1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
2 // expected-no-diagnostics
3 #ifndef HEADER
4 #define HEADER
5
6 // Test host codegen.
7 // RUN: %clang_cc1 -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
8 // 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
9 // 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
10 // 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
11 // 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
12 // 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
13
14 // 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}}"
15 // 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
16 // 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}}"
17 // 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}}"
18 // 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
19 // 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}}"
20 #ifdef CK1
21
22 template <typename T, int X, long long Y>
23 struct SS{
24 T a[X];
25 float b;
fooSS26 int foo(void) {
27
28 #pragma omp target
29 #pragma omp teams distribute
30 for(int i = 0; i < X; i++) {
31 a[i] = (T)0;
32 }
33 #pragma omp target
34 #pragma omp teams distribute dist_schedule(static)
35 for(int i = 0; i < X; i++) {
36 a[i] = (T)0;
37 }
38 #pragma omp target
39 #pragma omp teams distribute dist_schedule(static, X/2)
40 for(int i = 0; i < X; i++) {
41 a[i] = (T)0;
42 }
43
44
45
46
47
48
49 return a[0];
50 }
51 };
52
teams_template_struct(void)53 int teams_template_struct(void) {
54 SS<int, 123, 456> V;
55 return V.foo();
56
57 }
58 #endif // CK1
59
60 // Test host codegen.
61 // 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
62 // 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
63 // 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
64 // 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
65 // 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
66 // 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
67
68 // 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}}"
69 // 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
70 // 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}}"
71 // 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}}"
72 // 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
73 // 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}}"
74 #ifdef CK2
75
76 template <typename T, int n>
tmain(T argc)77 int tmain(T argc) {
78 T a[n];
79 #pragma omp target
80 #pragma omp teams distribute
81 for(int i = 0; i < n; i++) {
82 a[i] = (T)0;
83 }
84 #pragma omp target
85 #pragma omp teams distribute dist_schedule(static)
86 for(int i = 0; i < n; i++) {
87 a[i] = (T)0;
88 }
89 #pragma omp target
90 #pragma omp teams distribute dist_schedule(static, n)
91 for(int i = 0; i < n; i++) {
92 a[i] = (T)0;
93 }
94 return 0;
95 }
96
main(int argc,char ** argv)97 int main (int argc, char **argv) {
98 int n = 100;
99 int a[n];
100 #pragma omp target
101 #pragma omp teams distribute
102 for(int i = 0; i < n; i++) {
103 a[i] = 0;
104 }
105 #pragma omp target
106 #pragma omp teams distribute dist_schedule(static)
107 for(int i = 0; i < n; i++) {
108 a[i] = 0;
109 }
110 #pragma omp target
111 #pragma omp teams distribute dist_schedule(static, n)
112 for(int i = 0; i < n; i++) {
113 a[i] = 0;
114 }
115 return tmain<int, 10>(argc);
116 }
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132 #endif // CK2
133 #endif // #ifndef HEADER
134 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
135 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
136 // CHECK1-NEXT: entry:
137 // CHECK1-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
138 // CHECK1-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
139 // CHECK1-NEXT: ret i32 [[CALL]]
140 //
141 //
142 // CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
143 // CHECK1-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
144 // CHECK1-NEXT: entry:
145 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
146 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
147 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
148 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
149 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
150 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
151 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
152 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
153 // CHECK1-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
154 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x i8*], align 8
155 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 8
156 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 8
157 // CHECK1-NEXT: [[_TMP14:%.*]] = alloca i32, align 4
158 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
159 // CHECK1-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
160 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
161 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
162 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
163 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
164 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
165 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
166 // CHECK1-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
167 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
168 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
169 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
170 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
171 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
172 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
173 // CHECK1-NEXT: store i32 1, i32* [[TMP7]], align 4
174 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
175 // CHECK1-NEXT: store i32 1, i32* [[TMP8]], align 4
176 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
177 // CHECK1-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8
178 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
179 // CHECK1-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8
180 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
181 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 8
182 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
183 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 8
184 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
185 // CHECK1-NEXT: store i8** null, i8*** [[TMP13]], align 8
186 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
187 // CHECK1-NEXT: store i8** null, i8*** [[TMP14]], align 8
188 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
189 // CHECK1-NEXT: store i64 123, i64* [[TMP15]], align 8
190 // CHECK1-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_l28.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
191 // CHECK1-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
192 // CHECK1-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
193 // CHECK1: omp_offload.failed:
194 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
195 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
196 // CHECK1: omp_offload.cont:
197 // CHECK1-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
198 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
199 // CHECK1-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
200 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
201 // CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
202 // CHECK1-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
203 // CHECK1-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP21]], align 8
204 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
205 // CHECK1-NEXT: store i8* null, i8** [[TMP22]], align 8
206 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
207 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
208 // CHECK1-NEXT: [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
209 // CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
210 // CHECK1-NEXT: store i32 1, i32* [[TMP25]], align 4
211 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
212 // CHECK1-NEXT: store i32 1, i32* [[TMP26]], align 4
213 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
214 // CHECK1-NEXT: store i8** [[TMP23]], i8*** [[TMP27]], align 8
215 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
216 // CHECK1-NEXT: store i8** [[TMP24]], i8*** [[TMP28]], align 8
217 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
218 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP29]], align 8
219 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
220 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP30]], align 8
221 // CHECK1-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
222 // CHECK1-NEXT: store i8** null, i8*** [[TMP31]], align 8
223 // CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
224 // CHECK1-NEXT: store i8** null, i8*** [[TMP32]], align 8
225 // CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
226 // CHECK1-NEXT: store i64 123, i64* [[TMP33]], align 8
227 // CHECK1-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
228 // CHECK1-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
229 // CHECK1-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
230 // CHECK1: omp_offload.failed8:
231 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33(%struct.SS* [[THIS1]]) #[[ATTR2]]
232 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT9]]
233 // CHECK1: omp_offload.cont9:
234 // CHECK1-NEXT: [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
235 // CHECK1-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
236 // CHECK1-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
237 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
238 // CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
239 // CHECK1-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
240 // CHECK1-NEXT: store [123 x i32]* [[A10]], [123 x i32]** [[TMP39]], align 8
241 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 0
242 // CHECK1-NEXT: store i8* null, i8** [[TMP40]], align 8
243 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
244 // CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
245 // CHECK1-NEXT: [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
246 // CHECK1-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
247 // CHECK1-NEXT: store i32 1, i32* [[TMP43]], align 4
248 // CHECK1-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
249 // CHECK1-NEXT: store i32 1, i32* [[TMP44]], align 4
250 // CHECK1-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
251 // CHECK1-NEXT: store i8** [[TMP41]], i8*** [[TMP45]], align 8
252 // CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
253 // CHECK1-NEXT: store i8** [[TMP42]], i8*** [[TMP46]], align 8
254 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
255 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64** [[TMP47]], align 8
256 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
257 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i64** [[TMP48]], align 8
258 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
259 // CHECK1-NEXT: store i8** null, i8*** [[TMP49]], align 8
260 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
261 // CHECK1-NEXT: store i8** null, i8*** [[TMP50]], align 8
262 // CHECK1-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
263 // CHECK1-NEXT: store i64 123, i64* [[TMP51]], align 8
264 // CHECK1-NEXT: [[TMP52:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
265 // CHECK1-NEXT: [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
266 // CHECK1-NEXT: br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
267 // CHECK1: omp_offload.failed16:
268 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38(%struct.SS* [[THIS1]]) #[[ATTR2]]
269 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT17]]
270 // CHECK1: omp_offload.cont17:
271 // CHECK1-NEXT: [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
272 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A18]], i64 0, i64 0
273 // CHECK1-NEXT: [[TMP54:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
274 // CHECK1-NEXT: ret i32 [[TMP54]]
275 //
276 //
277 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
278 // CHECK1-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
279 // CHECK1-NEXT: entry:
280 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
281 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
282 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
283 // CHECK1-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]])
284 // CHECK1-NEXT: ret void
285 //
286 //
287 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
288 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
289 // CHECK1-NEXT: entry:
290 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
291 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
292 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
293 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
294 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
295 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
296 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
297 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
298 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
299 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
300 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
301 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
302 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
303 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
304 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
305 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
306 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
307 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
308 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
309 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
310 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
311 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
312 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
313 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
314 // CHECK1: cond.true:
315 // CHECK1-NEXT: br label [[COND_END:%.*]]
316 // CHECK1: cond.false:
317 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
318 // CHECK1-NEXT: br label [[COND_END]]
319 // CHECK1: cond.end:
320 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
321 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
322 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
323 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
324 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
325 // CHECK1: omp.inner.for.cond:
326 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
327 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
328 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
329 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
330 // CHECK1: omp.inner.for.body:
331 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
332 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
333 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
334 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
335 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
336 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
337 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
338 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
339 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
340 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
341 // CHECK1: omp.body.continue:
342 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
343 // CHECK1: omp.inner.for.inc:
344 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
345 // CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
346 // CHECK1-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
347 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
348 // CHECK1: omp.inner.for.end:
349 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
350 // CHECK1: omp.loop.exit:
351 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
352 // CHECK1-NEXT: ret void
353 //
354 //
355 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33
356 // CHECK1-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
357 // CHECK1-NEXT: entry:
358 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
359 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
360 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
361 // CHECK1-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..1 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
362 // CHECK1-NEXT: ret void
363 //
364 //
365 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
366 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
367 // CHECK1-NEXT: entry:
368 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
369 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
370 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
371 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
372 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
373 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
374 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
375 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
376 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
377 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
378 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
379 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
380 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
381 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
382 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
383 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
384 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
385 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
386 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
387 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
388 // CHECK1-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)
389 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
390 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
391 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
392 // CHECK1: cond.true:
393 // CHECK1-NEXT: br label [[COND_END:%.*]]
394 // CHECK1: cond.false:
395 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
396 // CHECK1-NEXT: br label [[COND_END]]
397 // CHECK1: cond.end:
398 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
399 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
400 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
401 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
402 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
403 // CHECK1: omp.inner.for.cond:
404 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
405 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
406 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
407 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
408 // CHECK1: omp.inner.for.body:
409 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
410 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
411 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
412 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
413 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
414 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
415 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
416 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
417 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
418 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
419 // CHECK1: omp.body.continue:
420 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
421 // CHECK1: omp.inner.for.inc:
422 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
423 // CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
424 // CHECK1-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
425 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
426 // CHECK1: omp.inner.for.end:
427 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
428 // CHECK1: omp.loop.exit:
429 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
430 // CHECK1-NEXT: ret void
431 //
432 //
433 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38
434 // CHECK1-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
435 // CHECK1-NEXT: entry:
436 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
437 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
438 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
439 // CHECK1-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..4 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
440 // CHECK1-NEXT: ret void
441 //
442 //
443 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
444 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
445 // CHECK1-NEXT: entry:
446 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
447 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
448 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
449 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
450 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
