338 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1339 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1340 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
1341 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1342 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1343 // CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1344 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
1345 // 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)
1346 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1347 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1348 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1349 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1350 // CHECK11: cond.true:
1351 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1352 // CHECK11-NEXT: br label [[COND_END:%.*]]
1353 // CHECK11: cond.false:
1354 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1355 // CHECK11-NEXT: br label [[COND_END]]
1356 // CHECK11: cond.end:
1357 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1358 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1359 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1360 // CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1361 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1362 // CHECK11: omp.inner.for.cond:
1363 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1364 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1365 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1366 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1367 // CHECK11: omp.inner.for.body:
1368 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1369 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
1370 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1371 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
1372 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
1373 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP17]]
1374 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1375 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1376 // CHECK11: omp.body.continue:
1377 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1378 // CHECK11: omp.inner.for.inc:
1379 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1380 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
1381 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1382 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
1383 // CHECK11: omp.inner.for.end:
1384 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1385 // CHECK11: omp.loop.exit:
1386 // CHECK11-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1387 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
1388 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
1389 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
1390 // CHECK11: omp.precond.end:
1391 // CHECK11-NEXT: ret void
1392 //
1393 //
1394 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1395 // CHECK11-SAME: () #[[ATTR5:[0-9]+]] {
1396 // CHECK11-NEXT: entry:
1397 // CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
1398 // CHECK11-NEXT: ret void
1399 //
1400 //
1401 // CHECK17-LABEL: define {{[^@]+}}@_Z21teams_template_structv
1402 // CHECK17-SAME: () #[[ATTR0:[0-9]+]] {
1403 // CHECK17-NEXT: entry:
1404 // CHECK17-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
1405 // CHECK17-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
1406 // CHECK17-NEXT: ret i32 [[CALL]]
1407 //
1408 //
1409 // CHECK17-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
1410 // CHECK17-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
1411 // CHECK17-NEXT: entry:
1412 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1413 // CHECK17-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
1414 // CHECK17-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
1415 // CHECK17-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
1416 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
1417 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1418 // CHECK17-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1419 // CHECK17-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
1420 // CHECK17-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1421 // CHECK17-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
1422 // CHECK17-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
1423 // CHECK17-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1424 // CHECK17-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
1425 // CHECK17-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
1426 // CHECK17-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1427 // CHECK17-NEXT: store i8* null, i8** [[TMP4]], align 8
1428 // CHECK17-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1429 // CHECK17-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1430 // CHECK17-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1431 // CHECK17-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1432 // CHECK17-NEXT: store i32 1, i32* [[TMP7]], align 4
1433 // CHECK17-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1434 // CHECK17-NEXT: store i32 1, i32* [[TMP8]], align 4
1435 // CHECK17-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1436 // CHECK17-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8
1437 // CHECK17-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1438 // CHECK17-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8
1439 // CHECK17-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1440 // CHECK17-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 8
1441 // CHECK17-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1442 // CHECK17-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 8
1443 // CHECK17-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1444 // CHECK17-NEXT: store i8** null, i8*** [[TMP13]], align 8
1445 // CHECK17-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1446 // CHECK17-NEXT: store i8** null, i8*** [[TMP14]], align 8
1447 // CHECK17-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1448 // CHECK17-NEXT: store i64 123, i64* [[TMP15]], align 8
1449 // CHECK17-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1450 // CHECK17-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
1451 // CHECK17-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1452 // CHECK17: omp_offload.failed:
1453 // CHECK17-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
1454 // CHECK17-NEXT: br label [[OMP_OFFLOAD_CONT]]
1455 // CHECK17: omp_offload.cont:
1456 // CHECK17-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1457 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A2]], i64 0, i64 0
1458 // CHECK17-NEXT: [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1459 // CHECK17-NEXT: ret i32 [[TMP18]]
1460 //
1461 //
1462 // CHECK17-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109
1463 // CHECK17-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
1464 // CHECK17-NEXT: entry:
1465 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1466 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1467 // CHECK17-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1468 // 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]])
1469 // CHECK17-NEXT: ret void
1470 //
1471 //
1472 // CHECK17-LABEL: define {{[^@]+}}@.omp_outlined.
