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 parallel for
30 for(int i = 0; i < X; i++) {
31 a[i] = (T)0;
32 }
33 #pragma omp target
34 #pragma omp teams distribute parallel for dist_schedule(static)
35 for(int i = 0; i < X; i++) {
36 a[i] = (T)0;
37 }
38 #pragma omp target
39 #pragma omp teams distribute parallel for dist_schedule(static, X/2)
40 for(int i = 0; i < X; i++) {
41 a[i] = (T)0;
42 }
43
44
45
46
47
48
49
50
51
52
53 return a[0];
54 }
55 };
56
teams_template_struct(void)57 int teams_template_struct(void) {
58 SS<int, 123, 456> V;
59 return V.foo();
60
61 }
62 #endif // CK1
63
64 // Test host codegen.
65 // 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
66 // 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
67 // 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
68 // 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
69 // 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
70 // 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
71
72 // 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}}"
73 // 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
74 // 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}}"
75 // 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}}"
76 // 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
77 // 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}}"
78 #ifdef CK2
79
80 template <typename T, int n>
tmain(T argc)81 int tmain(T argc) {
82 T a[n];
83 int m = 10;
84 #pragma omp target
85 #pragma omp teams distribute parallel for
86 for(int i = 0; i < n; i++) {
87 a[i] = (T)0;
88 }
89 #pragma omp target
90 #pragma omp teams distribute parallel for dist_schedule(static)
91 for(int i = 0; i < n; i++) {
92 a[i] = (T)0;
93 }
94 #pragma omp target
95 #pragma omp teams distribute parallel for dist_schedule(static, m)
96 for(int i = 0; i < n; i++) {
97 a[i] = (T)0;
98 }
99 return 0;
100 }
101
main(int argc,char ** argv)102 int main (int argc, char **argv) {
103 int n = 100;
104 int a[n];
105 int m = 10;
106 #pragma omp target
107 #pragma omp teams distribute parallel for
108 for(int i = 0; i < n; i++) {
109 a[i] = 0;
110 }
111 #pragma omp target
112 #pragma omp teams distribute parallel for dist_schedule(static)
113 for(int i = 0; i < n; i++) {
114 a[i] = 0;
115 }
116 #pragma omp target
117 #pragma omp teams distribute parallel for dist_schedule(static, m)
118 for(int i = 0; i < n; i++) {
119 a[i] = 0;
120 }
121 return tmain<int, 10>(argc);
122 }
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144 #endif // CK2
145 #endif // #ifndef HEADER
146 // CHECK1-LABEL: define {{[^@]+}}@_Z21teams_template_structv
147 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
148 // CHECK1-NEXT: entry:
149 // CHECK1-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
150 // CHECK1-NEXT: [[CALL:%.*]] = call noundef signext i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
151 // CHECK1-NEXT: ret i32 [[CALL]]
152 //
153 //
154 // CHECK1-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
155 // CHECK1-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
156 // CHECK1-NEXT: entry:
157 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
158 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
159 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
160 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
161 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
162 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 8
163 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 8
164 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 8
165 // CHECK1-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
166 // CHECK1-NEXT: [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x i8*], align 8
167 // CHECK1-NEXT: [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 8
168 // CHECK1-NEXT: [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 8
169 // CHECK1-NEXT: [[_TMP14:%.*]] = alloca i32, align 4
170 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
171 // CHECK1-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
172 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
173 // CHECK1-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
174 // CHECK1-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
175 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 8
176 // CHECK1-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
177 // CHECK1-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
178 // CHECK1-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 8
179 // CHECK1-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
180 // CHECK1-NEXT: store i8* null, i8** [[TMP4]], align 8
181 // CHECK1-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
182 // CHECK1-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
183 // CHECK1-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
184 // CHECK1-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
185 // CHECK1-NEXT: store i32 1, i32* [[TMP7]], align 4
186 // CHECK1-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
187 // CHECK1-NEXT: store i32 1, i32* [[TMP8]], align 4
188 // CHECK1-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
189 // CHECK1-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8
190 // CHECK1-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
191 // CHECK1-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8
192 // CHECK1-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
193 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 8
194 // CHECK1-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
195 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 8
196 // CHECK1-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
197 // CHECK1-NEXT: store i8** null, i8*** [[TMP13]], align 8
198 // CHECK1-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
199 // CHECK1-NEXT: store i8** null, i8*** [[TMP14]], align 8
200 // CHECK1-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
201 // CHECK1-NEXT: store i64 123, i64* [[TMP15]], align 8
202 // CHECK1-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[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]])
203 // CHECK1-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
204 // CHECK1-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
205 // CHECK1: omp_offload.failed:
206 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
207 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT]]
208 // CHECK1: omp_offload.cont:
209 // CHECK1-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
210 // CHECK1-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
211 // CHECK1-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
212 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 8
213 // CHECK1-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
214 // CHECK1-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
215 // CHECK1-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP21]], align 8
216 // CHECK1-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i64 0, i64 0
217 // CHECK1-NEXT: store i8* null, i8** [[TMP22]], align 8
218 // CHECK1-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
219 // CHECK1-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
220 // CHECK1-NEXT: [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
221 // CHECK1-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
222 // CHECK1-NEXT: store i32 1, i32* [[TMP25]], align 4
223 // CHECK1-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
224 // CHECK1-NEXT: store i32 1, i32* [[TMP26]], align 4
225 // CHECK1-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
226 // CHECK1-NEXT: store i8** [[TMP23]], i8*** [[TMP27]], align 8
227 // CHECK1-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
228 // CHECK1-NEXT: store i8** [[TMP24]], i8*** [[TMP28]], align 8
229 // CHECK1-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
230 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP29]], align 8
231 // CHECK1-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
232 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP30]], align 8
233 // CHECK1-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
234 // CHECK1-NEXT: store i8** null, i8*** [[TMP31]], align 8
235 // CHECK1-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
236 // CHECK1-NEXT: store i8** null, i8*** [[TMP32]], align 8
237 // CHECK1-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
238 // CHECK1-NEXT: store i64 123, i64* [[TMP33]], align 8
239 // CHECK1-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], 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]])
240 // CHECK1-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
241 // CHECK1-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
242 // CHECK1: omp_offload.failed8:
243 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33(%struct.SS* [[THIS1]]) #[[ATTR2]]
244 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT9]]
245 // CHECK1: omp_offload.cont9:
246 // CHECK1-NEXT: [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
247 // CHECK1-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
248 // CHECK1-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
249 // CHECK1-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 8
250 // CHECK1-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
251 // CHECK1-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
252 // CHECK1-NEXT: store [123 x i32]* [[A10]], [123 x i32]** [[TMP39]], align 8
253 // CHECK1-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i64 0, i64 0
254 // CHECK1-NEXT: store i8* null, i8** [[TMP40]], align 8
255 // CHECK1-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
256 // CHECK1-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
257 // CHECK1-NEXT: [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
258 // CHECK1-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
259 // CHECK1-NEXT: store i32 1, i32* [[TMP43]], align 4
260 // CHECK1-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
261 // CHECK1-NEXT: store i32 1, i32* [[TMP44]], align 4
262 // CHECK1-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
263 // CHECK1-NEXT: store i8** [[TMP41]], i8*** [[TMP45]], align 8