451 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
452 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
453 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
454 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
455 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
456 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
457 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
458 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
459 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
460 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
461 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
462 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
463 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
464 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
465 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
466 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
467 // CHECK1-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
468 // CHECK1: omp.dispatch.cond:
469 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
470 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
471 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
472 // CHECK1: cond.true:
473 // CHECK1-NEXT: br label [[COND_END:%.*]]
474 // CHECK1: cond.false:
475 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
476 // CHECK1-NEXT: br label [[COND_END]]
477 // CHECK1: cond.end:
478 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
479 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
480 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
481 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
482 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
483 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
484 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
485 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
486 // CHECK1: omp.dispatch.body:
487 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
488 // CHECK1: omp.inner.for.cond:
489 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
490 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !8
491 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
492 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
493 // CHECK1: omp.inner.for.body:
494 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
495 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
496 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
497 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !8
498 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
499 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !8
500 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
501 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
502 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !8
503 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
504 // CHECK1: omp.body.continue:
505 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
506 // CHECK1: omp.inner.for.inc:
507 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
508 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
509 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !8
510 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
511 // CHECK1: omp.inner.for.end:
512 // CHECK1-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
513 // CHECK1: omp.dispatch.inc:
514 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
515 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
516 // CHECK1-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
517 // CHECK1-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
518 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
519 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
520 // CHECK1-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
521 // CHECK1-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
522 // CHECK1-NEXT: br label [[OMP_DISPATCH_COND]]
523 // CHECK1: omp.dispatch.end:
524 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
525 // CHECK1-NEXT: ret void
526 //
527 //
528 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
529 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
530 // CHECK1-NEXT: entry:
531 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
532 // CHECK1-NEXT: ret void
533 //
534 //
535 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
536 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
537 // CHECK3-NEXT: entry:
538 // CHECK3-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
539 // CHECK3-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
540 // CHECK3-NEXT: ret i32 [[CALL]]
541 //
542 //
543 // CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
544 // CHECK3-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
545 // CHECK3-NEXT: entry:
546 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
547 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
548 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
549 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
550 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
551 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
552 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
553 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
554 // CHECK3-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
555 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x i8*], align 4
556 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 4
557 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 4
558 // CHECK3-NEXT: [[_TMP14:%.*]] = alloca i32, align 4
559 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
560 // CHECK3-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
561 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
562 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
563 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
564 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
565 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
566 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
567 // CHECK3-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
568 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
569 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
570 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
571 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
572 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
573 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
574 // CHECK3-NEXT: store i32 1, i32* [[TMP7]], align 4
575 // CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
576 // CHECK3-NEXT: store i32 1, i32* [[TMP8]], align 4
577 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
578 // CHECK3-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4
579 // CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
580 // CHECK3-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4
581 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
582 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 4
583 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
584 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 4
585 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
586 // CHECK3-NEXT: store i8** null, i8*** [[TMP13]], align 4
587 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
588 // CHECK3-NEXT: store i8** null, i8*** [[TMP14]], align 4
589 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
590 // CHECK3-NEXT: store i64 123, i64* [[TMP15]], align 8
591 // CHECK3-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_l28.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
592 // CHECK3-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
593 // CHECK3-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
594 // CHECK3: omp_offload.failed:
595 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
596 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
597 // CHECK3: omp_offload.cont:
598 // CHECK3-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
599 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
600 // CHECK3-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
601 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
602 // CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
603 // CHECK3-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
604 // CHECK3-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP21]], align 4
605 // CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
606 // CHECK3-NEXT: store i8* null, i8** [[TMP22]], align 4
607 // CHECK3-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
608 // CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
609 // CHECK3-NEXT: [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
610 // CHECK3-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
611 // CHECK3-NEXT: store i32 1, i32* [[TMP25]], align 4
612 // CHECK3-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
613 // CHECK3-NEXT: store i32 1, i32* [[TMP26]], align 4
614 // CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
615 // CHECK3-NEXT: store i8** [[TMP23]], i8*** [[TMP27]], align 4
616 // CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
617 // CHECK3-NEXT: store i8** [[TMP24]], i8*** [[TMP28]], align 4
618 // CHECK3-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
619 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.2, i32 0, i32 0), i64** [[TMP29]], align 4
620 // CHECK3-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
621 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP30]], align 4
622 // CHECK3-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
623 // CHECK3-NEXT: store i8** null, i8*** [[TMP31]], align 4
624 // CHECK3-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
625 // CHECK3-NEXT: store i8** null, i8*** [[TMP32]], align 4
626 // CHECK3-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
627 // CHECK3-NEXT: store i64 123, i64* [[TMP33]], align 8
628 // CHECK3-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]])
629 // CHECK3-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
630 // CHECK3-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
631 // CHECK3: omp_offload.failed8:
632 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33(%struct.SS* [[THIS1]]) #[[ATTR2]]
633 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT9]]
634 // CHECK3: omp_offload.cont9:
635 // CHECK3-NEXT: [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
636 // CHECK3-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
637 // CHECK3-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
638 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
639 // CHECK3-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
640 // CHECK3-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
641 // CHECK3-NEXT: store [123 x i32]* [[A10]], [123 x i32]** [[TMP39]], align 4
642 // CHECK3-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i32 0, i32 0
643 // CHECK3-NEXT: store i8* null, i8** [[TMP40]], align 4
644 // CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
645 // CHECK3-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
646 // CHECK3-NEXT: [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
647 // CHECK3-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
648 // CHECK3-NEXT: store i32 1, i32* [[TMP43]], align 4
649 // CHECK3-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
650 // CHECK3-NEXT: store i32 1, i32* [[TMP44]], align 4
651 // CHECK3-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
652 // CHECK3-NEXT: store i8** [[TMP41]], i8*** [[TMP45]], align 4
653 // CHECK3-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
654 // CHECK3-NEXT: store i8** [[TMP42]], i8*** [[TMP46]], align 4
655 // CHECK3-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
656 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.5, i32 0, i32 0), i64** [[TMP47]], align 4
657 // CHECK3-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
658 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.6, i32 0, i32 0), i64** [[TMP48]], align 4
659 // CHECK3-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
660 // CHECK3-NEXT: store i8** null, i8*** [[TMP49]], align 4
661 // CHECK3-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
662 // CHECK3-NEXT: store i8** null, i8*** [[TMP50]], align 4
663 // CHECK3-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
664 // CHECK3-NEXT: store i64 123, i64* [[TMP51]], align 8
665 // CHECK3-NEXT: [[TMP52:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
666 // CHECK3-NEXT: [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
667 // CHECK3-NEXT: br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
668 // CHECK3: omp_offload.failed16:
669 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38(%struct.SS* [[THIS1]]) #[[ATTR2]]
670 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT17]]
671 // CHECK3: omp_offload.cont17:
672 // CHECK3-NEXT: [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
673 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A18]], i32 0, i32 0
674 // CHECK3-NEXT: [[TMP54:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
675 // CHECK3-NEXT: ret i32 [[TMP54]]
676 //
677 //
678 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
679 // CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
680 // CHECK3-NEXT: entry:
681 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
682 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
683 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
684 // CHECK3-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]])
685 // CHECK3-NEXT: ret void
686 //
687 //
688 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
689 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
690 // CHECK3-NEXT: entry:
691 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
692 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
693 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
694 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
695 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
696 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
697 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
698 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
699 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
700 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
701 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
702 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
703 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
704 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
705 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
706 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
707 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
708 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
709 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
710 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
711 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
712 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
713 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
714 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
715 // CHECK3: cond.true:
716 // CHECK3-NEXT: br label [[COND_END:%.*]]
717 // CHECK3: cond.false:
718 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
719 // CHECK3-NEXT: br label [[COND_END]]
720 // CHECK3: cond.end:
721 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
722 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
723 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
724 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
725 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
726 // CHECK3: omp.inner.for.cond:
727 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
728 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
729 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
730 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
731 // CHECK3: omp.inner.for.body:
732 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
733 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
734 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
735 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4
736 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
737 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
738 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP9]]
739 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
740 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
741 // CHECK3: omp.body.continue:
742 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
743 // CHECK3: omp.inner.for.inc:
744 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
745 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
746 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
747 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
748 // CHECK3: omp.inner.for.end:
749 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
750 // CHECK3: omp.loop.exit:
751 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
752 // CHECK3-NEXT: ret void
753 //
754 //
755 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33
756 // CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
757 // CHECK3-NEXT: entry:
758 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
759 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
760 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
761 // CHECK3-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..1 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
762 // CHECK3-NEXT: ret void
763 //
764 //
765 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
766 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
767 // CHECK3-NEXT: entry:
768 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
769 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
770 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
771 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
772 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
773 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
774 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
775 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
776 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
777 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
778 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
779 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
780 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
781 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
782 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
783 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
784 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
785 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
786 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
787 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
788 // CHECK3-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)
789 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
790 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
791 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
792 // CHECK3: cond.true:
793 // CHECK3-NEXT: br label [[COND_END:%.*]]
794 // CHECK3: cond.false:
795 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
796 // CHECK3-NEXT: br label [[COND_END]]
797 // CHECK3: cond.end:
798 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
799 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
800 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
801 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
802 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
803 // CHECK3: omp.inner.for.cond:
804 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
805 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
806 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
807 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
808 // CHECK3: omp.inner.for.body:
809 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
810 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
811 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
812 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4
813 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
814 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
815 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP9]]
816 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
817 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
818 // CHECK3: omp.body.continue:
819 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
820 // CHECK3: omp.inner.for.inc:
821 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
822 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
823 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
824 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
825 // CHECK3: omp.inner.for.end:
826 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
827 // CHECK3: omp.loop.exit:
828 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
829 // CHECK3-NEXT: ret void
830 //
831 //