1473 // CHECK17-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1474 // CHECK17-NEXT: entry:
1475 // CHECK17-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1476 // CHECK17-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1477 // CHECK17-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
1478 // CHECK17-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1479 // CHECK17-NEXT: [[TMP:%.*]] = alloca i32, align 4
1480 // CHECK17-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1481 // CHECK17-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1482 // CHECK17-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1483 // CHECK17-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1484 // CHECK17-NEXT: [[I:%.*]] = alloca i32, align 4
1485 // CHECK17-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1486 // CHECK17-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1487 // CHECK17-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
1488 // CHECK17-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
1489 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1490 // CHECK17-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1491 // CHECK17-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1492 // CHECK17-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1493 // CHECK17-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1494 // CHECK17-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1495 // 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)
1496 // CHECK17-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1497 // CHECK17-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1498 // CHECK17-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1499 // CHECK17: cond.true:
1500 // CHECK17-NEXT: br label [[COND_END:%.*]]
1501 // CHECK17: cond.false:
1502 // CHECK17-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1503 // CHECK17-NEXT: br label [[COND_END]]
1504 // CHECK17: cond.end:
1505 // CHECK17-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1506 // CHECK17-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1507 // CHECK17-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1508 // CHECK17-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1509 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1510 // CHECK17: omp.inner.for.cond:
1511 // CHECK17-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1512 // CHECK17-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1513 // CHECK17-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1514 // CHECK17-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1515 // CHECK17: omp.inner.for.body:
1516 // CHECK17-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1517 // CHECK17-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1518 // CHECK17-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1519 // CHECK17-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1520 // CHECK17-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1521 // CHECK17-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
1522 // CHECK17-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
1523 // CHECK17-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
1524 // CHECK17-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1525 // CHECK17-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1526 // CHECK17: omp.body.continue:
1527 // CHECK17-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1528 // CHECK17: omp.inner.for.inc:
1529 // CHECK17-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1530 // CHECK17-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
1531 // CHECK17-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
1532 // CHECK17-NEXT: br label [[OMP_INNER_FOR_COND]]
1533 // CHECK17: omp.inner.for.end:
1534 // CHECK17-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1535 // CHECK17: omp.loop.exit:
1536 // CHECK17-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1537 // CHECK17-NEXT: ret void
1538 //
1539 //
1540 // CHECK17-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1541 // CHECK17-SAME: () #[[ATTR3:[0-9]+]] {
1542 // CHECK17-NEXT: entry:
1543 // CHECK17-NEXT: call void @__tgt_register_requires(i64 1)
1544 // CHECK17-NEXT: ret void
1545 //
1546 //
1547 // CHECK19-LABEL: define {{[^@]+}}@_Z21teams_template_structv
1548 // CHECK19-SAME: () #[[ATTR0:[0-9]+]] {
1549 // CHECK19-NEXT: entry:
1550 // CHECK19-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
1551 // CHECK19-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
1552 // CHECK19-NEXT: ret i32 [[CALL]]
1553 //
1554 //
1555 // CHECK19-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
1556 // CHECK19-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
1557 // CHECK19-NEXT: entry:
1558 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1559 // CHECK19-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
1560 // CHECK19-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
1561 // CHECK19-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
1562 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
1563 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1564 // CHECK19-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1565 // CHECK19-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
1566 // CHECK19-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1567 // CHECK19-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
1568 // CHECK19-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
1569 // CHECK19-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1570 // CHECK19-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
1571 // CHECK19-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
1572 // CHECK19-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1573 // CHECK19-NEXT: store i8* null, i8** [[TMP4]], align 4
1574 // CHECK19-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1575 // CHECK19-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1576 // CHECK19-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1577 // CHECK19-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1578 // CHECK19-NEXT: store i32 1, i32* [[TMP7]], align 4
1579 // CHECK19-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1580 // CHECK19-NEXT: store i32 1, i32* [[TMP8]], align 4
1581 // CHECK19-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1582 // CHECK19-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4
1583 // CHECK19-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1584 // CHECK19-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4
1585 // CHECK19-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1586 // CHECK19-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 4
1587 // CHECK19-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1588 // CHECK19-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 4
1589 // CHECK19-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1590 // CHECK19-NEXT: store i8** null, i8*** [[TMP13]], align 4
1591 // CHECK19-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1592 // CHECK19-NEXT: store i8** null, i8*** [[TMP14]], align 4
1593 // CHECK19-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1594 // CHECK19-NEXT: store i64 123, i64* [[TMP15]], align 8
1595 // CHECK19-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1596 // CHECK19-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
1597 // CHECK19-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1598 // CHECK19: omp_offload.failed:
1599 // CHECK19-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
1600 // CHECK19-NEXT: br label [[OMP_OFFLOAD_CONT]]
1601 // CHECK19: omp_offload.cont:
1602 // CHECK19-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
1603 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A2]], i32 0, i32 0
1604 // CHECK19-NEXT: [[TMP18:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
1605 // CHECK19-NEXT: ret i32 [[TMP18]]
1606 //
1607 //
1608 // CHECK19-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l109
1609 // CHECK19-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
1610 // CHECK19-NEXT: entry:
1611 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1612 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1613 // CHECK19-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1614 // 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]])
1615 // CHECK19-NEXT: ret void
1616 //
1617 //
1618 // CHECK19-LABEL: define {{[^@]+}}@.omp_outlined.