264 // CHECK1-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
265 // CHECK1-NEXT: store i8** [[TMP42]], i8*** [[TMP46]], align 8
266 // CHECK1-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
267 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP47]], align 8
268 // CHECK1-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
269 // CHECK1-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP48]], align 8
270 // CHECK1-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
271 // CHECK1-NEXT: store i8** null, i8*** [[TMP49]], align 8
272 // CHECK1-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
273 // CHECK1-NEXT: store i8** null, i8*** [[TMP50]], align 8
274 // CHECK1-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
275 // CHECK1-NEXT: store i64 123, i64* [[TMP51]], align 8
276 // CHECK1-NEXT: [[TMP52:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], 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]])
277 // CHECK1-NEXT: [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
278 // CHECK1-NEXT: br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
279 // CHECK1: omp_offload.failed16:
280 // CHECK1-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38(%struct.SS* [[THIS1]]) #[[ATTR2]]
281 // CHECK1-NEXT: br label [[OMP_OFFLOAD_CONT17]]
282 // CHECK1: omp_offload.cont17:
283 // CHECK1-NEXT: [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
284 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A18]], i64 0, i64 0
285 // CHECK1-NEXT: [[TMP54:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
286 // CHECK1-NEXT: ret i32 [[TMP54]]
287 //
288 //
289 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
290 // CHECK1-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
291 // CHECK1-NEXT: entry:
292 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
293 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
294 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
295 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
296 // CHECK1-NEXT: ret void
297 //
298 //
299 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
300 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
301 // CHECK1-NEXT: entry:
302 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
303 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
304 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
305 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
306 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
307 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
308 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
309 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
310 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
311 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
312 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
313 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
314 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
315 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
316 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
317 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
318 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
319 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
320 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
321 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
322 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
323 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
324 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
325 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
326 // CHECK1: cond.true:
327 // CHECK1-NEXT: br label [[COND_END:%.*]]
328 // CHECK1: cond.false:
329 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
330 // CHECK1-NEXT: br label [[COND_END]]
331 // CHECK1: cond.end:
332 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
333 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
334 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
335 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
336 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
337 // CHECK1: omp.inner.for.cond:
338 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
339 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
340 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
341 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
342 // CHECK1: omp.inner.for.body:
343 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
344 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
345 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
346 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
347 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
348 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
349 // CHECK1: omp.inner.for.inc:
350 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
351 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
352 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
353 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
354 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
355 // CHECK1: omp.inner.for.end:
356 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
357 // CHECK1: omp.loop.exit:
358 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
359 // CHECK1-NEXT: ret void
360 //
361 //
362 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
363 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
364 // CHECK1-NEXT: entry:
365 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
366 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
367 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
368 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
369 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
370 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
371 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
372 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
373 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
374 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
375 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
376 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
377 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
378 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
379 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
380 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
381 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
382 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
383 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
384 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
385 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
386 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
387 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
388 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
389 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
390 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
391 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
392 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
393 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
394 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
395 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
396 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
397 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
398 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
399 // CHECK1: cond.true:
400 // CHECK1-NEXT: br label [[COND_END:%.*]]
401 // CHECK1: cond.false:
402 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
403 // CHECK1-NEXT: br label [[COND_END]]
404 // CHECK1: cond.end:
405 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
406 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
407 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
408 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
409 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
410 // CHECK1: omp.inner.for.cond:
411 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
412 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
413 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
414 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
415 // CHECK1: omp.inner.for.body:
416 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
417 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
418 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
419 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
420 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
421 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
422 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
423 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
424 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
425 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
426 // CHECK1: omp.body.continue:
427 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
428 // CHECK1: omp.inner.for.inc:
429 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
430 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
431 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
432 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
433 // CHECK1: omp.inner.for.end:
434 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
435 // CHECK1: omp.loop.exit:
436 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
437 // CHECK1-NEXT: ret void
438 //
439 //
440 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33
441 // CHECK1-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
442 // CHECK1-NEXT: entry:
443 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
444 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
445 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
446 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
447 // CHECK1-NEXT: ret void
448 //
449 //
450 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
451 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
452 // CHECK1-NEXT: entry:
453 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
454 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
455 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
456 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
457 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