832 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38
833 // CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
834 // CHECK3-NEXT: entry:
835 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
836 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
837 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
838 // CHECK3-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..4 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
839 // CHECK3-NEXT: ret void
840 //
841 //
842 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
843 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
844 // CHECK3-NEXT: entry:
845 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
846 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
847 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
848 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
849 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
850 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
851 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
852 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
853 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
854 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
855 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
856 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
857 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
858 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
859 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
860 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
861 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
862 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
863 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
864 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
865 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
866 // CHECK3-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
867 // CHECK3: omp.dispatch.cond:
868 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
869 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
870 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
871 // CHECK3: cond.true:
872 // CHECK3-NEXT: br label [[COND_END:%.*]]
873 // CHECK3: cond.false:
874 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
875 // CHECK3-NEXT: br label [[COND_END]]
876 // CHECK3: cond.end:
877 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
878 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
879 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
880 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
881 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
882 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
883 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
884 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
885 // CHECK3: omp.dispatch.body:
886 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
887 // CHECK3: omp.inner.for.cond:
888 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
889 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
890 // CHECK3-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
891 // CHECK3-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
892 // CHECK3: omp.inner.for.body:
893 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
894 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
895 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
896 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
897 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
898 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !9
899 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
900 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !9
901 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
902 // CHECK3: omp.body.continue:
903 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
904 // CHECK3: omp.inner.for.inc:
905 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
906 // CHECK3-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
907 // CHECK3-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
908 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
909 // CHECK3: omp.inner.for.end:
910 // CHECK3-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
911 // CHECK3: omp.dispatch.inc:
912 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
913 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
914 // CHECK3-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
915 // CHECK3-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
916 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
917 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
918 // CHECK3-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
919 // CHECK3-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
920 // CHECK3-NEXT: br label [[OMP_DISPATCH_COND]]
921 // CHECK3: omp.dispatch.end:
922 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
923 // CHECK3-NEXT: ret void
924 //
925 //
926 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
927 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
928 // CHECK3-NEXT: entry:
929 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
930 // CHECK3-NEXT: ret void
931 //
932 //
933 // CHECK9-LABEL: define {{[^@]+}}@main
934 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
935 // CHECK9-NEXT: entry:
936 // CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
937 // CHECK9-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
938 // CHECK9-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
939 // CHECK9-NEXT: [[N:%.*]] = alloca i32, align 4
940 // CHECK9-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
941 // CHECK9-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
942 // CHECK9-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
943 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
944 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
945 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
946 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
947 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
948 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
949 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
950 // CHECK9-NEXT: [[N_CASTED3:%.*]] = alloca i64, align 8
951 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
952 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
953 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
954 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
955 // CHECK9-NEXT: [[_TMP9:%.*]] = alloca i32, align 4
956 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
957 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
958 // CHECK9-NEXT: [[N_CASTED19:%.*]] = alloca i64, align 8
959 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS21:%.*]] = alloca [3 x i8*], align 8
960 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS22:%.*]] = alloca [3 x i8*], align 8
961 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS23:%.*]] = alloca [3 x i8*], align 8
962 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES24:%.*]] = alloca [3 x i64], align 8
963 // CHECK9-NEXT: [[_TMP25:%.*]] = alloca i32, align 4
964 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_26:%.*]] = alloca i32, align 4
965 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
966 // CHECK9-NEXT: store i32 0, i32* [[RETVAL]], align 4
967 // CHECK9-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
968 // CHECK9-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
969 // CHECK9-NEXT: store i32 100, i32* [[N]], align 4
970 // CHECK9-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
971 // CHECK9-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
972 // CHECK9-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
973 // CHECK9-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
974 // CHECK9-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
975 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
976 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
977 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
978 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
979 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
980 // CHECK9-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
981 // CHECK9-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
982 // 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)
983 // CHECK9-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
984 // CHECK9-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
985 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP8]], align 8
986 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
987 // CHECK9-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
988 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP10]], align 8
989 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
990 // CHECK9-NEXT: store i8* null, i8** [[TMP11]], align 8
991 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
992 // CHECK9-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
993 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
994 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
995 // CHECK9-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
996 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
997 // CHECK9-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
998 // CHECK9-NEXT: store i8* null, i8** [[TMP16]], align 8
999 // CHECK9-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1000 // CHECK9-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
1001 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 8
1002 // CHECK9-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1003 // CHECK9-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
1004 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 8
1005 // CHECK9-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1006 // CHECK9-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 8
1007 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1008 // CHECK9-NEXT: store i8* null, i8** [[TMP22]], align 8
1009 // CHECK9-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1010 // CHECK9-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1011 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1012 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
1013 // CHECK9-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
1014 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1015 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
1016 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1017 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1018 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1019 // CHECK9-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1020 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
1021 // CHECK9-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
1022 // CHECK9-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1023 // CHECK9-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1024 // CHECK9-NEXT: store i32 1, i32* [[TMP30]], align 4
1025 // CHECK9-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1026 // CHECK9-NEXT: store i32 3, i32* [[TMP31]], align 4
1027 // CHECK9-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1028 // CHECK9-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 8
1029 // CHECK9-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1030 // CHECK9-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 8
1031 // CHECK9-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1032 // CHECK9-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 8
1033 // CHECK9-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1034 // CHECK9-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 8
1035 // CHECK9-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1036 // CHECK9-NEXT: store i8** null, i8*** [[TMP36]], align 8
1037 // CHECK9-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1038 // CHECK9-NEXT: store i8** null, i8*** [[TMP37]], align 8
1039 // CHECK9-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1040 // CHECK9-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
1041 // 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]+}}_main_l100.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1042 // CHECK9-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
1043 // CHECK9-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1044 // CHECK9: omp_offload.failed:
1045 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
1046 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
1047 // CHECK9: omp_offload.cont:
1048 // CHECK9-NEXT: [[TMP41:%.*]] = load i32, i32* [[N]], align 4
1049 // CHECK9-NEXT: [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
1050 // CHECK9-NEXT: store i32 [[TMP41]], i32* [[CONV4]], align 4
1051 // CHECK9-NEXT: [[TMP42:%.*]] = load i64, i64* [[N_CASTED3]], align 8
1052 // CHECK9-NEXT: [[TMP43:%.*]] = mul nuw i64 [[TMP1]], 4
1053 // CHECK9-NEXT: [[TMP44:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES8]] to i8*
1054 // CHECK9-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP44]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes.2 to i8*), i64 24, i1 false)
1055 // CHECK9-NEXT: [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
1056 // CHECK9-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
1057 // CHECK9-NEXT: store i64 [[TMP42]], i64* [[TMP46]], align 8
1058 // CHECK9-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
1059 // CHECK9-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
1060 // CHECK9-NEXT: store i64 [[TMP42]], i64* [[TMP48]], align 8
1061 // CHECK9-NEXT: [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
1062 // CHECK9-NEXT: store i8* null, i8** [[TMP49]], align 8
1063 // CHECK9-NEXT: [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
1064 // CHECK9-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
1065 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP51]], align 8
1066 // CHECK9-NEXT: [[TMP52:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
1067 // CHECK9-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i64*
1068 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP53]], align 8
1069 // CHECK9-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
1070 // CHECK9-NEXT: store i8* null, i8** [[TMP54]], align 8
1071 // CHECK9-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
1072 // CHECK9-NEXT: [[TMP56:%.*]] = bitcast i8** [[TMP55]] to i32**
1073 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP56]], align 8
1074 // CHECK9-NEXT: [[TMP57:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
1075 // CHECK9-NEXT: [[TMP58:%.*]] = bitcast i8** [[TMP57]] to i32**
1076 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP58]], align 8
1077 // CHECK9-NEXT: [[TMP59:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
1078 // CHECK9-NEXT: store i64 [[TMP43]], i64* [[TMP59]], align 8
1079 // CHECK9-NEXT: [[TMP60:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
1080 // CHECK9-NEXT: store i8* null, i8** [[TMP60]], align 8
1081 // CHECK9-NEXT: [[TMP61:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
1082 // CHECK9-NEXT: [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
1083 // CHECK9-NEXT: [[TMP63:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
1084 // CHECK9-NEXT: [[TMP64:%.*]] = load i32, i32* [[N]], align 4
1085 // CHECK9-NEXT: store i32 [[TMP64]], i32* [[DOTCAPTURE_EXPR_10]], align 4
1086 // CHECK9-NEXT: [[TMP65:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
1087 // CHECK9-NEXT: [[SUB12:%.*]] = sub nsw i32 [[TMP65]], 0
1088 // CHECK9-NEXT: [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
1089 // CHECK9-NEXT: [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
1090 // CHECK9-NEXT: store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
1091 // CHECK9-NEXT: [[TMP66:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
1092 // CHECK9-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP66]], 1
1093 // CHECK9-NEXT: [[TMP67:%.*]] = zext i32 [[ADD15]] to i64
1094 // CHECK9-NEXT: [[KERNEL_ARGS16:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1095 // CHECK9-NEXT: [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 0
1096 // CHECK9-NEXT: store i32 1, i32* [[TMP68]], align 4
1097 // CHECK9-NEXT: [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 1
1098 // CHECK9-NEXT: store i32 3, i32* [[TMP69]], align 4
1099 // CHECK9-NEXT: [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 2
1100 // CHECK9-NEXT: store i8** [[TMP61]], i8*** [[TMP70]], align 8
1101 // CHECK9-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 3
1102 // CHECK9-NEXT: store i8** [[TMP62]], i8*** [[TMP71]], align 8
1103 // CHECK9-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 4
1104 // CHECK9-NEXT: store i64* [[TMP63]], i64** [[TMP72]], align 8
1105 // CHECK9-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 5
1106 // CHECK9-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP73]], align 8
1107 // CHECK9-NEXT: [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 6
1108 // CHECK9-NEXT: store i8** null, i8*** [[TMP74]], align 8
1109 // CHECK9-NEXT: [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 7
1110 // CHECK9-NEXT: store i8** null, i8*** [[TMP75]], align 8
1111 // CHECK9-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 8
1112 // CHECK9-NEXT: store i64 [[TMP67]], i64* [[TMP76]], align 8
1113 // CHECK9-NEXT: [[TMP77:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]])
1114 // CHECK9-NEXT: [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
1115 // CHECK9-NEXT: br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
1116 // CHECK9: omp_offload.failed17:
1117 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105(i64 [[TMP42]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
1118 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT18]]
1119 // CHECK9: omp_offload.cont18:
1120 // CHECK9-NEXT: [[TMP79:%.*]] = load i32, i32* [[N]], align 4
1121 // CHECK9-NEXT: [[CONV20:%.*]] = bitcast i64* [[N_CASTED19]] to i32*
1122 // CHECK9-NEXT: store i32 [[TMP79]], i32* [[CONV20]], align 4
1123 // CHECK9-NEXT: [[TMP80:%.*]] = load i64, i64* [[N_CASTED19]], align 8
1124 // CHECK9-NEXT: [[TMP81:%.*]] = mul nuw i64 [[TMP1]], 4
1125 // CHECK9-NEXT: [[TMP82:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES24]] to i8*
1126 // CHECK9-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP82]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes.5 to i8*), i64 24, i1 false)
1127 // CHECK9-NEXT: [[TMP83:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
1128 // CHECK9-NEXT: [[TMP84:%.*]] = bitcast i8** [[TMP83]] to i64*
1129 // CHECK9-NEXT: store i64 [[TMP80]], i64* [[TMP84]], align 8
1130 // CHECK9-NEXT: [[TMP85:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
1131 // CHECK9-NEXT: [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
1132 // CHECK9-NEXT: store i64 [[TMP80]], i64* [[TMP86]], align 8
1133 // CHECK9-NEXT: [[TMP87:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 0
1134 // CHECK9-NEXT: store i8* null, i8** [[TMP87]], align 8
1135 // CHECK9-NEXT: [[TMP88:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 1
1136 // CHECK9-NEXT: [[TMP89:%.*]] = bitcast i8** [[TMP88]] to i64*
1137 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP89]], align 8
1138 // CHECK9-NEXT: [[TMP90:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 1
1139 // CHECK9-NEXT: [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i64*
1140 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP91]], align 8
1141 // CHECK9-NEXT: [[TMP92:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 1
1142 // CHECK9-NEXT: store i8* null, i8** [[TMP92]], align 8
1143 // CHECK9-NEXT: [[TMP93:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 2
1144 // CHECK9-NEXT: [[TMP94:%.*]] = bitcast i8** [[TMP93]] to i32**
1145 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP94]], align 8
1146 // CHECK9-NEXT: [[TMP95:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 2
1147 // CHECK9-NEXT: [[TMP96:%.*]] = bitcast i8** [[TMP95]] to i32**
1148 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP96]], align 8
1149 // CHECK9-NEXT: [[TMP97:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 2
1150 // CHECK9-NEXT: store i64 [[TMP81]], i64* [[TMP97]], align 8
1151 // CHECK9-NEXT: [[TMP98:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS23]], i64 0, i64 2
1152 // CHECK9-NEXT: store i8* null, i8** [[TMP98]], align 8
1153 // CHECK9-NEXT: [[TMP99:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS21]], i32 0, i32 0
1154 // CHECK9-NEXT: [[TMP100:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS22]], i32 0, i32 0
1155 // CHECK9-NEXT: [[TMP101:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES24]], i32 0, i32 0
1156 // CHECK9-NEXT: [[TMP102:%.*]] = load i32, i32* [[N]], align 4
1157 // CHECK9-NEXT: store i32 [[TMP102]], i32* [[DOTCAPTURE_EXPR_26]], align 4
1158 // CHECK9-NEXT: [[TMP103:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_26]], align 4
1159 // CHECK9-NEXT: [[SUB28:%.*]] = sub nsw i32 [[TMP103]], 0