1619 // CHECK19-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1620 // CHECK19-NEXT: entry:
1621 // CHECK19-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1622 // CHECK19-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1623 // CHECK19-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1624 // CHECK19-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1625 // CHECK19-NEXT: [[TMP:%.*]] = alloca i32, align 4
1626 // CHECK19-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1627 // CHECK19-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1628 // CHECK19-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1629 // CHECK19-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1630 // CHECK19-NEXT: [[I:%.*]] = alloca i32, align 4
1631 // CHECK19-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1632 // CHECK19-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1633 // CHECK19-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1634 // CHECK19-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1635 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1636 // CHECK19-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1637 // CHECK19-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1638 // CHECK19-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1639 // CHECK19-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1640 // CHECK19-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1641 // 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)
1642 // CHECK19-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1643 // CHECK19-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1644 // CHECK19-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1645 // CHECK19: cond.true:
1646 // CHECK19-NEXT: br label [[COND_END:%.*]]
1647 // CHECK19: cond.false:
1648 // CHECK19-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1649 // CHECK19-NEXT: br label [[COND_END]]
1650 // CHECK19: cond.end:
1651 // CHECK19-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1652 // CHECK19-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1653 // CHECK19-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1654 // CHECK19-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1655 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1656 // CHECK19: omp.inner.for.cond:
1657 // CHECK19-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1658 // CHECK19-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1659 // CHECK19-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1660 // CHECK19-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1661 // CHECK19: omp.inner.for.body:
1662 // CHECK19-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1663 // CHECK19-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1664 // CHECK19-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1665 // CHECK19-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1666 // CHECK19-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1667 // CHECK19-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
1668 // CHECK19-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP9]]
1669 // CHECK19-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1670 // CHECK19-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1671 // CHECK19: omp.body.continue:
1672 // CHECK19-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1673 // CHECK19: omp.inner.for.inc:
1674 // CHECK19-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1675 // CHECK19-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
1676 // CHECK19-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
1677 // CHECK19-NEXT: br label [[OMP_INNER_FOR_COND]]
1678 // CHECK19: omp.inner.for.end:
1679 // CHECK19-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1680 // CHECK19: omp.loop.exit:
1681 // CHECK19-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1682 // CHECK19-NEXT: ret void
1683 //
1684 //
1685 // CHECK19-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1686 // CHECK19-SAME: () #[[ATTR3:[0-9]+]] {
1687 // CHECK19-NEXT: entry:
1688 // CHECK19-NEXT: call void @__tgt_register_requires(i64 1)
1689 // CHECK19-NEXT: ret void
1690 //
1691 //
1692 // CHECK25-LABEL: define {{[^@]+}}@main
1693 // CHECK25-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1694 // CHECK25-NEXT: entry:
1695 // CHECK25-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
1696 // CHECK25-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
1697 // CHECK25-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
1698 // CHECK25-NEXT: [[N:%.*]] = alloca i32, align 4
1699 // CHECK25-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
1700 // CHECK25-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1701 // CHECK25-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
1702 // CHECK25-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1703 // CHECK25-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1704 // CHECK25-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1705 // CHECK25-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
1706 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4
1707 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1708 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1709 // CHECK25-NEXT: store i32 0, i32* [[RETVAL]], align 4
1710 // CHECK25-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1711 // CHECK25-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
1712 // CHECK25-NEXT: store i32 100, i32* [[N]], align 4
1713 // CHECK25-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1714 // CHECK25-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
1715 // CHECK25-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
1716 // CHECK25-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
1717 // CHECK25-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
1718 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
1719 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
1720 // CHECK25-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
1721 // CHECK25-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
1722 // CHECK25-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
1723 // CHECK25-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
1724 // CHECK25-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1725 // CHECK25-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes to i8*), i64 24, i1 false)
1726 // CHECK25-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1727 // CHECK25-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
1728 // CHECK25-NEXT: store i64 [[TMP4]], i64* [[TMP8]], align 8
1729 // CHECK25-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1730 // CHECK25-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
1731 // CHECK25-NEXT: store i64 [[TMP4]], i64* [[TMP10]], align 8
1732 // CHECK25-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1733 // CHECK25-NEXT: store i8* null, i8** [[TMP11]], align 8
1734 // CHECK25-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1735 // CHECK25-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1736 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
1737 // CHECK25-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1738 // CHECK25-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
1739 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
1740 // CHECK25-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1741 // CHECK25-NEXT: store i8* null, i8** [[TMP16]], align 8
1742 // CHECK25-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1743 // CHECK25-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
1744 // CHECK25-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 8
1745 // CHECK25-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1746 // CHECK25-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
1747 // CHECK25-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 8
1748 // CHECK25-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1749 // CHECK25-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 8
1750 // CHECK25-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1751 // CHECK25-NEXT: store i8* null, i8** [[TMP22]], align 8
1752 // CHECK25-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1753 // CHECK25-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1754 // CHECK25-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1755 // CHECK25-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
1756 // CHECK25-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
1757 // CHECK25-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1758 // CHECK25-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
1759 // CHECK25-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1760 // CHECK25-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1761 // CHECK25-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1762 // CHECK25-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1763 // CHECK25-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
1764 // CHECK25-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
1765 // CHECK25-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1766 // CHECK25-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1767 // CHECK25-NEXT: store i32 1, i32* [[TMP30]], align 4
1768 // CHECK25-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1769 // CHECK25-NEXT: store i32 3, i32* [[TMP31]], align 4
1770 // CHECK25-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1771 // CHECK25-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 8
1772 // CHECK25-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1773 // CHECK25-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 8
1774 // CHECK25-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1775 // CHECK25-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 8
1776 // CHECK25-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1777 // CHECK25-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 8
1778 // CHECK25-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1779 // CHECK25-NEXT: store i8** null, i8*** [[TMP36]], align 8
1780 // CHECK25-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1781 // CHECK25-NEXT: store i8** null, i8*** [[TMP37]], align 8
1782 // CHECK25-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1783 // CHECK25-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
1784 // CHECK25-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1785 // CHECK25-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
1786 // CHECK25-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1787 // CHECK25: omp_offload.failed:
1788 // CHECK25-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
1789 // CHECK25-NEXT: br label [[OMP_OFFLOAD_CONT]]
1790 // CHECK25: omp_offload.cont:
1791 // CHECK25-NEXT: [[TMP41:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1792 // CHECK25-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP41]])
1793 // CHECK25-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
1794 // CHECK25-NEXT: [[TMP42:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1795 // CHECK25-NEXT: call void @llvm.stackrestore(i8* [[TMP42]])
1796 // CHECK25-NEXT: [[TMP43:%.*]] = load i32, i32* [[RETVAL]], align 4
1797 // CHECK25-NEXT: ret i32 [[TMP43]]
1798 //
1799 //
1800 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162
1801 // CHECK25-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1802 // CHECK25-NEXT: entry:
1803 // CHECK25-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
1804 // CHECK25-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1805 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1806 // CHECK25-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
1807 // CHECK25-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1808 // CHECK25-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1809 // CHECK25-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
1810 // CHECK25-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1811 // CHECK25-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1812 // 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]])
1813 // CHECK25-NEXT: ret void
1814 //
1815 //
1816 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined.