458 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
459 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
460 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
461 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
462 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
463 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
464 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
465 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
466 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
467 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
468 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
469 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
470 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
471 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
472 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
473 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
474 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
475 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
476 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
477 // CHECK1: cond.true:
478 // CHECK1-NEXT: br label [[COND_END:%.*]]
479 // CHECK1: cond.false:
480 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
481 // CHECK1-NEXT: br label [[COND_END]]
482 // CHECK1: cond.end:
483 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
484 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
485 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
486 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
487 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
488 // CHECK1: omp.inner.for.cond:
489 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
490 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
491 // CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
492 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
493 // CHECK1: omp.inner.for.body:
494 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
495 // CHECK1-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
496 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
497 // CHECK1-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
498 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], %struct.SS* [[TMP0]])
499 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
500 // CHECK1: omp.inner.for.inc:
501 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
502 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
503 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
504 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
505 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
506 // CHECK1: omp.inner.for.end:
507 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
508 // CHECK1: omp.loop.exit:
509 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
510 // CHECK1-NEXT: ret void
511 //
512 //
513 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
514 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
515 // CHECK1-NEXT: entry:
516 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
517 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
518 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
519 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
520 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
521 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
522 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
523 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
524 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
525 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
526 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
527 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
528 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
529 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
530 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
531 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
532 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
533 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
534 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
535 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
536 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
537 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
538 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
539 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
540 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
541 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
542 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
543 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
544 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
545 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
546 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
547 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
548 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
549 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
550 // CHECK1: cond.true:
551 // CHECK1-NEXT: br label [[COND_END:%.*]]
552 // CHECK1: cond.false:
553 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
554 // CHECK1-NEXT: br label [[COND_END]]
555 // CHECK1: cond.end:
556 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
557 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
558 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
559 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
560 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
561 // CHECK1: omp.inner.for.cond:
562 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
563 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
564 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
565 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
566 // CHECK1: omp.inner.for.body:
567 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
568 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
569 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
570 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
571 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
572 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
573 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
574 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
575 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
576 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
577 // CHECK1: omp.body.continue:
578 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
579 // CHECK1: omp.inner.for.inc:
580 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
581 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
582 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
583 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
584 // CHECK1: omp.inner.for.end:
585 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
586 // CHECK1: omp.loop.exit:
587 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
588 // CHECK1-NEXT: ret void
589 //
590 //
591 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38
592 // CHECK1-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
593 // CHECK1-NEXT: entry:
594 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
595 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
596 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
597 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
598 // CHECK1-NEXT: ret void
599 //
600 //
601 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
602 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
603 // CHECK1-NEXT: entry:
604 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
605 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
606 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
607 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
608 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
609 // CHECK1-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
610 // CHECK1-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
611 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
612 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
613 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
614 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
615 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
616 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
617 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
618 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
619 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
620 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
621 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
622 // CHECK1-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
623 // CHECK1-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
624 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
625 // CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
626 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
627 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
628 // CHECK1: cond.true:
629 // CHECK1-NEXT: br label [[COND_END:%.*]]
630 // CHECK1: cond.false:
631 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
632 // CHECK1-NEXT: br label [[COND_END]]
633 // CHECK1: cond.end:
634 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
635 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
636 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
637 // CHECK1-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
638 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
639 // CHECK1: omp.inner.for.cond:
640 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
641 // CHECK1-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 123
642 // CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
643 // CHECK1: omp.inner.for.body:
644 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
645 // CHECK1-NEXT: [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
646 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
647 // CHECK1-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
648 // CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]], %struct.SS* [[TMP0]])
649 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
650 // CHECK1: omp.inner.for.inc:
651 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