1160 // CHECK9-NEXT: [[DIV29:%.*]] = sdiv i32 [[SUB28]], 1
1161 // CHECK9-NEXT: [[SUB30:%.*]] = sub nsw i32 [[DIV29]], 1
1162 // CHECK9-NEXT: store i32 [[SUB30]], i32* [[DOTCAPTURE_EXPR_27]], align 4
1163 // CHECK9-NEXT: [[TMP104:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
1164 // CHECK9-NEXT: [[ADD31:%.*]] = add nsw i32 [[TMP104]], 1
1165 // CHECK9-NEXT: [[TMP105:%.*]] = zext i32 [[ADD31]] to i64
1166 // CHECK9-NEXT: [[KERNEL_ARGS32:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1167 // CHECK9-NEXT: [[TMP106:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS32]], i32 0, i32 0
1168 // CHECK9-NEXT: store i32 1, i32* [[TMP106]], align 4
1169 // CHECK9-NEXT: [[TMP107:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS32]], i32 0, i32 1
1170 // CHECK9-NEXT: store i32 3, i32* [[TMP107]], align 4
1171 // CHECK9-NEXT: [[TMP108:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS32]], i32 0, i32 2
1172 // CHECK9-NEXT: store i8** [[TMP99]], i8*** [[TMP108]], align 8
1173 // CHECK9-NEXT: [[TMP109:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS32]], i32 0, i32 3
1174 // CHECK9-NEXT: store i8** [[TMP100]], i8*** [[TMP109]], align 8
1175 // CHECK9-NEXT: [[TMP110:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS32]], i32 0, i32 4
1176 // CHECK9-NEXT: store i64* [[TMP101]], i64** [[TMP110]], align 8
1177 // CHECK9-NEXT: [[TMP111:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS32]], i32 0, i32 5
1178 // CHECK9-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.6, i32 0, i32 0), i64** [[TMP111]], align 8
1179 // CHECK9-NEXT: [[TMP112:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS32]], i32 0, i32 6
1180 // CHECK9-NEXT: store i8** null, i8*** [[TMP112]], align 8
1181 // CHECK9-NEXT: [[TMP113:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS32]], i32 0, i32 7
1182 // CHECK9-NEXT: store i8** null, i8*** [[TMP113]], align 8
1183 // CHECK9-NEXT: [[TMP114:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS32]], i32 0, i32 8
1184 // CHECK9-NEXT: store i64 [[TMP105]], i64* [[TMP114]], align 8
1185 // CHECK9-NEXT: [[TMP115:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS32]])
1186 // CHECK9-NEXT: [[TMP116:%.*]] = icmp ne i32 [[TMP115]], 0
1187 // CHECK9-NEXT: br i1 [[TMP116]], label [[OMP_OFFLOAD_FAILED33:%.*]], label [[OMP_OFFLOAD_CONT34:%.*]]
1188 // CHECK9: omp_offload.failed33:
1189 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110(i64 [[TMP80]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
1190 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT34]]
1191 // CHECK9: omp_offload.cont34:
1192 // CHECK9-NEXT: [[TMP117:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1193 // CHECK9-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP117]])
1194 // CHECK9-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
1195 // CHECK9-NEXT: [[TMP118:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1196 // CHECK9-NEXT: call void @llvm.stackrestore(i8* [[TMP118]])
1197 // CHECK9-NEXT: [[TMP119:%.*]] = load i32, i32* [[RETVAL]], align 4
1198 // CHECK9-NEXT: ret i32 [[TMP119]]
1199 //
1200 //
1201 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100
1202 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1203 // CHECK9-NEXT: entry:
1204 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
1205 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1206 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1207 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
1208 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1209 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1210 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
1211 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1212 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1213 // 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]])
1214 // CHECK9-NEXT: ret void
1215 //
1216 //
1217 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
1218 // 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]] {
1219 // CHECK9-NEXT: entry:
1220 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1221 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1222 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
1223 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1224 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1225 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1226 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1227 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1228 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1229 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1230 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1231 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1232 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1233 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1234 // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4
1235 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1236 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1237 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
1238 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1239 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1240 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
1241 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1242 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1243 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1244 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1245 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1246 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1247 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1248 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1249 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1250 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
1251 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1252 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1253 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1254 // CHECK9: omp.precond.then:
1255 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1256 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1257 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
1258 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1259 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1260 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1261 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
1262 // 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)
1263 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1264 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1265 // CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1266 // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1267 // CHECK9: cond.true:
1268 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1269 // CHECK9-NEXT: br label [[COND_END:%.*]]
1270 // CHECK9: cond.false:
1271 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1272 // CHECK9-NEXT: br label [[COND_END]]
1273 // CHECK9: cond.end:
1274 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1275 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1276 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1277 // CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1278 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1279 // CHECK9: omp.inner.for.cond:
1280 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1281 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1282 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1283 // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1284 // CHECK9: omp.inner.for.body:
1285 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1286 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
1287 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1288 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
1289 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
1290 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
1291 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
1292 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1293 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1294 // CHECK9: omp.body.continue:
1295 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1296 // CHECK9: omp.inner.for.inc:
1297 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1298 // CHECK9-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
1299 // CHECK9-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1300 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
1301 // CHECK9: omp.inner.for.end:
1302 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1303 // CHECK9: omp.loop.exit:
1304 // CHECK9-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1305 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
1306 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
1307 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
1308 // CHECK9: omp.precond.end:
1309 // CHECK9-NEXT: ret void
1310 //
1311 //
1312 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105
1313 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1314 // CHECK9-NEXT: entry:
1315 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
1316 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1317 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1318 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
1319 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1320 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1321 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
1322 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1323 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1324 // 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..1 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
1325 // CHECK9-NEXT: ret void
1326 //
1327 //
1328 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
1329 // 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]] {
1330 // CHECK9-NEXT: entry:
1331 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1332 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1333 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
1334 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1335 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1336 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1337 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1338 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1339 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1340 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1341 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1342 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1343 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1344 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1345 // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4
1346 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1347 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1348 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
1349 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1350 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1351 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
1352 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1353 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1354 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1355 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1356 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1357 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1358 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1359 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1360 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1361 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
1362 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1363 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1364 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1365 // CHECK9: omp.precond.then:
1366 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1367 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1368 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
1369 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1370 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1371 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1372 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
1373 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1374 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1375 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1376 // CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1377 // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1378 // CHECK9: cond.true:
1379 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1380 // CHECK9-NEXT: br label [[COND_END:%.*]]
1381 // CHECK9: cond.false:
1382 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1383 // CHECK9-NEXT: br label [[COND_END]]
1384 // CHECK9: cond.end:
1385 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1386 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1387 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1388 // CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1389 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1390 // CHECK9: omp.inner.for.cond:
1391 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1392 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1393 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1394 // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1395 // CHECK9: omp.inner.for.body:
1396 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1397 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
1398 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1399 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
1400 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
1401 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP17]] to i64
1402 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
1403 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1404 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1405 // CHECK9: omp.body.continue:
1406 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1407 // CHECK9: omp.inner.for.inc:
1408 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1409 // CHECK9-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
1410 // CHECK9-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
1411 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
1412 // CHECK9: omp.inner.for.end:
1413 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1414 // CHECK9: omp.loop.exit:
1415 // CHECK9-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1416 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
1417 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
1418 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
1419 // CHECK9: omp.precond.end:
1420 // CHECK9-NEXT: ret void
1421 //
1422 //
1423 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110
1424 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1425 // CHECK9-NEXT: entry:
1426 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
1427 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1428 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1429 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1430 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
1431 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
1432 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1433 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1434 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
1435 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1436 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1437 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
1438 // CHECK9-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
1439 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1440 // CHECK9-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
1441 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV1]], align 4
1442 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
1443 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
1444 // CHECK9-NEXT: ret void
1445 //
1446 //
1447 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
1448 // 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:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
1449 // CHECK9-NEXT: entry:
1450 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1451 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1452 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
1453 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1454 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1455 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
1456 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1457 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1458 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1459 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
1460 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1461 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1462 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1463 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1464 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1465 // CHECK9-NEXT: [[I4:%.*]] = alloca i32, align 4
1466 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1467 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1468 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
1469 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1470 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1471 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
1472 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
1473 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1474 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1475 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
1476 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1477 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1478 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1479 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1480 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1481 // CHECK9-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
1482 // CHECK9-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
1483 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
1484 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1485 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1486 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1487 // CHECK9: omp.precond.then:
1488 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1489 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
1490 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
1491 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1492 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1493 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[CONV]], align 4
1494 // CHECK9-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1495 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
1496 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
1497 // CHECK9-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
1498 // CHECK9: omp.dispatch.cond:
1499 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1500 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
1501 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
1502 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1503 // CHECK9: cond.true:
1504 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
1505 // CHECK9-NEXT: br label [[COND_END:%.*]]
1506 // CHECK9: cond.false:
1507 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1508 // CHECK9-NEXT: br label [[COND_END]]
1509 // CHECK9: cond.end:
1510 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
1511 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1512 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1513 // CHECK9-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
1514 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1515 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1516 // CHECK9-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
1517 // CHECK9-NEXT: br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1518 // CHECK9: omp.dispatch.body:
1519 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1520 // CHECK9: omp.inner.for.cond:
1521 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1522 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !11
1523 // CHECK9-NEXT: [[CMP7:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
1524 // CHECK9-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1525 // CHECK9: omp.inner.for.body:
1526 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1527 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
1528 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1529 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !11