1817 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1818 // CHECK25-NEXT: entry:
1819 // CHECK25-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1820 // CHECK25-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1821 // CHECK25-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
1822 // CHECK25-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1823 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1824 // CHECK25-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1825 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4
1826 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1827 // CHECK25-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1828 // CHECK25-NEXT: [[I:%.*]] = alloca i32, align 4
1829 // CHECK25-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1830 // CHECK25-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1831 // CHECK25-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1832 // CHECK25-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1833 // CHECK25-NEXT: [[I3:%.*]] = alloca i32, align 4
1834 // CHECK25-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1835 // CHECK25-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1836 // CHECK25-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
1837 // CHECK25-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1838 // CHECK25-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1839 // CHECK25-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
1840 // CHECK25-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1841 // CHECK25-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1842 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1843 // CHECK25-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1844 // CHECK25-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1845 // CHECK25-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1846 // CHECK25-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1847 // CHECK25-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1848 // CHECK25-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1849 // CHECK25-NEXT: store i32 0, i32* [[I]], align 4
1850 // CHECK25-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1851 // CHECK25-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1852 // CHECK25-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1853 // CHECK25: omp.precond.then:
1854 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1855 // CHECK25-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1856 // CHECK25-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
1857 // CHECK25-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1858 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1859 // CHECK25-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1860 // CHECK25-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
1861 // 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)
1862 // CHECK25-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1863 // CHECK25-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1864 // CHECK25-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1865 // CHECK25-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1866 // CHECK25: cond.true:
1867 // CHECK25-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1868 // CHECK25-NEXT: br label [[COND_END:%.*]]
1869 // CHECK25: cond.false:
1870 // CHECK25-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1871 // CHECK25-NEXT: br label [[COND_END]]
1872 // CHECK25: cond.end:
1873 // CHECK25-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1874 // CHECK25-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1875 // CHECK25-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1876 // CHECK25-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1877 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1878 // CHECK25: omp.inner.for.cond:
1879 // CHECK25-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1880 // CHECK25-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1881 // CHECK25-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1882 // CHECK25-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1883 // CHECK25: omp.inner.for.body:
1884 // CHECK25-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1885 // CHECK25-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
1886 // CHECK25-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1887 // CHECK25-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
1888 // CHECK25-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
1889 // CHECK25-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
1890 // CHECK25-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
1891 // CHECK25-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1892 // CHECK25-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1893 // CHECK25: omp.body.continue:
1894 // CHECK25-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1895 // CHECK25: omp.inner.for.inc:
1896 // CHECK25-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1897 // CHECK25-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
1898 // CHECK25-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1899 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND]]
1900 // CHECK25: omp.inner.for.end:
1901 // CHECK25-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1902 // CHECK25: omp.loop.exit:
1903 // CHECK25-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1904 // CHECK25-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
1905 // CHECK25-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
1906 // CHECK25-NEXT: br label [[OMP_PRECOND_END]]
1907 // CHECK25: omp.precond.end:
1908 // CHECK25-NEXT: ret void
1909 //
1910 //
1911 // CHECK25-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
1912 // CHECK25-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
1913 // CHECK25-NEXT: entry:
1914 // CHECK25-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
1915 // CHECK25-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
1916 // CHECK25-NEXT: [[TE:%.*]] = alloca i32, align 4
1917 // CHECK25-NEXT: [[TH:%.*]] = alloca i32, align 4
1918 // CHECK25-NEXT: [[TE_CASTED:%.*]] = alloca i64, align 8
1919 // CHECK25-NEXT: [[TH_CASTED:%.*]] = alloca i64, align 8
1920 // CHECK25-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1921 // CHECK25-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1922 // CHECK25-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1923 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4
1924 // CHECK25-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1925 // CHECK25-NEXT: store i32 0, i32* [[TE]], align 4
1926 // CHECK25-NEXT: store i32 128, i32* [[TH]], align 4
1927 // CHECK25-NEXT: [[TMP0:%.*]] = load i32, i32* [[TE]], align 4
1928 // CHECK25-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_CASTED]] to i32*
1929 // CHECK25-NEXT: store i32 [[TMP0]], i32* [[CONV]], align 4
1930 // CHECK25-NEXT: [[TMP1:%.*]] = load i64, i64* [[TE_CASTED]], align 8
1931 // CHECK25-NEXT: [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
1932 // CHECK25-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_CASTED]] to i32*
1933 // CHECK25-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
1934 // CHECK25-NEXT: [[TMP3:%.*]] = load i64, i64* [[TH_CASTED]], align 8
1935 // CHECK25-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1936 // CHECK25-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
1937 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[TMP5]], align 8
1938 // CHECK25-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1939 // CHECK25-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
1940 // CHECK25-NEXT: store i64 [[TMP1]], i64* [[TMP7]], align 8
1941 // CHECK25-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1942 // CHECK25-NEXT: store i8* null, i8** [[TMP8]], align 8
1943 // CHECK25-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1944 // CHECK25-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
1945 // CHECK25-NEXT: store i64 [[TMP3]], i64* [[TMP10]], align 8
1946 // CHECK25-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1947 // CHECK25-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i64*