652 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
653 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
654 // CHECK1-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
655 // CHECK1-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
656 // CHECK1-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
657 // CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
658 // CHECK1-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4
659 // CHECK1-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
660 // CHECK1-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
661 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP15]], [[TMP16]]
662 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4
663 // CHECK1-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
664 // CHECK1-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP17]], 122
665 // CHECK1-NEXT: br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
666 // CHECK1: cond.true5:
667 // CHECK1-NEXT: br label [[COND_END7:%.*]]
668 // CHECK1: cond.false6:
669 // CHECK1-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
670 // CHECK1-NEXT: br label [[COND_END7]]
671 // CHECK1: cond.end7:
672 // CHECK1-NEXT: [[COND8:%.*]] = phi i32 [ 122, [[COND_TRUE5]] ], [ [[TMP18]], [[COND_FALSE6]] ]
673 // CHECK1-NEXT: store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4
674 // CHECK1-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
675 // CHECK1-NEXT: store i32 [[TMP19]], i32* [[DOTOMP_IV]], align 4
676 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
677 // CHECK1: omp.inner.for.end:
678 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
679 // CHECK1: omp.loop.exit:
680 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
681 // CHECK1-NEXT: ret void
682 //
683 //
684 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
685 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
686 // CHECK1-NEXT: entry:
687 // CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
688 // CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
689 // CHECK1-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
690 // CHECK1-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
691 // CHECK1-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 8
692 // CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
693 // CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
694 // CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
695 // CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
696 // CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
697 // CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
698 // CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
699 // CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
700 // CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
701 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
702 // CHECK1-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
703 // CHECK1-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 8
704 // CHECK1-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 8
705 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
706 // CHECK1-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
707 // CHECK1-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
708 // CHECK1-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
709 // CHECK1-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
710 // CHECK1-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
711 // CHECK1-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
712 // CHECK1-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
713 // CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
714 // CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
715 // CHECK1-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
716 // CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
717 // CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
718 // CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
719 // CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
720 // CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
721 // CHECK1: cond.true:
722 // CHECK1-NEXT: br label [[COND_END:%.*]]
723 // CHECK1: cond.false:
724 // CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
725 // CHECK1-NEXT: br label [[COND_END]]
726 // CHECK1: cond.end:
727 // CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
728 // CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
729 // CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
730 // CHECK1-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
731 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
732 // CHECK1: omp.inner.for.cond:
733 // CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
734 // CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
735 // CHECK1-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
736 // CHECK1-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
737 // CHECK1: omp.inner.for.body:
738 // CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
739 // CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
740 // CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
741 // CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
742 // CHECK1-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
743 // CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
744 // CHECK1-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
745 // CHECK1-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i64 0, i64 [[IDXPROM]]
746 // CHECK1-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
747 // CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
748 // CHECK1: omp.body.continue:
749 // CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
750 // CHECK1: omp.inner.for.inc:
751 // CHECK1-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
752 // CHECK1-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
753 // CHECK1-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
754 // CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]]
755 // CHECK1: omp.inner.for.end:
756 // CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
757 // CHECK1: omp.loop.exit:
758 // CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
759 // CHECK1-NEXT: ret void
760 //
761 //
762 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
763 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
764 // CHECK1-NEXT: entry:
765 // CHECK1-NEXT: call void @__tgt_register_requires(i64 1)
766 // CHECK1-NEXT: ret void
767 //
768 //
769 // CHECK3-LABEL: define {{[^@]+}}@_Z21teams_template_structv
770 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
771 // CHECK3-NEXT: entry:
772 // CHECK3-NEXT: [[V:%.*]] = alloca [[STRUCT_SS:%.*]], align 4
773 // CHECK3-NEXT: [[CALL:%.*]] = call noundef i32 @_ZN2SSIiLi123ELx456EE3fooEv(%struct.SS* noundef nonnull align 4 dereferenceable(496) [[V]])
774 // CHECK3-NEXT: ret i32 [[CALL]]
775 //
776 //
777 // CHECK3-LABEL: define {{[^@]+}}@_ZN2SSIiLi123ELx456EE3fooEv
778 // CHECK3-SAME: (%struct.SS* noundef nonnull align 4 dereferenceable(496) [[THIS:%.*]]) #[[ATTR0]] comdat align 2 {
779 // CHECK3-NEXT: entry:
780 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
781 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
782 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
783 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
784 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
785 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS3:%.*]] = alloca [1 x i8*], align 4
786 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS4:%.*]] = alloca [1 x i8*], align 4
787 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS5:%.*]] = alloca [1 x i8*], align 4
788 // CHECK3-NEXT: [[_TMP6:%.*]] = alloca i32, align 4
789 // CHECK3-NEXT: [[DOTOFFLOAD_BASEPTRS11:%.*]] = alloca [1 x i8*], align 4
790 // CHECK3-NEXT: [[DOTOFFLOAD_PTRS12:%.*]] = alloca [1 x i8*], align 4
791 // CHECK3-NEXT: [[DOTOFFLOAD_MAPPERS13:%.*]] = alloca [1 x i8*], align 4
792 // CHECK3-NEXT: [[_TMP14:%.*]] = alloca i32, align 4
793 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
794 // CHECK3-NEXT: [[THIS1:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
795 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[THIS1]], i32 0, i32 0
796 // CHECK3-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
797 // CHECK3-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to %struct.SS**
798 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP1]], align 4
799 // CHECK3-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
800 // CHECK3-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [123 x i32]**
801 // CHECK3-NEXT: store [123 x i32]* [[A]], [123 x i32]** [[TMP3]], align 4
802 // CHECK3-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
803 // CHECK3-NEXT: store i8* null, i8** [[TMP4]], align 4
804 // CHECK3-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
805 // CHECK3-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
806 // CHECK3-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
807 // CHECK3-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
808 // CHECK3-NEXT: store i32 1, i32* [[TMP7]], align 4
809 // CHECK3-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
810 // CHECK3-NEXT: store i32 1, i32* [[TMP8]], align 4
811 // CHECK3-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
812 // CHECK3-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4
813 // CHECK3-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
814 // CHECK3-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4
815 // CHECK3-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
816 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64** [[TMP11]], align 4
817 // CHECK3-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
818 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP12]], align 4
819 // CHECK3-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
820 // CHECK3-NEXT: store i8** null, i8*** [[TMP13]], align 4
821 // CHECK3-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
822 // CHECK3-NEXT: store i8** null, i8*** [[TMP14]], align 4
823 // CHECK3-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