1530 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !11
1531 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP20]] to i64
1532 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
1533 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !11
1534 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1535 // CHECK9: omp.body.continue:
1536 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1537 // CHECK9: omp.inner.for.inc:
1538 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1539 // CHECK9-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP21]], 1
1540 // CHECK9-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1541 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
1542 // CHECK9: omp.inner.for.end:
1543 // CHECK9-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
1544 // CHECK9: omp.dispatch.inc:
1545 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1546 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1547 // CHECK9-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
1548 // CHECK9-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
1549 // CHECK9-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1550 // CHECK9-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1551 // CHECK9-NEXT: [[ADD10:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
1552 // CHECK9-NEXT: store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
1553 // CHECK9-NEXT: br label [[OMP_DISPATCH_COND]]
1554 // CHECK9: omp.dispatch.end:
1555 // CHECK9-NEXT: [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1556 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
1557 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP27]])
1558 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
1559 // CHECK9: omp.precond.end:
1560 // CHECK9-NEXT: ret void
1561 //
1562 //
1563 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
1564 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
1565 // CHECK9-NEXT: entry:
1566 // CHECK9-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
1567 // CHECK9-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
1568 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
1569 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
1570 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
1571 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1572 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
1573 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
1574 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
1575 // CHECK9-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
1576 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [1 x i8*], align 8
1577 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS9:%.*]] = alloca [1 x i8*], align 8
1578 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [1 x i8*], align 8
1579 // CHECK9-NEXT: [[_TMP11:%.*]] = alloca i32, align 4
1580 // CHECK9-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1581 // CHECK9-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1582 // CHECK9-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
1583 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
1584 // CHECK9-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1585 // CHECK9-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
1586 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
1587 // CHECK9-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1588 // CHECK9-NEXT: store i8* null, i8** [[TMP4]], align 8
1589 // CHECK9-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1590 // CHECK9-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1591 // CHECK9-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1592 // CHECK9-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1593 // CHECK9-NEXT: store i32 1, i32* [[TMP7]], align 4
1594 // CHECK9-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1595 // CHECK9-NEXT: store i32 1, i32* [[TMP8]], align 4
1596 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1597 // CHECK9-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8
1598 // CHECK9-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1599 // CHECK9-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8
1600 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1601 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP11]], align 8
1602 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1603 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP12]], align 8
1604 // CHECK9-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1605 // CHECK9-NEXT: store i8** null, i8*** [[TMP13]], align 8
1606 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1607 // CHECK9-NEXT: store i8** null, i8*** [[TMP14]], align 8
1608 // CHECK9-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1609 // CHECK9-NEXT: store i64 10, i64* [[TMP15]], align 8
1610 // CHECK9-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1611 // CHECK9-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
1612 // CHECK9-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1613 // CHECK9: omp_offload.failed:
1614 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79([10 x i32]* [[A]]) #[[ATTR3]]
1615 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
1616 // CHECK9: omp_offload.cont:
1617 // CHECK9-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
1618 // CHECK9-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to [10 x i32]**
1619 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP19]], align 8
1620 // CHECK9-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
1621 // CHECK9-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [10 x i32]**
1622 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP21]], align 8
1623 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
1624 // CHECK9-NEXT: store i8* null, i8** [[TMP22]], align 8
1625 // CHECK9-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
1626 // CHECK9-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
1627 // CHECK9-NEXT: [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1628 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 0
1629 // CHECK9-NEXT: store i32 1, i32* [[TMP25]], align 4
1630 // CHECK9-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 1
1631 // CHECK9-NEXT: store i32 1, i32* [[TMP26]], align 4
1632 // CHECK9-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 2
1633 // CHECK9-NEXT: store i8** [[TMP23]], i8*** [[TMP27]], align 8
1634 // CHECK9-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 3
1635 // CHECK9-NEXT: store i8** [[TMP24]], i8*** [[TMP28]], align 8
1636 // CHECK9-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 4
1637 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.11, i32 0, i32 0), i64** [[TMP29]], align 8
1638 // CHECK9-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 5
1639 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.12, i32 0, i32 0), i64** [[TMP30]], align 8
1640 // CHECK9-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 6
1641 // CHECK9-NEXT: store i8** null, i8*** [[TMP31]], align 8
1642 // CHECK9-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 7
1643 // CHECK9-NEXT: store i8** null, i8*** [[TMP32]], align 8
1644 // CHECK9-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 8
1645 // CHECK9-NEXT: store i64 10, i64* [[TMP33]], align 8
1646 // CHECK9-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]])
1647 // CHECK9-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
1648 // CHECK9-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
1649 // CHECK9: omp_offload.failed6:
1650 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84([10 x i32]* [[A]]) #[[ATTR3]]
1651 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT7]]
1652 // CHECK9: omp_offload.cont7:
1653 // CHECK9-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
1654 // CHECK9-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
1655 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 8
1656 // CHECK9-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
1657 // CHECK9-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [10 x i32]**
1658 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP39]], align 8
1659 // CHECK9-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
1660 // CHECK9-NEXT: store i8* null, i8** [[TMP40]], align 8
1661 // CHECK9-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
1662 // CHECK9-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
1663 // CHECK9-NEXT: [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1664 // CHECK9-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 0
1665 // CHECK9-NEXT: store i32 1, i32* [[TMP43]], align 4
1666 // CHECK9-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 1
1667 // CHECK9-NEXT: store i32 1, i32* [[TMP44]], align 4
1668 // CHECK9-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 2
1669 // CHECK9-NEXT: store i8** [[TMP41]], i8*** [[TMP45]], align 8
1670 // CHECK9-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 3
1671 // CHECK9-NEXT: store i8** [[TMP42]], i8*** [[TMP46]], align 8
1672 // CHECK9-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 4
1673 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.14, i32 0, i32 0), i64** [[TMP47]], align 8
1674 // CHECK9-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 5
1675 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.15, i32 0, i32 0), i64** [[TMP48]], align 8
1676 // CHECK9-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 6
1677 // CHECK9-NEXT: store i8** null, i8*** [[TMP49]], align 8
1678 // CHECK9-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 7
1679 // CHECK9-NEXT: store i8** null, i8*** [[TMP50]], align 8
1680 // CHECK9-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 8
1681 // CHECK9-NEXT: store i64 10, i64* [[TMP51]], align 8
1682 // CHECK9-NEXT: [[TMP52:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]])
1683 // CHECK9-NEXT: [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
1684 // CHECK9-NEXT: br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
1685 // CHECK9: omp_offload.failed13:
1686 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89([10 x i32]* [[A]]) #[[ATTR3]]
1687 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT14]]
1688 // CHECK9: omp_offload.cont14:
1689 // CHECK9-NEXT: ret i32 0
1690 //
1691 //
1692 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79
1693 // CHECK9-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1694 // CHECK9-NEXT: entry:
1695 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
1696 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
1697 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
1698 // CHECK9-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..7 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
1699 // CHECK9-NEXT: ret void
1700 //
1701 //
1702 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..7
1703 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1704 // CHECK9-NEXT: entry:
1705 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1706 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1707 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
1708 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1709 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1710 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1711 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1712 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1713 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1714 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1715 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1716 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1717 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
1718 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
1719 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1720 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
1721 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1722 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1723 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1724 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1725 // CHECK9-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)
1726 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1727 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
1728 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1729 // CHECK9: cond.true:
1730 // CHECK9-NEXT: br label [[COND_END:%.*]]
1731 // CHECK9: cond.false:
1732 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1733 // CHECK9-NEXT: br label [[COND_END]]
1734 // CHECK9: cond.end:
1735 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1736 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1737 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1738 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1739 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1740 // CHECK9: omp.inner.for.cond:
1741 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1742 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1743 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1744 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1745 // CHECK9: omp.inner.for.body:
1746 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1747 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1748 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1749 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1750 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
1751 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
1752 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
1753 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1754 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1755 // CHECK9: omp.body.continue:
1756 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1757 // CHECK9: omp.inner.for.inc:
1758 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1759 // CHECK9-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
1760 // CHECK9-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
1761 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
1762 // CHECK9: omp.inner.for.end:
1763 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1764 // CHECK9: omp.loop.exit:
1765 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1766 // CHECK9-NEXT: ret void
1767 //
1768 //
1769 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84
1770 // CHECK9-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1771 // CHECK9-NEXT: entry:
1772 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
1773 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
1774 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
1775 // CHECK9-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..10 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
1776 // CHECK9-NEXT: ret void
1777 //
1778 //
1779 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..10
1780 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1781 // CHECK9-NEXT: entry:
1782 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1783 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1784 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
1785 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1786 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1787 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1788 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1789 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1790 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1791 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1792 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1793 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1794 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
1795 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
1796 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1797 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
1798 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1799 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1800 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1801 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1802 // CHECK9-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)
1803 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1804 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
1805 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1806 // CHECK9: cond.true:
1807 // CHECK9-NEXT: br label [[COND_END:%.*]]
1808 // CHECK9: cond.false:
1809 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1810 // CHECK9-NEXT: br label [[COND_END]]
1811 // CHECK9: cond.end:
1812 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1813 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1814 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1815 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1816 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1817 // CHECK9: omp.inner.for.cond:
1818 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1819 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1820 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1821 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1822 // CHECK9: omp.inner.for.body:
1823 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1824 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
1825 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1826 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1827 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
1828 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP9]] to i64
1829 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
1830 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1831 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1832 // CHECK9: omp.body.continue:
1833 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1834 // CHECK9: omp.inner.for.inc:
1835 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1836 // CHECK9-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
1837 // CHECK9-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
1838 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
1839 // CHECK9: omp.inner.for.end:
1840 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1841 // CHECK9: omp.loop.exit:
1842 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1843 // CHECK9-NEXT: ret void
1844 //
1845 //
1846 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89
1847 // CHECK9-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1848 // CHECK9-NEXT: entry:
1849 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
1850 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
1851 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
1852 // CHECK9-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..13 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
1853 // CHECK9-NEXT: ret void
1854 //
1855 //
1856 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..13
1857 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
1858 // CHECK9-NEXT: entry:
1859 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1860 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1861 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
1862 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1863 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1864 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1865 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1866 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1867 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1868 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1869 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1870 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1871 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
1872 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
1873 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1874 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
1875 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1876 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1877 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1878 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1879 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 10)
1880 // CHECK9-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
1881 // CHECK9: omp.dispatch.cond:
1882 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1883 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
1884 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1885 // CHECK9: cond.true:
1886 // CHECK9-NEXT: br label [[COND_END:%.*]]
1887 // CHECK9: cond.false:
1888 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1889 // CHECK9-NEXT: br label [[COND_END]]
1890 // CHECK9: cond.end:
1891 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1892 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1893 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1894 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1895 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1896 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1897 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1898 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1899 // CHECK9: omp.dispatch.body:
1900 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1901 // CHECK9: omp.inner.for.cond:
1902 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
1903 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
1904 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1905 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1906 // CHECK9: omp.inner.for.body:
1907 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
1908 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1909 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1910 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !14
1911 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !14
1912 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
1913 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
1914 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !14
1915 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1916 // CHECK9: omp.body.continue:
1917 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1918 // CHECK9: omp.inner.for.inc:
1919 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
1920 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
1921 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
1922 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
1923 // CHECK9: omp.inner.for.end:
1924 // CHECK9-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
1925 // CHECK9: omp.dispatch.inc:
1926 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1927 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1928 // CHECK9-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
1929 // CHECK9-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
1930 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1931 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1932 // CHECK9-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
1933 // CHECK9-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
1934 // CHECK9-NEXT: br label [[OMP_DISPATCH_COND]]
1935 // CHECK9: omp.dispatch.end:
1936 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1937 // CHECK9-NEXT: ret void
1938 //
1939 //
1940 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1941 // CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
1942 // CHECK9-NEXT: entry:
1943 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
1944 // CHECK9-NEXT: ret void
1945 //
1946 //
1947 // CHECK11-LABEL: define {{[^@]+}}@main
1948 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1949 // CHECK11-NEXT: entry:
1950 // CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
1951 // CHECK11-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
1952 // CHECK11-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
1953 // CHECK11-NEXT: [[N:%.*]] = alloca i32, align 4
1954 // CHECK11-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
1955 // CHECK11-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
1956 // CHECK11-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
1957 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
1958 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
1959 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
1960 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
1961 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
1962 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1963 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1964 // CHECK11-NEXT: [[N_CASTED3:%.*]] = alloca i32, align 4
1965 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
1966 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
1967 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
1968 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
1969 // CHECK11-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
1970 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
1971 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
1972 // CHECK11-NEXT: [[N_CASTED18:%.*]] = alloca i32, align 4
1973 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [3 x i8*], align 4
1974 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS20:%.*]] = alloca [3 x i8*], align 4
1975 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [3 x i8*], align 4
1976 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES22:%.*]] = alloca [3 x i64], align 4
1977 // CHECK11-NEXT: [[_TMP23:%.*]] = alloca i32, align 4
1978 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
1979 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_25:%.*]] = alloca i32, align 4
1980 // CHECK11-NEXT: store i32 0, i32* [[RETVAL]], align 4
1981 // CHECK11-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1982 // CHECK11-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
1983 // CHECK11-NEXT: store i32 100, i32* [[N]], align 4
1984 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1985 // CHECK11-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
1986 // CHECK11-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
1987 // CHECK11-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
1988 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
1989 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
1990 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
1991 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
1992 // CHECK11-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
1993 // CHECK11-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
1994 // CHECK11-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1995 // 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)
1996 // CHECK11-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1997 // CHECK11-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
1998 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP8]], align 4
1999 // CHECK11-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2000 // CHECK11-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
2001 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
2002 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2003 // CHECK11-NEXT: store i8* null, i8** [[TMP11]], align 4
2004 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2005 // CHECK11-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
2006 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
2007 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2008 // CHECK11-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
2009 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
2010 // CHECK11-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
2011 // CHECK11-NEXT: store i8* null, i8** [[TMP16]], align 4
2012 // CHECK11-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2013 // CHECK11-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
2014 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 4
2015 // CHECK11-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2016 // CHECK11-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
2017 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 4
2018 // CHECK11-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
2019 // CHECK11-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 4
2020 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
2021 // CHECK11-NEXT: store i8* null, i8** [[TMP22]], align 4
2022 // CHECK11-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2023 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2024 // CHECK11-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
2025 // CHECK11-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
2026 // CHECK11-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
2027 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2028 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
2029 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2030 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2031 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2032 // CHECK11-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2033 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
2034 // CHECK11-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
2035 // CHECK11-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2036 // CHECK11-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
2037 // CHECK11-NEXT: store i32 1, i32* [[TMP30]], align 4
2038 // CHECK11-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
2039 // CHECK11-NEXT: store i32 3, i32* [[TMP31]], align 4
2040 // CHECK11-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
2041 // CHECK11-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 4
2042 // CHECK11-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
2043 // CHECK11-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 4
2044 // CHECK11-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
2045 // CHECK11-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 4
2046 // CHECK11-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
2047 // CHECK11-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 4
2048 // CHECK11-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
2049 // CHECK11-NEXT: store i8** null, i8*** [[TMP36]], align 4
2050 // CHECK11-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
2051 // CHECK11-NEXT: store i8** null, i8*** [[TMP37]], align 4
2052 // CHECK11-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
2053 // CHECK11-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
2054 // 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]+}}_main_l100.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
2055 // CHECK11-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
2056 // CHECK11-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2057 // CHECK11: omp_offload.failed:
2058 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
2059 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
2060 // CHECK11: omp_offload.cont:
2061 // CHECK11-NEXT: [[TMP41:%.*]] = load i32, i32* [[N]], align 4
2062 // CHECK11-NEXT: store i32 [[TMP41]], i32* [[N_CASTED3]], align 4
2063 // CHECK11-NEXT: [[TMP42:%.*]] = load i32, i32* [[N_CASTED3]], align 4
2064 // CHECK11-NEXT: [[TMP43:%.*]] = mul nuw i32 [[TMP0]], 4
2065 // CHECK11-NEXT: [[TMP44:%.*]] = sext i32 [[TMP43]] to i64
2066 // CHECK11-NEXT: [[TMP45:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES7]] to i8*
2067 // CHECK11-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP45]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes.2 to i8*), i32 24, i1 false)
2068 // CHECK11-NEXT: [[TMP46:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
2069 // CHECK11-NEXT: [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i32*
2070 // CHECK11-NEXT: store i32 [[TMP42]], i32* [[TMP47]], align 4
2071 // CHECK11-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
2072 // CHECK11-NEXT: [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32*
2073 // CHECK11-NEXT: store i32 [[TMP42]], i32* [[TMP49]], align 4
2074 // CHECK11-NEXT: [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
2075 // CHECK11-NEXT: store i8* null, i8** [[TMP50]], align 4
2076 // CHECK11-NEXT: [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
2077 // CHECK11-NEXT: [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
2078 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP52]], align 4
2079 // CHECK11-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
2080 // CHECK11-NEXT: [[TMP54:%.*]] = bitcast i8** [[TMP53]] to i32*
2081 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP54]], align 4
2082 // CHECK11-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
2083 // CHECK11-NEXT: store i8* null, i8** [[TMP55]], align 4
2084 // CHECK11-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
2085 // CHECK11-NEXT: [[TMP57:%.*]] = bitcast i8** [[TMP56]] to i32**
2086 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP57]], align 4
2087 // CHECK11-NEXT: [[TMP58:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
2088 // CHECK11-NEXT: [[TMP59:%.*]] = bitcast i8** [[TMP58]] to i32**
2089 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP59]], align 4
2090 // CHECK11-NEXT: [[TMP60:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
2091 // CHECK11-NEXT: store i64 [[TMP44]], i64* [[TMP60]], align 4
2092 // CHECK11-NEXT: [[TMP61:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
2093 // CHECK11-NEXT: store i8* null, i8** [[TMP61]], align 4
2094 // CHECK11-NEXT: [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
2095 // CHECK11-NEXT: [[TMP63:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
2096 // CHECK11-NEXT: [[TMP64:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
2097 // CHECK11-NEXT: [[TMP65:%.*]] = load i32, i32* [[N]], align 4
2098 // CHECK11-NEXT: store i32 [[TMP65]], i32* [[DOTCAPTURE_EXPR_9]], align 4
2099 // CHECK11-NEXT: [[TMP66:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
2100 // CHECK11-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP66]], 0
2101 // CHECK11-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
2102 // CHECK11-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
2103 // CHECK11-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
2104 // CHECK11-NEXT: [[TMP67:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
2105 // CHECK11-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP67]], 1
2106 // CHECK11-NEXT: [[TMP68:%.*]] = zext i32 [[ADD14]] to i64
2107 // CHECK11-NEXT: [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2108 // CHECK11-NEXT: [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
2109 // CHECK11-NEXT: store i32 1, i32* [[TMP69]], align 4
2110 // CHECK11-NEXT: [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
2111 // CHECK11-NEXT: store i32 3, i32* [[TMP70]], align 4
2112 // CHECK11-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
2113 // CHECK11-NEXT: store i8** [[TMP62]], i8*** [[TMP71]], align 4
2114 // CHECK11-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
2115 // CHECK11-NEXT: store i8** [[TMP63]], i8*** [[TMP72]], align 4
2116 // CHECK11-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
2117 // CHECK11-NEXT: store i64* [[TMP64]], i64** [[TMP73]], align 4
2118 // CHECK11-NEXT: [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
2119 // CHECK11-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.3, i32 0, i32 0), i64** [[TMP74]], align 4
2120 // CHECK11-NEXT: [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
2121 // CHECK11-NEXT: store i8** null, i8*** [[TMP75]], align 4
2122 // CHECK11-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
2123 // CHECK11-NEXT: store i8** null, i8*** [[TMP76]], align 4
2124 // CHECK11-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
2125 // CHECK11-NEXT: store i64 [[TMP68]], i64* [[TMP77]], align 8
2126 // CHECK11-NEXT: [[TMP78:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
2127 // CHECK11-NEXT: [[TMP79:%.*]] = icmp ne i32 [[TMP78]], 0
2128 // CHECK11-NEXT: br i1 [[TMP79]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
2129 // CHECK11: omp_offload.failed16:
2130 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105(i32 [[TMP42]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
2131 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT17]]
2132 // CHECK11: omp_offload.cont17:
2133 // CHECK11-NEXT: [[TMP80:%.*]] = load i32, i32* [[N]], align 4
2134 // CHECK11-NEXT: store i32 [[TMP80]], i32* [[N_CASTED18]], align 4
2135 // CHECK11-NEXT: [[TMP81:%.*]] = load i32, i32* [[N_CASTED18]], align 4
2136 // CHECK11-NEXT: [[TMP82:%.*]] = mul nuw i32 [[TMP0]], 4
2137 // CHECK11-NEXT: [[TMP83:%.*]] = sext i32 [[TMP82]] to i64
2138 // CHECK11-NEXT: [[TMP84:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES22]] to i8*
2139 // CHECK11-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP84]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes.5 to i8*), i32 24, i1 false)
2140 // CHECK11-NEXT: [[TMP85:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
2141 // CHECK11-NEXT: [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i32*
2142 // CHECK11-NEXT: store i32 [[TMP81]], i32* [[TMP86]], align 4
2143 // CHECK11-NEXT: [[TMP87:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
2144 // CHECK11-NEXT: [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
2145 // CHECK11-NEXT: store i32 [[TMP81]], i32* [[TMP88]], align 4
2146 // CHECK11-NEXT: [[TMP89:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 0
2147 // CHECK11-NEXT: store i8* null, i8** [[TMP89]], align 4
2148 // CHECK11-NEXT: [[TMP90:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 1
2149 // CHECK11-NEXT: [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i32*
2150 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP91]], align 4
2151 // CHECK11-NEXT: [[TMP92:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 1
2152 // CHECK11-NEXT: [[TMP93:%.*]] = bitcast i8** [[TMP92]] to i32*
2153 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP93]], align 4
2154 // CHECK11-NEXT: [[TMP94:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 1
2155 // CHECK11-NEXT: store i8* null, i8** [[TMP94]], align 4
2156 // CHECK11-NEXT: [[TMP95:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 2
2157 // CHECK11-NEXT: [[TMP96:%.*]] = bitcast i8** [[TMP95]] to i32**
2158 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP96]], align 4
2159 // CHECK11-NEXT: [[TMP97:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 2
2160 // CHECK11-NEXT: [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i32**
2161 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP98]], align 4
2162 // CHECK11-NEXT: [[TMP99:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES22]], i32 0, i32 2
2163 // CHECK11-NEXT: store i64 [[TMP83]], i64* [[TMP99]], align 4
2164 // CHECK11-NEXT: [[TMP100:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 2
2165 // CHECK11-NEXT: store i8* null, i8** [[TMP100]], align 4
2166 // CHECK11-NEXT: [[TMP101:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
2167 // CHECK11-NEXT: [[TMP102:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
2168 // CHECK11-NEXT: [[TMP103:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES22]], i32 0, i32 0
2169 // CHECK11-NEXT: [[TMP104:%.*]] = load i32, i32* [[N]], align 4
2170 // CHECK11-NEXT: store i32 [[TMP104]], i32* [[DOTCAPTURE_EXPR_24]], align 4
2171 // CHECK11-NEXT: [[TMP105:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
2172 // CHECK11-NEXT: [[SUB26:%.*]] = sub nsw i32 [[TMP105]], 0
2173 // CHECK11-NEXT: [[DIV27:%.*]] = sdiv i32 [[SUB26]], 1
2174 // CHECK11-NEXT: [[SUB28:%.*]] = sub nsw i32 [[DIV27]], 1
2175 // CHECK11-NEXT: store i32 [[SUB28]], i32* [[DOTCAPTURE_EXPR_25]], align 4
2176 // CHECK11-NEXT: [[TMP106:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_25]], align 4
2177 // CHECK11-NEXT: [[ADD29:%.*]] = add nsw i32 [[TMP106]], 1
2178 // CHECK11-NEXT: [[TMP107:%.*]] = zext i32 [[ADD29]] to i64
2179 // CHECK11-NEXT: [[KERNEL_ARGS30:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2180 // CHECK11-NEXT: [[TMP108:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 0
2181 // CHECK11-NEXT: store i32 1, i32* [[TMP108]], align 4
2182 // CHECK11-NEXT: [[TMP109:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 1
2183 // CHECK11-NEXT: store i32 3, i32* [[TMP109]], align 4
2184 // CHECK11-NEXT: [[TMP110:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 2
2185 // CHECK11-NEXT: store i8** [[TMP101]], i8*** [[TMP110]], align 4
2186 // CHECK11-NEXT: [[TMP111:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 3
2187 // CHECK11-NEXT: store i8** [[TMP102]], i8*** [[TMP111]], align 4
2188 // CHECK11-NEXT: [[TMP112:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 4
2189 // CHECK11-NEXT: store i64* [[TMP103]], i64** [[TMP112]], align 4
2190 // CHECK11-NEXT: [[TMP113:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 5
2191 // CHECK11-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.6, i32 0, i32 0), i64** [[TMP113]], align 4
2192 // CHECK11-NEXT: [[TMP114:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 6
2193 // CHECK11-NEXT: store i8** null, i8*** [[TMP114]], align 4
2194 // CHECK11-NEXT: [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 7
2195 // CHECK11-NEXT: store i8** null, i8*** [[TMP115]], align 4
2196 // CHECK11-NEXT: [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 8
2197 // CHECK11-NEXT: store i64 [[TMP107]], i64* [[TMP116]], align 8
2198 // CHECK11-NEXT: [[TMP117:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]])
2199 // CHECK11-NEXT: [[TMP118:%.*]] = icmp ne i32 [[TMP117]], 0
2200 // CHECK11-NEXT: br i1 [[TMP118]], label [[OMP_OFFLOAD_FAILED31:%.*]], label [[OMP_OFFLOAD_CONT32:%.*]]
2201 // CHECK11: omp_offload.failed31:
2202 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110(i32 [[TMP81]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
2203 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT32]]
2204 // CHECK11: omp_offload.cont32:
2205 // CHECK11-NEXT: [[TMP119:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
2206 // CHECK11-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP119]])
2207 // CHECK11-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
2208 // CHECK11-NEXT: [[TMP120:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
2209 // CHECK11-NEXT: call void @llvm.stackrestore(i8* [[TMP120]])
2210 // CHECK11-NEXT: [[TMP121:%.*]] = load i32, i32* [[RETVAL]], align 4
2211 // CHECK11-NEXT: ret i32 [[TMP121]]
2212 //
2213 //
2214 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l100
2215 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
2216 // CHECK11-NEXT: entry:
2217 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
2218 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
2219 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
2220 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
2221 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2222 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
2223 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2224 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
2225 // 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]])
2226 // CHECK11-NEXT: ret void
2227 //
2228 //
2229 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
2230 // 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]] {
2231 // CHECK11-NEXT: entry:
2232 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2233 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2234 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
2235 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
2236 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
2237 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2238 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
2239 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2240 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2241 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
2242 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2243 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2244 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2245 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2246 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
2247 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2248 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2249 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
2250 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2251 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
2252 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
2253 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2254 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
2255 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2256 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
2257 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2258 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2259 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2260 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2261 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2262 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
2263 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2264 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2265 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2266 // CHECK11: omp.precond.then:
2267 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2268 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2269 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
2270 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2271 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2272 // CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2273 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
2274 // 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)
2275 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2276 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2277 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
2278 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2279 // CHECK11: cond.true:
2280 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2281 // CHECK11-NEXT: br label [[COND_END:%.*]]
2282 // CHECK11: cond.false:
2283 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2284 // CHECK11-NEXT: br label [[COND_END]]
2285 // CHECK11: cond.end:
2286 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2287 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2288 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2289 // CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
2290 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2291 // CHECK11: omp.inner.for.cond:
2292 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2293 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2294 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2295 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2296 // CHECK11: omp.inner.for.body:
2297 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2298 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
2299 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2300 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
2301 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
2302 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP17]]
2303 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2304 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2305 // CHECK11: omp.body.continue:
2306 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2307 // CHECK11: omp.inner.for.inc:
2308 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2309 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
2310 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
2311 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
2312 // CHECK11: omp.inner.for.end:
2313 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2314 // CHECK11: omp.loop.exit:
2315 // CHECK11-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2316 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
2317 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
2318 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
2319 // CHECK11: omp.precond.end:
2320 // CHECK11-NEXT: ret void
2321 //
2322 //
2323 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l105
2324 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2325 // CHECK11-NEXT: entry:
2326 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
2327 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
2328 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
2329 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
2330 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2331 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
2332 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2333 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
2334 // 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..1 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
2335 // CHECK11-NEXT: ret void
2336 //
2337 //
2338 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
2339 // 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]] {
2340 // CHECK11-NEXT: entry:
2341 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2342 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2343 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
2344 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
2345 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
2346 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2347 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
2348 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2349 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2350 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
2351 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2352 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2353 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2354 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2355 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
2356 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2357 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2358 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
2359 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2360 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
2361 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
2362 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2363 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
2364 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2365 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
2366 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2367 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2368 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2369 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2370 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2371 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
2372 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2373 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2374 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2375 // CHECK11: omp.precond.then:
2376 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2377 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2378 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
2379 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2380 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2381 // CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2382 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
2383 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2384 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2385 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2386 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
2387 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2388 // CHECK11: cond.true:
2389 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2390 // CHECK11-NEXT: br label [[COND_END:%.*]]
2391 // CHECK11: cond.false:
2392 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2393 // CHECK11-NEXT: br label [[COND_END]]
2394 // CHECK11: cond.end:
2395 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2396 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2397 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2398 // CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
2399 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2400 // CHECK11: omp.inner.for.cond:
2401 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2402 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2403 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2404 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2405 // CHECK11: omp.inner.for.body:
2406 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2407 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP16]], 1
2408 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2409 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
2410 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[I3]], align 4
2411 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP17]]
2412 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2413 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2414 // CHECK11: omp.body.continue:
2415 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2416 // CHECK11: omp.inner.for.inc:
2417 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2418 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP18]], 1
2419 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
2420 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
2421 // CHECK11: omp.inner.for.end:
2422 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2423 // CHECK11: omp.loop.exit:
2424 // CHECK11-NEXT: [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2425 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
2426 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP20]])
2427 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
2428 // CHECK11: omp.precond.end:
2429 // CHECK11-NEXT: ret void
2430 //
2431 //
2432 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l110
2433 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2434 // CHECK11-NEXT: entry:
2435 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
2436 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
2437 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
2438 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2439 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
2440 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
2441 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2442 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
2443 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2444 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
2445 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[N_ADDR]], align 4
2446 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
2447 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2448 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
2449 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
2450 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB2]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..4 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
2451 // CHECK11-NEXT: ret void
2452 //
2453 //
2454 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
2455 // 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:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
2456 // CHECK11-NEXT: entry:
2457 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2458 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2459 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
2460 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
2461 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
2462 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
2463 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2464 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
2465 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2466 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
2467 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
2468 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2469 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2470 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2471 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2472 // CHECK11-NEXT: [[I4:%.*]] = alloca i32, align 4
2473 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2474 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2475 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
2476 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
2477 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
2478 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2479 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
2480 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
2481 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
2482 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2483 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2484 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2485 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2486 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2487 // CHECK11-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
2488 // CHECK11-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
2489 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
2490 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2491 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2492 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2493 // CHECK11: omp.precond.then:
2494 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2495 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2496 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
2497 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2498 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2499 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
2500 // CHECK11-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2501 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
2502 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
2503 // CHECK11-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
2504 // CHECK11: omp.dispatch.cond:
2505 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2506 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2507 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
2508 // CHECK11-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2509 // CHECK11: cond.true:
2510 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2511 // CHECK11-NEXT: br label [[COND_END:%.*]]
2512 // CHECK11: cond.false:
2513 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2514 // CHECK11-NEXT: br label [[COND_END]]
2515 // CHECK11: cond.end:
2516 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
2517 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2518 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2519 // CHECK11-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
2520 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2521 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2522 // CHECK11-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
2523 // CHECK11-NEXT: br i1 [[CMP6]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2524 // CHECK11: omp.dispatch.body:
2525 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2526 // CHECK11: omp.inner.for.cond:
2527 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
2528 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !12
2529 // CHECK11-NEXT: [[CMP7:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
2530 // CHECK11-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2531 // CHECK11: omp.inner.for.body:
2532 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
2533 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP19]], 1
2534 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2535 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I4]], align 4, !llvm.access.group !12
2536 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[I4]], align 4, !llvm.access.group !12
2537 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP20]]
2538 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !12
2539 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2540 // CHECK11: omp.body.continue:
2541 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2542 // CHECK11: omp.inner.for.inc:
2543 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
2544 // CHECK11-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP21]], 1
2545 // CHECK11-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !12
2546 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP13:![0-9]+]]
2547 // CHECK11: omp.inner.for.end:
2548 // CHECK11-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
2549 // CHECK11: omp.dispatch.inc:
2550 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2551 // CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2552 // CHECK11-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
2553 // CHECK11-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_LB]], align 4
2554 // CHECK11-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2555 // CHECK11-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2556 // CHECK11-NEXT: [[ADD10:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
2557 // CHECK11-NEXT: store i32 [[ADD10]], i32* [[DOTOMP_UB]], align 4
2558 // CHECK11-NEXT: br label [[OMP_DISPATCH_COND]]
2559 // CHECK11: omp.dispatch.end:
2560 // CHECK11-NEXT: [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2561 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
2562 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP27]])
2563 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
2564 // CHECK11: omp.precond.end:
2565 // CHECK11-NEXT: ret void
2566 //
2567 //
2568 // CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
2569 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
2570 // CHECK11-NEXT: entry:
2571 // CHECK11-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
2572 // CHECK11-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
2573 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
2574 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
2575 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
2576 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
2577 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
2578 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
2579 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
2580 // CHECK11-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
2581 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [1 x i8*], align 4
2582 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS9:%.*]] = alloca [1 x i8*], align 4