1948 // CHECK25-NEXT: store i64 [[TMP3]], i64* [[TMP12]], align 8
1949 // CHECK25-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1950 // CHECK25-NEXT: store i8* null, i8** [[TMP13]], align 8
1951 // CHECK25-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1952 // CHECK25-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [10 x i32]**
1953 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP15]], align 8
1954 // CHECK25-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1955 // CHECK25-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [10 x i32]**
1956 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP17]], align 8
1957 // CHECK25-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1958 // CHECK25-NEXT: store i8* null, i8** [[TMP18]], align 8
1959 // CHECK25-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1960 // CHECK25-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1961 // CHECK25-NEXT: [[TMP21:%.*]] = load i32, i32* [[TE]], align 4
1962 // CHECK25-NEXT: [[TMP22:%.*]] = load i32, i32* [[TH]], align 4
1963 // CHECK25-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1964 // CHECK25-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1965 // CHECK25-NEXT: store i32 1, i32* [[TMP23]], align 4
1966 // CHECK25-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1967 // CHECK25-NEXT: store i32 3, i32* [[TMP24]], align 4
1968 // CHECK25-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1969 // CHECK25-NEXT: store i8** [[TMP19]], i8*** [[TMP25]], align 8
1970 // CHECK25-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1971 // CHECK25-NEXT: store i8** [[TMP20]], i8*** [[TMP26]], align 8
1972 // CHECK25-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1973 // CHECK25-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP27]], align 8
1974 // CHECK25-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1975 // CHECK25-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP28]], align 8
1976 // CHECK25-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1977 // CHECK25-NEXT: store i8** null, i8*** [[TMP29]], align 8
1978 // CHECK25-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1979 // CHECK25-NEXT: store i8** null, i8*** [[TMP30]], align 8
1980 // CHECK25-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1981 // CHECK25-NEXT: store i64 10, i64* [[TMP31]], align 8
1982 // CHECK25-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 [[TMP21]], i32 [[TMP22]], i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1983 // CHECK25-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
1984 // CHECK25-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1985 // CHECK25: omp_offload.failed:
1986 // CHECK25-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151(i64 [[TMP1]], i64 [[TMP3]], [10 x i32]* [[A]]) #[[ATTR3]]
1987 // CHECK25-NEXT: br label [[OMP_OFFLOAD_CONT]]
1988 // CHECK25: omp_offload.cont:
1989 // CHECK25-NEXT: ret i32 0
1990 //
1991 //
1992 // CHECK25-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151
1993 // CHECK25-SAME: (i64 noundef [[TE:%.*]], i64 noundef [[TH:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1994 // CHECK25-NEXT: entry:
1995 // CHECK25-NEXT: [[TE_ADDR:%.*]] = alloca i64, align 8
1996 // CHECK25-NEXT: [[TH_ADDR:%.*]] = alloca i64, align 8
1997 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
1998 // CHECK25-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
1999 // CHECK25-NEXT: store i64 [[TE]], i64* [[TE_ADDR]], align 8
2000 // CHECK25-NEXT: store i64 [[TH]], i64* [[TH_ADDR]], align 8
2001 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2002 // CHECK25-NEXT: [[CONV:%.*]] = bitcast i64* [[TE_ADDR]] to i32*
2003 // CHECK25-NEXT: [[CONV1:%.*]] = bitcast i64* [[TH_ADDR]] to i32*
2004 // CHECK25-NEXT: [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2005 // CHECK25-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
2006 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[CONV1]], align 4
2007 // CHECK25-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
2008 // 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]])
2009 // CHECK25-NEXT: ret void
2010 //
2011 //
2012 // CHECK25-LABEL: define {{[^@]+}}@.omp_outlined..1
2013 // CHECK25-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2014 // CHECK25-NEXT: entry:
2015 // CHECK25-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2016 // CHECK25-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2017 // CHECK25-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2018 // CHECK25-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2019 // CHECK25-NEXT: [[TMP:%.*]] = alloca i32, align 4
2020 // CHECK25-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2021 // CHECK25-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2022 // CHECK25-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2023 // CHECK25-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2024 // CHECK25-NEXT: [[I:%.*]] = alloca i32, align 4
2025 // CHECK25-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2026 // CHECK25-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2027 // CHECK25-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2028 // CHECK25-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2029 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2030 // CHECK25-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
2031 // CHECK25-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2032 // CHECK25-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2033 // CHECK25-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2034 // CHECK25-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2035 // 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)
2036 // CHECK25-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2037 // CHECK25-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2038 // CHECK25-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2039 // CHECK25: cond.true:
2040 // CHECK25-NEXT: br label [[COND_END:%.*]]
2041 // CHECK25: cond.false:
2042 // CHECK25-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2043 // CHECK25-NEXT: br label [[COND_END]]
2044 // CHECK25: cond.end:
2045 // CHECK25-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2046 // CHECK25-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2047 // CHECK25-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2048 // CHECK25-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2049 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2050 // CHECK25: omp.inner.for.cond:
2051 // CHECK25-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2052 // CHECK25-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2053 // CHECK25-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2054 // CHECK25-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2055 // CHECK25: omp.inner.for.body:
2056 // CHECK25-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2057 // CHECK25-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2058 // CHECK25-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2059 // CHECK25-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2060 // CHECK25-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
2061 // CHECK25-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
2062 // CHECK25-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
2063 // CHECK25-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2064 // CHECK25-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2065 // CHECK25: omp.body.continue:
2066 // CHECK25-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2067 // CHECK25: omp.inner.for.inc:
2068 // CHECK25-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2069 // CHECK25-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
2070 // CHECK25-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2071 // CHECK25-NEXT: br label [[OMP_INNER_FOR_COND]]
2072 // CHECK25: omp.inner.for.end:
2073 // CHECK25-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2074 // CHECK25: omp.loop.exit:
2075 // CHECK25-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2076 // CHECK25-NEXT: ret void
2077 //
2078 //