824 // CHECK3-NEXT: store i64 123, i64* [[TMP15]], align 8
825 // CHECK3-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[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]])
826 // CHECK3-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
827 // CHECK3-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
828 // CHECK3: omp_offload.failed:
829 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28(%struct.SS* [[THIS1]]) #[[ATTR2:[0-9]+]]
830 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT]]
831 // CHECK3: omp_offload.cont:
832 // CHECK3-NEXT: [[A2:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
833 // CHECK3-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
834 // CHECK3-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to %struct.SS**
835 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP19]], align 4
836 // CHECK3-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
837 // CHECK3-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [123 x i32]**
838 // CHECK3-NEXT: store [123 x i32]* [[A2]], [123 x i32]** [[TMP21]], align 4
839 // CHECK3-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS5]], i32 0, i32 0
840 // CHECK3-NEXT: store i8* null, i8** [[TMP22]], align 4
841 // CHECK3-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS3]], i32 0, i32 0
842 // CHECK3-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS4]], i32 0, i32 0
843 // CHECK3-NEXT: [[KERNEL_ARGS7:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
844 // CHECK3-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 0
845 // CHECK3-NEXT: store i32 1, i32* [[TMP25]], align 4
846 // CHECK3-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 1
847 // CHECK3-NEXT: store i32 1, i32* [[TMP26]], align 4
848 // CHECK3-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 2
849 // CHECK3-NEXT: store i8** [[TMP23]], i8*** [[TMP27]], align 4
850 // CHECK3-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 3
851 // CHECK3-NEXT: store i8** [[TMP24]], i8*** [[TMP28]], align 4
852 // CHECK3-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 4
853 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.4, i32 0, i32 0), i64** [[TMP29]], align 4
854 // CHECK3-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 5
855 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP30]], align 4
856 // CHECK3-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 6
857 // CHECK3-NEXT: store i8** null, i8*** [[TMP31]], align 4
858 // CHECK3-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 7
859 // CHECK3-NEXT: store i8** null, i8*** [[TMP32]], align 4
860 // CHECK3-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS7]], i32 0, i32 8
861 // CHECK3-NEXT: store i64 123, i64* [[TMP33]], align 8
862 // CHECK3-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], 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]])
863 // CHECK3-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
864 // CHECK3-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED8:%.*]], label [[OMP_OFFLOAD_CONT9:%.*]]
865 // CHECK3: omp_offload.failed8:
866 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33(%struct.SS* [[THIS1]]) #[[ATTR2]]
867 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT9]]
868 // CHECK3: omp_offload.cont9:
869 // CHECK3-NEXT: [[A10:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
870 // CHECK3-NEXT: [[TMP36:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
871 // CHECK3-NEXT: [[TMP37:%.*]] = bitcast i8** [[TMP36]] to %struct.SS**
872 // CHECK3-NEXT: store %struct.SS* [[THIS1]], %struct.SS** [[TMP37]], align 4
873 // CHECK3-NEXT: [[TMP38:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
874 // CHECK3-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to [123 x i32]**
875 // CHECK3-NEXT: store [123 x i32]* [[A10]], [123 x i32]** [[TMP39]], align 4
876 // CHECK3-NEXT: [[TMP40:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS13]], i32 0, i32 0
877 // CHECK3-NEXT: store i8* null, i8** [[TMP40]], align 4
878 // CHECK3-NEXT: [[TMP41:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS11]], i32 0, i32 0
879 // CHECK3-NEXT: [[TMP42:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS12]], i32 0, i32 0
880 // CHECK3-NEXT: [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
881 // CHECK3-NEXT: [[TMP43:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
882 // CHECK3-NEXT: store i32 1, i32* [[TMP43]], align 4
883 // CHECK3-NEXT: [[TMP44:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
884 // CHECK3-NEXT: store i32 1, i32* [[TMP44]], align 4
885 // CHECK3-NEXT: [[TMP45:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
886 // CHECK3-NEXT: store i8** [[TMP41]], i8*** [[TMP45]], align 4
887 // CHECK3-NEXT: [[TMP46:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
888 // CHECK3-NEXT: store i8** [[TMP42]], i8*** [[TMP46]], align 4
889 // CHECK3-NEXT: [[TMP47:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
890 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.8, i32 0, i32 0), i64** [[TMP47]], align 4
891 // CHECK3-NEXT: [[TMP48:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
892 // CHECK3-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP48]], align 4
893 // CHECK3-NEXT: [[TMP49:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
894 // CHECK3-NEXT: store i8** null, i8*** [[TMP49]], align 4
895 // CHECK3-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
896 // CHECK3-NEXT: store i8** null, i8*** [[TMP50]], align 4
897 // CHECK3-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
898 // CHECK3-NEXT: store i64 123, i64* [[TMP51]], align 8
899 // CHECK3-NEXT: [[TMP52:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], 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]])
900 // CHECK3-NEXT: [[TMP53:%.*]] = icmp ne i32 [[TMP52]], 0
901 // CHECK3-NEXT: br i1 [[TMP53]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
902 // CHECK3: omp_offload.failed16:
903 // CHECK3-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38(%struct.SS* [[THIS1]]) #[[ATTR2]]
904 // CHECK3-NEXT: br label [[OMP_OFFLOAD_CONT17]]
905 // CHECK3: omp_offload.cont17:
906 // CHECK3-NEXT: [[A18:%.*]] = getelementptr inbounds [[STRUCT_SS]], %struct.SS* [[THIS1]], i32 0, i32 0
907 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A18]], i32 0, i32 0
908 // CHECK3-NEXT: [[TMP54:%.*]] = load i32, i32* [[ARRAYIDX]], align 4
909 // CHECK3-NEXT: ret i32 [[TMP54]]
910 //
911 //
912 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l28
913 // CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1:[0-9]+]] {
914 // CHECK3-NEXT: entry:
915 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
916 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
917 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
918 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
919 // CHECK3-NEXT: ret void
920 //
921 //
922 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
923 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
924 // CHECK3-NEXT: entry:
925 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
926 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
927 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
928 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
929 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
930 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
931 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
932 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
933 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
934 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
935 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
936 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
937 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
938 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
939 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
940 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
941 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
942 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
943 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
944 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
945 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
946 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
947 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
948 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
949 // CHECK3: cond.true:
950 // CHECK3-NEXT: br label [[COND_END:%.*]]
951 // CHECK3: cond.false:
952 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
953 // CHECK3-NEXT: br label [[COND_END]]
954 // CHECK3: cond.end:
955 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
956 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
957 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
958 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
959 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
960 // CHECK3: omp.inner.for.cond:
961 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
962 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
963 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
964 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
965 // CHECK3: omp.inner.for.body:
966 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
967 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
968 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
969 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
970 // CHECK3: omp.inner.for.inc:
971 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
972 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
973 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
974 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
975 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
976 // CHECK3: omp.inner.for.end:
977 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
978 // CHECK3: omp.loop.exit:
979 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
980 // CHECK3-NEXT: ret void
981 //
982 //
983 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
984 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
985 // CHECK3-NEXT: entry:
986 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
987 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