2583 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [1 x i8*], align 4
2584 // CHECK11-NEXT: [[_TMP11:%.*]] = alloca i32, align 4
2585 // CHECK11-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
2586 // CHECK11-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2587 // CHECK11-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
2588 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
2589 // CHECK11-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2590 // CHECK11-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
2591 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
2592 // CHECK11-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2593 // CHECK11-NEXT: store i8* null, i8** [[TMP4]], align 4
2594 // CHECK11-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2595 // CHECK11-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2596 // CHECK11-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2597 // CHECK11-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
2598 // CHECK11-NEXT: store i32 1, i32* [[TMP7]], align 4
2599 // CHECK11-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
2600 // CHECK11-NEXT: store i32 1, i32* [[TMP8]], align 4
2601 // CHECK11-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
2602 // CHECK11-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4
2603 // CHECK11-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
2604 // CHECK11-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4
2605 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
2606 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP11]], align 4
2607 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
2608 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP12]], align 4
2609 // CHECK11-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
2610 // CHECK11-NEXT: store i8** null, i8*** [[TMP13]], align 4
2611 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
2612 // CHECK11-NEXT: store i8** null, i8*** [[TMP14]], align 4
2613 // CHECK11-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
2614 // CHECK11-NEXT: store i64 10, i64* [[TMP15]], align 8
2615 // CHECK11-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
2616 // CHECK11-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
2617 // CHECK11-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2618 // CHECK11: omp_offload.failed:
2619 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79([10 x i32]* [[A]]) #[[ATTR3]]
2620 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
2621 // CHECK11: omp_offload.cont:
2622 // CHECK11-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
2623 // CHECK11-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to [10 x i32]**
2624 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP19]], align 4
2625 // CHECK11-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
2626 // CHECK11-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [10 x i32]**
2627 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP21]], align 4
2628 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
2629 // CHECK11-NEXT: store i8* null, i8** [[TMP22]], align 4
2630 // CHECK11-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
2631 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
2632 // CHECK11-NEXT: [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2633 // CHECK11-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 0
2634 // CHECK11-NEXT: store i32 1, i32* [[TMP25]], align 4
2635 // CHECK11-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 1
2636 // CHECK11-NEXT: store i32 1, i32* [[TMP26]], align 4
2637 // CHECK11-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 2
2638 // CHECK11-NEXT: store i8** [[TMP23]], i8*** [[TMP27]], align 4
2639 // CHECK11-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 3
2640 // CHECK11-NEXT: store i8** [[TMP24]], i8*** [[TMP28]], align 4
2641 // CHECK11-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 4
2642 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.11, i32 0, i32 0), i64** [[TMP29]], align 4
2643 // CHECK11-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 5
2644 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.12, i32 0, i32 0), i64** [[TMP30]], align 4
2645 // CHECK11-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 6
2646 // CHECK11-NEXT: store i8** null, i8*** [[TMP31]], align 4
2647 // CHECK11-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 7
2648 // CHECK11-NEXT: store i8** null, i8*** [[TMP32]], align 4
2649 // CHECK11-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 8
2650 // CHECK11-NEXT: store i64 10, i64* [[TMP33]], align 8
2651 // CHECK11-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]])
2652 // CHECK11-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
2653 // CHECK11-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
2654 // CHECK11: omp_offload.failed6:
2655 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84([10 x i32]* [[A]]) #[[ATTR3]]
2656 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT7]]
2657 // CHECK11: omp_offload.cont7:
2658 // CHECK11-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
2659 // CHECK11-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to [10 x i32]**
2660 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP37]], align 4
2661 // CHECK11-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
2662 // CHECK11-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [10 x i32]**
2663 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP39]], align 4
2664 // CHECK11-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
2665 // CHECK11-NEXT: store i8* null, i8** [[TMP40]], align 4
2666 // CHECK11-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
2667 // CHECK11-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
2668 // CHECK11-NEXT: [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2669 // CHECK11-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 0
2670 // CHECK11-NEXT: store i32 1, i32* [[TMP43]], align 4
2671 // CHECK11-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 1
2672 // CHECK11-NEXT: store i32 1, i32* [[TMP44]], align 4
2673 // CHECK11-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 2
2674 // CHECK11-NEXT: store i8** [[TMP41]], i8*** [[TMP45]], align 4
2675 // CHECK11-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 3
2676 // CHECK11-NEXT: store i8** [[TMP42]], i8*** [[TMP46]], align 4
2677 // CHECK11-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 4
2678 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.14, i32 0, i32 0), i64** [[TMP47]], align 4
2679 // CHECK11-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 5
2680 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.15, i32 0, i32 0), i64** [[TMP48]], align 4
2681 // CHECK11-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 6
2682 // CHECK11-NEXT: store i8** null, i8*** [[TMP49]], align 4
2683 // CHECK11-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 7
2684 // CHECK11-NEXT: store i8** null, i8*** [[TMP50]], align 4
2685 // CHECK11-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 8
2686 // CHECK11-NEXT: store i64 10, i64* [[TMP51]], align 8
2687 // CHECK11-NEXT: [[TMP52:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB2]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]])
2688 // CHECK11-NEXT: [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
2689 // CHECK11-NEXT: br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
2690 // CHECK11: omp_offload.failed13:
2691 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89([10 x i32]* [[A]]) #[[ATTR3]]
2692 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT14]]
2693 // CHECK11: omp_offload.cont14:
2694 // CHECK11-NEXT: ret i32 0
2695 //
2696 //
2697 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l79
2698 // CHECK11-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2699 // CHECK11-NEXT: entry:
2700 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
2701 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
2702 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
2703 // CHECK11-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..7 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
2704 // CHECK11-NEXT: ret void
2705 //
2706 //
2707 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..7
2708 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2709 // CHECK11-NEXT: entry:
2710 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2711 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2712 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
2713 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2714 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
2715 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2716 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2717 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2718 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2719 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
2720 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2721 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2722 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
2723 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
2724 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2725 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
2726 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2727 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2728 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2729 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2730 // CHECK11-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)
2731 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2732 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2733 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2734 // CHECK11: cond.true:
2735 // CHECK11-NEXT: br label [[COND_END:%.*]]
2736 // CHECK11: cond.false:
2737 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2738 // CHECK11-NEXT: br label [[COND_END]]
2739 // CHECK11: cond.end:
2740 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2741 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2742 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2743 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2744 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2745 // CHECK11: omp.inner.for.cond:
2746 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2747 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2748 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2749 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2750 // CHECK11: omp.inner.for.body:
2751 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2752 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2753 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2754 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2755 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
2756 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP9]]
2757 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2758 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2759 // CHECK11: omp.body.continue:
2760 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2761 // CHECK11: omp.inner.for.inc:
2762 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2763 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
2764 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2765 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
2766 // CHECK11: omp.inner.for.end:
2767 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2768 // CHECK11: omp.loop.exit:
2769 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2770 // CHECK11-NEXT: ret void
2771 //
2772 //
2773 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84
2774 // CHECK11-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2775 // CHECK11-NEXT: entry:
2776 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
2777 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
2778 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
2779 // CHECK11-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..10 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
2780 // CHECK11-NEXT: ret void
2781 //
2782 //
2783 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..10
2784 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2785 // CHECK11-NEXT: entry:
2786 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2787 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2788 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
2789 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2790 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
2791 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2792 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2793 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2794 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2795 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
2796 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2797 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2798 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
2799 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
2800 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2801 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
2802 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2803 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2804 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2805 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2806 // CHECK11-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)
2807 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2808 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2809 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2810 // CHECK11: cond.true:
2811 // CHECK11-NEXT: br label [[COND_END:%.*]]
2812 // CHECK11: cond.false:
2813 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2814 // CHECK11-NEXT: br label [[COND_END]]
2815 // CHECK11: cond.end:
2816 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2817 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2818 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2819 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2820 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2821 // CHECK11: omp.inner.for.cond:
2822 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2823 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2824 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2825 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2826 // CHECK11: omp.inner.for.body:
2827 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2828 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], 1
2829 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2830 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2831 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[I]], align 4
2832 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP9]]
2833 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2834 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2835 // CHECK11: omp.body.continue:
2836 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2837 // CHECK11: omp.inner.for.inc:
2838 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2839 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
2840 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
2841 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
2842 // CHECK11: omp.inner.for.end:
2843 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2844 // CHECK11: omp.loop.exit:
2845 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2846 // CHECK11-NEXT: ret void
2847 //
2848 //
2849 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89
2850 // CHECK11-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2851 // CHECK11-NEXT: entry:
2852 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
2853 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
2854 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
2855 // CHECK11-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..13 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
2856 // CHECK11-NEXT: ret void
2857 //
2858 //
2859 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..13
2860 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2861 // CHECK11-NEXT: entry:
2862 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
2863 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
2864 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
2865 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2866 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
2867 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2868 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2869 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2870 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2871 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
2872 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
2873 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
2874 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
2875 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
2876 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2877 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
2878 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2879 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2880 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
2881 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2882 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 10)
2883 // CHECK11-NEXT: br label [[OMP_DISPATCH_COND:%.*]]
2884 // CHECK11: omp.dispatch.cond:
2885 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2886 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2887 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2888 // CHECK11: cond.true:
2889 // CHECK11-NEXT: br label [[COND_END:%.*]]
2890 // CHECK11: cond.false:
2891 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2892 // CHECK11-NEXT: br label [[COND_END]]
2893 // CHECK11: cond.end:
2894 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2895 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2896 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2897 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2898 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2899 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2900 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2901 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2902 // CHECK11: omp.dispatch.body:
2903 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2904 // CHECK11: omp.inner.for.cond:
2905 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
2906 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !15
2907 // CHECK11-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2908 // CHECK11-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2909 // CHECK11: omp.inner.for.body:
2910 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
2911 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2912 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2913 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !15
2914 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4, !llvm.access.group !15
2915 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
2916 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4, !llvm.access.group !15
2917 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2918 // CHECK11: omp.body.continue:
2919 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2920 // CHECK11: omp.inner.for.inc:
2921 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
2922 // CHECK11-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
2923 // CHECK11-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
2924 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
2925 // CHECK11: omp.inner.for.end:
2926 // CHECK11-NEXT: br label [[OMP_DISPATCH_INC:%.*]]
2927 // CHECK11: omp.dispatch.inc:
2928 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2929 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2930 // CHECK11-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
2931 // CHECK11-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_LB]], align 4
2932 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2933 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2934 // CHECK11-NEXT: [[ADD5:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
2935 // CHECK11-NEXT: store i32 [[ADD5]], i32* [[DOTOMP_UB]], align 4
2936 // CHECK11-NEXT: br label [[OMP_DISPATCH_COND]]
2937 // CHECK11: omp.dispatch.end:
2938 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2939 // CHECK11-NEXT: ret void
2940 //
2941 //
2942 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2943 // CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
2944 // CHECK11-NEXT: entry:
2945 // CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
2946 // CHECK11-NEXT: ret void
2947 //
2948