2079 // CHECK25-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2080 // CHECK25-SAME: () #[[ATTR6:[0-9]+]] {
2081 // CHECK25-NEXT: entry:
2082 // CHECK25-NEXT: call void @__tgt_register_requires(i64 1)
2083 // CHECK25-NEXT: ret void
2084 //
2085 //
2086 // CHECK27-LABEL: define {{[^@]+}}@main
2087 // CHECK27-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
2088 // CHECK27-NEXT: entry:
2089 // CHECK27-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
2090 // CHECK27-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
2091 // CHECK27-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
2092 // CHECK27-NEXT: [[N:%.*]] = alloca i32, align 4
2093 // CHECK27-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
2094 // CHECK27-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
2095 // CHECK27-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
2096 // CHECK27-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
2097 // CHECK27-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
2098 // CHECK27-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
2099 // CHECK27-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
2100 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4
2101 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2102 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2103 // CHECK27-NEXT: store i32 0, i32* [[RETVAL]], align 4
2104 // CHECK27-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
2105 // CHECK27-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
2106 // CHECK27-NEXT: store i32 100, i32* [[N]], align 4
2107 // CHECK27-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
2108 // CHECK27-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
2109 // CHECK27-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
2110 // CHECK27-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
2111 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
2112 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
2113 // CHECK27-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
2114 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
2115 // CHECK27-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
2116 // CHECK27-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
2117 // CHECK27-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
2118 // CHECK27-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP6]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes to i8*), i32 24, i1 false)
2119 // CHECK27-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2120 // CHECK27-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
2121 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[TMP8]], align 4
2122 // CHECK27-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2123 // CHECK27-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
2124 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
2125 // CHECK27-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2126 // CHECK27-NEXT: store i8* null, i8** [[TMP11]], align 4
2127 // CHECK27-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2128 // CHECK27-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
2129 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
2130 // CHECK27-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2131 // CHECK27-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
2132 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
2133 // CHECK27-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
2134 // CHECK27-NEXT: store i8* null, i8** [[TMP16]], align 4
2135 // CHECK27-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2136 // CHECK27-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
2137 // CHECK27-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 4
2138 // CHECK27-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2139 // CHECK27-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
2140 // CHECK27-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 4
2141 // CHECK27-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
2142 // CHECK27-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 4
2143 // CHECK27-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
2144 // CHECK27-NEXT: store i8* null, i8** [[TMP22]], align 4
2145 // CHECK27-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2146 // CHECK27-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2147 // CHECK27-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2148 // CHECK27-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
2149 // CHECK27-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
2150 // CHECK27-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2151 // CHECK27-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
2152 // CHECK27-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2153 // CHECK27-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2154 // CHECK27-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2155 // CHECK27-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2156 // CHECK27-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
2157 // CHECK27-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
2158 // CHECK27-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2159 // CHECK27-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
2160 // CHECK27-NEXT: store i32 1, i32* [[TMP30]], align 4
2161 // CHECK27-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
2162 // CHECK27-NEXT: store i32 3, i32* [[TMP31]], align 4
2163 // CHECK27-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
2164 // CHECK27-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 4
2165 // CHECK27-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
2166 // CHECK27-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 4
2167 // CHECK27-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
2168 // CHECK27-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 4
2169 // CHECK27-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
2170 // CHECK27-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 4
2171 // CHECK27-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
2172 // CHECK27-NEXT: store i8** null, i8*** [[TMP36]], align 4
2173 // CHECK27-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
2174 // CHECK27-NEXT: store i8** null, i8*** [[TMP37]], align 4
2175 // CHECK27-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
2176 // CHECK27-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
2177 // CHECK27-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
2178 // CHECK27-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
2179 // CHECK27-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2180 // CHECK27: omp_offload.failed:
2181 // CHECK27-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
2182 // CHECK27-NEXT: br label [[OMP_OFFLOAD_CONT]]
2183 // CHECK27: omp_offload.cont:
2184 // CHECK27-NEXT: [[TMP41:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2185 // CHECK27-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP41]])
2186 // CHECK27-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
2187 // CHECK27-NEXT: [[TMP42:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
2188 // CHECK27-NEXT: call void @llvm.stackrestore(i8* [[TMP42]])
2189 // CHECK27-NEXT: [[TMP43:%.*]] = load i32, i32* [[RETVAL]], align 4
2190 // CHECK27-NEXT: ret i32 [[TMP43]]
2191 //
2192 //
2193 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l162
2194 // CHECK27-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
2195 // CHECK27-NEXT: entry:
2196 // CHECK27-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
2197 // CHECK27-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
2198 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
2199 // CHECK27-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
2200 // CHECK27-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2201 // CHECK27-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
2202 // CHECK27-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2203 // CHECK27-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
2204 // 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]])
2205 // CHECK27-NEXT: ret void
2206 //
2207 //
2208 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined.
2209 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2210 // CHECK27-NEXT: entry:
2211 // CHECK27-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2212 // CHECK27-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2213 // CHECK27-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
2214 // CHECK27-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
2215 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
2216 // CHECK27-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2217 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4
2218 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2219 // CHECK27-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2220 // CHECK27-NEXT: [[I:%.*]] = alloca i32, align 4
2221 // CHECK27-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2222 // CHECK27-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2223 // CHECK27-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2224 // CHECK27-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2225 // CHECK27-NEXT: [[I3:%.*]] = alloca i32, align 4
2226 // CHECK27-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2227 // CHECK27-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2228 // CHECK27-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
2229 // CHECK27-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2230 // CHECK27-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
2231 // CHECK27-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
2232 // CHECK27-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2233 // CHECK27-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
2234 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2235 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
2236 // CHECK27-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2237 // CHECK27-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2238 // CHECK27-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2239 // CHECK27-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2240 // CHECK27-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2241 // CHECK27-NEXT: store i32 0, i32* [[I]], align 4
2242 // CHECK27-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2243 // CHECK27-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2244 // CHECK27-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2245 // CHECK27: omp.precond.then:
2246 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2247 // CHECK27-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2248 // CHECK27-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
2249 // CHECK27-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2250 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2251 // CHECK27-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2252 // CHECK27-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
2253 // 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)
2254 // CHECK27-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2255 // CHECK27-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2256 // CHECK27-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
2257 // CHECK27-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2258 // CHECK27: cond.true:
2259 // CHECK27-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2260 // CHECK27-NEXT: br label [[COND_END:%.*]]
2261 // CHECK27: cond.false:
2262 // CHECK27-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2263 // CHECK27-NEXT: br label [[COND_END]]
2264 // CHECK27: cond.end:
2265 // CHECK27-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2266 // CHECK27-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2267 // CHECK27-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2268 // CHECK27-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
2269 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2270 // CHECK27: omp.inner.for.cond:
2271 // CHECK27-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2272 // CHECK27-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2273 // CHECK27-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2274 // CHECK27-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2275 // CHECK27: omp.inner.for.body:
2276 // CHECK27-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2277 // CHECK27-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
2278 // CHECK27-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2279 // CHECK27-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
2280 // CHECK27-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
2281 // CHECK27-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP17]]
2282 // CHECK27-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2283 // CHECK27-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2284 // CHECK27: omp.body.continue:
2285 // CHECK27-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2286 // CHECK27: omp.inner.for.inc:
2287 // CHECK27-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2288 // CHECK27-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
2289 // CHECK27-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
2290 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND]]
2291 // CHECK27: omp.inner.for.end:
2292 // CHECK27-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2293 // CHECK27: omp.loop.exit:
2294 // CHECK27-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2295 // CHECK27-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
2296 // CHECK27-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
2297 // CHECK27-NEXT: br label [[OMP_PRECOND_END]]
2298 // CHECK27: omp.precond.end:
2299 // CHECK27-NEXT: ret void
2300 //
2301 //
2302 // CHECK27-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
2303 // CHECK27-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
2304 // CHECK27-NEXT: entry:
2305 // CHECK27-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
2306 // CHECK27-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
2307 // CHECK27-NEXT: [[TE:%.*]] = alloca i32, align 4
2308 // CHECK27-NEXT: [[TH:%.*]] = alloca i32, align 4
2309 // CHECK27-NEXT: [[TE_CASTED:%.*]] = alloca i32, align 4
2310 // CHECK27-NEXT: [[TH_CASTED:%.*]] = alloca i32, align 4
2311 // CHECK27-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
2312 // CHECK27-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
2313 // CHECK27-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
2314 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4
2315 // CHECK27-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
2316 // CHECK27-NEXT: store i32 0, i32* [[TE]], align 4
2317 // CHECK27-NEXT: store i32 128, i32* [[TH]], align 4
2318 // CHECK27-NEXT: [[TMP0:%.*]] = load i32, i32* [[TE]], align 4
2319 // CHECK27-NEXT: store i32 [[TMP0]], i32* [[TE_CASTED]], align 4
2320 // CHECK27-NEXT: [[TMP1:%.*]] = load i32, i32* [[TE_CASTED]], align 4
2321 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[TH]], align 4
2322 // CHECK27-NEXT: store i32 [[TMP2]], i32* [[TH_CASTED]], align 4
2323 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_CASTED]], align 4
2324 // CHECK27-NEXT: [[TMP4:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2325 // CHECK27-NEXT: [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32*
2326 // CHECK27-NEXT: store i32 [[TMP1]], i32* [[TMP5]], align 4
2327 // CHECK27-NEXT: [[TMP6:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2328 // CHECK27-NEXT: [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i32*
2329 // CHECK27-NEXT: store i32 [[TMP1]], i32* [[TMP7]], align 4
2330 // CHECK27-NEXT: [[TMP8:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2331 // CHECK27-NEXT: store i8* null, i8** [[TMP8]], align 4
2332 // CHECK27-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2333 // CHECK27-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
2334 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
2335 // CHECK27-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2336 // CHECK27-NEXT: [[TMP12:%.*]] = bitcast i8** [[TMP11]] to i32*
2337 // CHECK27-NEXT: store i32 [[TMP3]], i32* [[TMP12]], align 4
2338 // CHECK27-NEXT: [[TMP13:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
2339 // CHECK27-NEXT: store i8* null, i8** [[TMP13]], align 4
2340 // CHECK27-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2341 // CHECK27-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to [10 x i32]**
2342 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP15]], align 4
2343 // CHECK27-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2344 // CHECK27-NEXT: [[TMP17:%.*]] = bitcast i8** [[TMP16]] to [10 x i32]**
2345 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP17]], align 4
2346 // CHECK27-NEXT: [[TMP18:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
2347 // CHECK27-NEXT: store i8* null, i8** [[TMP18]], align 4
2348 // CHECK27-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2349 // CHECK27-NEXT: [[TMP20:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2350 // CHECK27-NEXT: [[TMP21:%.*]] = load i32, i32* [[TE]], align 4
2351 // CHECK27-NEXT: [[TMP22:%.*]] = load i32, i32* [[TH]], align 4
2352 // CHECK27-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2353 // CHECK27-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
2354 // CHECK27-NEXT: store i32 1, i32* [[TMP23]], align 4
2355 // CHECK27-NEXT: [[TMP24:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
2356 // CHECK27-NEXT: store i32 3, i32* [[TMP24]], align 4
2357 // CHECK27-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
2358 // CHECK27-NEXT: store i8** [[TMP19]], i8*** [[TMP25]], align 4
2359 // CHECK27-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
2360 // CHECK27-NEXT: store i8** [[TMP20]], i8*** [[TMP26]], align 4
2361 // CHECK27-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
2362 // CHECK27-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP27]], align 4
2363 // CHECK27-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
2364 // CHECK27-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP28]], align 4
2365 // CHECK27-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
2366 // CHECK27-NEXT: store i8** null, i8*** [[TMP29]], align 4
2367 // CHECK27-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
2368 // CHECK27-NEXT: store i8** null, i8*** [[TMP30]], align 4
2369 // CHECK27-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
2370 // CHECK27-NEXT: store i64 10, i64* [[TMP31]], align 8
2371 // CHECK27-NEXT: [[TMP32:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 [[TMP21]], i32 [[TMP22]], i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
2372 // CHECK27-NEXT: [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
2373 // CHECK27-NEXT: br i1 [[TMP33]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2374 // CHECK27: omp_offload.failed:
2375 // CHECK27-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151(i32 [[TMP1]], i32 [[TMP3]], [10 x i32]* [[A]]) #[[ATTR3]]
2376 // CHECK27-NEXT: br label [[OMP_OFFLOAD_CONT]]
2377 // CHECK27: omp_offload.cont:
2378 // CHECK27-NEXT: ret i32 0
2379 //
2380 //
2381 // CHECK27-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l151
2382 // CHECK27-SAME: (i32 noundef [[TE:%.*]], i32 noundef [[TH:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2383 // CHECK27-NEXT: entry:
2384 // CHECK27-NEXT: [[TE_ADDR:%.*]] = alloca i32, align 4
2385 // CHECK27-NEXT: [[TH_ADDR:%.*]] = alloca i32, align 4
2386 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
2387 // CHECK27-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
2388 // CHECK27-NEXT: store i32 [[TE]], i32* [[TE_ADDR]], align 4
2389 // CHECK27-NEXT: store i32 [[TH]], i32* [[TH_ADDR]], align 4
2390 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
2391 // CHECK27-NEXT: [[TMP1:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
2392 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[TE_ADDR]], align 4
2393 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[TH_ADDR]], align 4
2394 // CHECK27-NEXT: call void @__kmpc_push_num_teams(%struct.ident_t* @[[GLOB2]], i32 [[TMP0]], i32 [[TMP2]], i32 [[TMP3]])
2395 // 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]])
2396 // CHECK27-NEXT: ret void
2397 //
2398 //
2399 // CHECK27-LABEL: define {{[^@]+}}@.omp_outlined..1
2400 // CHECK27-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2401 // CHECK27-NEXT: entry:
2402 // CHECK27-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2403 // CHECK27-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2404 // CHECK27-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
2405 // CHECK27-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2406 // CHECK27-NEXT: [[TMP:%.*]] = alloca i32, align 4
2407 // CHECK27-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2408 // CHECK27-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2409 // CHECK27-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2410 // CHECK27-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2411 // CHECK27-NEXT: [[I:%.*]] = alloca i32, align 4
2412 // CHECK27-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2413 // CHECK27-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2414 // CHECK27-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
2415 // CHECK27-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
2416 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2417 // CHECK27-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
2418 // CHECK27-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2419 // CHECK27-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2420 // CHECK27-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2421 // CHECK27-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2422 // 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)
2423 // CHECK27-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2424 // CHECK27-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2425 // CHECK27-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2426 // CHECK27: cond.true:
2427 // CHECK27-NEXT: br label [[COND_END:%.*]]
2428 // CHECK27: cond.false:
2429 // CHECK27-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2430 // CHECK27-NEXT: br label [[COND_END]]
2431 // CHECK27: cond.end:
2432 // CHECK27-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2433 // CHECK27-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2434 // CHECK27-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2435 // CHECK27-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2436 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2437 // CHECK27: omp.inner.for.cond:
2438 // CHECK27-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2439 // CHECK27-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2440 // CHECK27-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2441 // CHECK27-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2442 // CHECK27: omp.inner.for.body:
2443 // CHECK27-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2444 // CHECK27-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2445 // CHECK27-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2446 // CHECK27-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2447 // CHECK27-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
2448 // CHECK27-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP9]]
2449 // CHECK27-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2450 // CHECK27-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2451 // CHECK27: omp.body.continue:
2452 // CHECK27-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2453 // CHECK27: omp.inner.for.inc:
2454 // CHECK27-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2455 // CHECK27-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
2456 // CHECK27-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2457 // CHECK27-NEXT: br label [[OMP_INNER_FOR_COND]]
2458 // CHECK27: omp.inner.for.end:
2459 // CHECK27-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2460 // CHECK27: omp.loop.exit:
2461 // CHECK27-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2462 // CHECK27-NEXT: ret void
2463 //
2464 //
2465 // CHECK27-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2466 // CHECK27-SAME: () #[[ATTR6:[0-9]+]] {
2467 // CHECK27-NEXT: entry:
2468 // CHECK27-NEXT: call void @__tgt_register_requires(i64 1)
2469 // CHECK27-NEXT: ret void
2470 //
2471