988 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
989 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
990 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
991 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
992 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
993 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
994 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
995 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
996 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
997 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
998 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
999 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1000 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1001 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1002 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1003 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1004 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1005 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1006 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1007 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1008 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
1009 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
1010 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1011 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1012 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1013 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1014 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1015 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1016 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1017 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1018 // CHECK3: cond.true:
1019 // CHECK3-NEXT: br label [[COND_END:%.*]]
1020 // CHECK3: cond.false:
1021 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1022 // CHECK3-NEXT: br label [[COND_END]]
1023 // CHECK3: cond.end:
1024 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1025 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1026 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1027 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1028 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1029 // CHECK3: omp.inner.for.cond:
1030 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1031 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1032 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1033 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1034 // CHECK3: omp.inner.for.body:
1035 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1036 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1037 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1038 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1039 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1040 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
1041 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
1042 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1043 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1044 // CHECK3: omp.body.continue:
1045 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1046 // CHECK3: omp.inner.for.inc:
1047 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1048 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
1049 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
1050 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
1051 // CHECK3: omp.inner.for.end:
1052 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1053 // CHECK3: omp.loop.exit:
1054 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1055 // CHECK3-NEXT: ret void
1056 //
1057 //
1058 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l33
1059 // CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1060 // CHECK3-NEXT: entry:
1061 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1062 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1063 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1064 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
1065 // CHECK3-NEXT: ret void
1066 //
1067 //
1068 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
1069 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1070 // CHECK3-NEXT: entry:
1071 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1072 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1073 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1074 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1075 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1076 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1077 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1078 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1079 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1080 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1081 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1082 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1083 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1084 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1085 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1086 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1087 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1088 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1089 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1090 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1091 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1092 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1093 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1094 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1095 // CHECK3: cond.true:
1096 // CHECK3-NEXT: br label [[COND_END:%.*]]
1097 // CHECK3: cond.false:
1098 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1099 // CHECK3-NEXT: br label [[COND_END]]
1100 // CHECK3: cond.end:
1101 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1102 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1103 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1104 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1105 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1106 // CHECK3: omp.inner.for.cond:
1107 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1108 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1109 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1110 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1111 // CHECK3: omp.inner.for.body:
1112 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1113 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1114 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], %struct.SS* [[TMP0]])
1115 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1116 // CHECK3: omp.inner.for.inc:
1117 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1118 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1119 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
1120 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1121 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
1122 // CHECK3: omp.inner.for.end:
1123 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1124 // CHECK3: omp.loop.exit:
1125 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1126 // CHECK3-NEXT: ret void
1127 //
1128 //
1129 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
1130 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1131 // CHECK3-NEXT: entry:
1132 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1133 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1134 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1135 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1136 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1137 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1138 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1139 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1140 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1141 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1142 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1143 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1144 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1145 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1146 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1147 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1148 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1149 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1150 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1151 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1152 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1153 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1154 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
1155 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
1156 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1157 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1158 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1159 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1160 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1161 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1162 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1163 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1164 // CHECK3: cond.true:
1165 // CHECK3-NEXT: br label [[COND_END:%.*]]
1166 // CHECK3: cond.false:
1167 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1168 // CHECK3-NEXT: br label [[COND_END]]
1169 // CHECK3: cond.end:
1170 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1171 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1172 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1173 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1174 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1175 // CHECK3: omp.inner.for.cond:
1176 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1177 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1178 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1179 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1180 // CHECK3: omp.inner.for.body:
1181 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1182 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1183 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1184 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1185 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1186 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
1187 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
1188 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1189 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1190 // CHECK3: omp.body.continue:
1191 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1192 // CHECK3: omp.inner.for.inc:
1193 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1194 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
1195 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
1196 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
1197 // CHECK3: omp.inner.for.end:
1198 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1199 // CHECK3: omp.loop.exit:
1200 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1201 // CHECK3-NEXT: ret void
1202 //
1203 //
1204 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__ZN2SSIiLi123ELx456EE3fooEv_l38
1205 // CHECK3-SAME: (%struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1206 // CHECK3-NEXT: entry:
1207 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1208 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1209 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1210 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.SS*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.SS* [[TMP0]])
1211 // CHECK3-NEXT: ret void
1212 //
1213 //
1214 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..6
1215 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1216 // CHECK3-NEXT: entry:
1217 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1218 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1219 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1220 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1221 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1222 // CHECK3-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1223 // CHECK3-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1224 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1225 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1226 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1227 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1228 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1229 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1230 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1231 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1232 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_COMB_UB]], align 4
1233 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1234 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1235 // CHECK3-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1236 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1237 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 61)
1238 // CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1239 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 122
1240 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1241 // CHECK3: cond.true:
1242 // CHECK3-NEXT: br label [[COND_END:%.*]]
1243 // CHECK3: cond.false:
1244 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1245 // CHECK3-NEXT: br label [[COND_END]]
1246 // CHECK3: cond.end:
1247 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1248 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1249 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1250 // CHECK3-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1251 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1252 // CHECK3: omp.inner.for.cond:
1253 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1254 // CHECK3-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP6]], 123
1255 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1256 // CHECK3: omp.inner.for.body:
1257 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1258 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1259 // CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, %struct.SS*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP7]], i32 [[TMP8]], %struct.SS* [[TMP0]])
1260 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1261 // CHECK3: omp.inner.for.inc:
1262 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1263 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1264 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP9]], [[TMP10]]
1265 // CHECK3-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1266 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1267 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1268 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1269 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4
1270 // CHECK3-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1271 // CHECK3-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1272 // CHECK3-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
1273 // CHECK3-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4
1274 // CHECK3-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1275 // CHECK3-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP15]], 122
1276 // CHECK3-NEXT: br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
1277 // CHECK3: cond.true5:
1278 // CHECK3-NEXT: br label [[COND_END7:%.*]]
1279 // CHECK3: cond.false6:
1280 // CHECK3-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1281 // CHECK3-NEXT: br label [[COND_END7]]
1282 // CHECK3: cond.end7:
1283 // CHECK3-NEXT: [[COND8:%.*]] = phi i32 [ 122, [[COND_TRUE5]] ], [ [[TMP16]], [[COND_FALSE6]] ]
1284 // CHECK3-NEXT: store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4
1285 // CHECK3-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1286 // CHECK3-NEXT: store i32 [[TMP17]], i32* [[DOTOMP_IV]], align 4
1287 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
1288 // CHECK3: omp.inner.for.end:
1289 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1290 // CHECK3: omp.loop.exit:
1291 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1292 // CHECK3-NEXT: ret void
1293 //
1294 //
1295 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
1296 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], %struct.SS* noundef [[THIS:%.*]]) #[[ATTR1]] {
1297 // CHECK3-NEXT: entry:
1298 // CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
1299 // CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
1300 // CHECK3-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1301 // CHECK3-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1302 // CHECK3-NEXT: [[THIS_ADDR:%.*]] = alloca %struct.SS*, align 4
1303 // CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1304 // CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
1305 // CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1306 // CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1307 // CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1308 // CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1309 // CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
1310 // CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
1311 // CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
1312 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1313 // CHECK3-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1314 // CHECK3-NEXT: store %struct.SS* [[THIS]], %struct.SS** [[THIS_ADDR]], align 4
1315 // CHECK3-NEXT: [[TMP0:%.*]] = load %struct.SS*, %struct.SS** [[THIS_ADDR]], align 4
1316 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1317 // CHECK3-NEXT: store i32 122, i32* [[DOTOMP_UB]], align 4
1318 // CHECK3-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
1319 // CHECK3-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
1320 // CHECK3-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
1321 // CHECK3-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
1322 // CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1323 // CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1324 // CHECK3-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
1325 // CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
1326 // CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1327 // CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1328 // CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 122
1329 // CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1330 // CHECK3: cond.true:
1331 // CHECK3-NEXT: br label [[COND_END:%.*]]
1332 // CHECK3: cond.false:
1333 // CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1334 // CHECK3-NEXT: br label [[COND_END]]
1335 // CHECK3: cond.end:
1336 // CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 122, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
1337 // CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1338 // CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1339 // CHECK3-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
1340 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1341 // CHECK3: omp.inner.for.cond:
1342 // CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1343 // CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1344 // CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
1345 // CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1346 // CHECK3: omp.inner.for.body:
1347 // CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1348 // CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
1349 // CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1350 // CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4
1351 // CHECK3-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_SS:%.*]], %struct.SS* [[TMP0]], i32 0, i32 0
1352 // CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
1353 // CHECK3-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [123 x i32], [123 x i32]* [[A]], i32 0, i32 [[TMP11]]
1354 // CHECK3-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1355 // CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1356 // CHECK3: omp.body.continue:
1357 // CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1358 // CHECK3: omp.inner.for.inc:
1359 // CHECK3-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1360 // CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
1361 // CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
1362 // CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]]
1363 // CHECK3: omp.inner.for.end:
1364 // CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1365 // CHECK3: omp.loop.exit:
1366 // CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
1367 // CHECK3-NEXT: ret void
1368 //
1369 //
1370 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1371 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
1372 // CHECK3-NEXT: entry:
1373 // CHECK3-NEXT: call void @__tgt_register_requires(i64 1)
1374 // CHECK3-NEXT: ret void
1375 //
1376 //
1377 // CHECK9-LABEL: define {{[^@]+}}@main
1378 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
1379 // CHECK9-NEXT: entry:
1380 // CHECK9-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
1381 // CHECK9-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
1382 // CHECK9-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 8
1383 // CHECK9-NEXT: [[N:%.*]] = alloca i32, align 4
1384 // CHECK9-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 8
1385 // CHECK9-NEXT: [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1386 // CHECK9-NEXT: [[M:%.*]] = alloca i32, align 4
1387 // CHECK9-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8
1388 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 8
1389 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 8
1390 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 8
1391 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 8
1392 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1393 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1394 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1395 // CHECK9-NEXT: [[N_CASTED3:%.*]] = alloca i64, align 8
1396 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS5:%.*]] = alloca [3 x i8*], align 8
1397 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS6:%.*]] = alloca [3 x i8*], align 8
1398 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS7:%.*]] = alloca [3 x i8*], align 8
1399 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES8:%.*]] = alloca [3 x i64], align 8
1400 // CHECK9-NEXT: [[_TMP9:%.*]] = alloca i32, align 4
1401 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
1402 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_11:%.*]] = alloca i32, align 4
1403 // CHECK9-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
1404 // CHECK9-NEXT: [[N_CASTED20:%.*]] = alloca i64, align 8
1405 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS22:%.*]] = alloca [4 x i8*], align 8
1406 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS23:%.*]] = alloca [4 x i8*], align 8
1407 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS24:%.*]] = alloca [4 x i8*], align 8
1408 // CHECK9-NEXT: [[DOTOFFLOAD_SIZES25:%.*]] = alloca [4 x i64], align 8
1409 // CHECK9-NEXT: [[_TMP26:%.*]] = alloca i32, align 4
1410 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_27:%.*]] = alloca i32, align 4
1411 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_28:%.*]] = alloca i32, align 4
1412 // CHECK9-NEXT: store i32 0, i32* [[RETVAL]], align 4
1413 // CHECK9-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
1414 // CHECK9-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 8
1415 // CHECK9-NEXT: store i32 100, i32* [[N]], align 4
1416 // CHECK9-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
1417 // CHECK9-NEXT: [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
1418 // CHECK9-NEXT: [[TMP2:%.*]] = call i8* @llvm.stacksave()
1419 // CHECK9-NEXT: store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
1420 // CHECK9-NEXT: [[VLA:%.*]] = alloca i32, i64 [[TMP1]], align 4
1421 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
1422 // CHECK9-NEXT: store i32 10, i32* [[M]], align 4
1423 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[N]], align 4
1424 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_CASTED]] to i32*
1425 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV]], align 4
1426 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[N_CASTED]], align 8
1427 // CHECK9-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP1]], 4
1428 // CHECK9-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
1429 // 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)
1430 // CHECK9-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1431 // CHECK9-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i64*
1432 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP8]], align 8
1433 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1434 // CHECK9-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i64*
1435 // CHECK9-NEXT: store i64 [[TMP4]], i64* [[TMP10]], align 8
1436 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1437 // CHECK9-NEXT: store i8* null, i8** [[TMP11]], align 8
1438 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1439 // CHECK9-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i64*
1440 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP13]], align 8
1441 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1442 // CHECK9-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i64*
1443 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP15]], align 8
1444 // CHECK9-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1445 // CHECK9-NEXT: store i8* null, i8** [[TMP16]], align 8
1446 // CHECK9-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1447 // CHECK9-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
1448 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 8
1449 // CHECK9-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1450 // CHECK9-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
1451 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 8
1452 // CHECK9-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
1453 // CHECK9-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 8
1454 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1455 // CHECK9-NEXT: store i8* null, i8** [[TMP22]], align 8
1456 // CHECK9-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1457 // CHECK9-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1458 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
1459 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
1460 // CHECK9-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
1461 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1462 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
1463 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1464 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1465 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1466 // CHECK9-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1467 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
1468 // CHECK9-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
1469 // CHECK9-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1470 // CHECK9-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
1471 // CHECK9-NEXT: store i32 1, i32* [[TMP30]], align 4
1472 // CHECK9-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
1473 // CHECK9-NEXT: store i32 3, i32* [[TMP31]], align 4
1474 // CHECK9-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
1475 // CHECK9-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 8
1476 // CHECK9-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
1477 // CHECK9-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 8
1478 // CHECK9-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
1479 // CHECK9-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 8
1480 // CHECK9-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
1481 // CHECK9-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 8
1482 // CHECK9-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
1483 // CHECK9-NEXT: store i8** null, i8*** [[TMP36]], align 8
1484 // CHECK9-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
1485 // CHECK9-NEXT: store i8** null, i8*** [[TMP37]], align 8
1486 // CHECK9-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
1487 // CHECK9-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
1488 // CHECK9-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l106.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
1489 // CHECK9-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
1490 // CHECK9-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1491 // CHECK9: omp_offload.failed:
1492 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l106(i64 [[TMP4]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
1493 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
1494 // CHECK9: omp_offload.cont:
1495 // CHECK9-NEXT: [[TMP41:%.*]] = load i32, i32* [[N]], align 4
1496 // CHECK9-NEXT: [[CONV4:%.*]] = bitcast i64* [[N_CASTED3]] to i32*
1497 // CHECK9-NEXT: store i32 [[TMP41]], i32* [[CONV4]], align 4
1498 // CHECK9-NEXT: [[TMP42:%.*]] = load i64, i64* [[N_CASTED3]], align 8
1499 // CHECK9-NEXT: [[TMP43:%.*]] = mul nuw i64 [[TMP1]], 4
1500 // CHECK9-NEXT: [[TMP44:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES8]] to i8*
1501 // CHECK9-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP44]], i8* align 8 bitcast ([3 x i64]* @.offload_sizes.4 to i8*), i64 24, i1 false)
1502 // CHECK9-NEXT: [[TMP45:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
1503 // CHECK9-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to i64*
1504 // CHECK9-NEXT: store i64 [[TMP42]], i64* [[TMP46]], align 8
1505 // CHECK9-NEXT: [[TMP47:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
1506 // CHECK9-NEXT: [[TMP48:%.*]] = bitcast i8** [[TMP47]] to i64*
1507 // CHECK9-NEXT: store i64 [[TMP42]], i64* [[TMP48]], align 8
1508 // CHECK9-NEXT: [[TMP49:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 0
1509 // CHECK9-NEXT: store i8* null, i8** [[TMP49]], align 8
1510 // CHECK9-NEXT: [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 1
1511 // CHECK9-NEXT: [[TMP51:%.*]] = bitcast i8** [[TMP50]] to i64*
1512 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP51]], align 8
1513 // CHECK9-NEXT: [[TMP52:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 1
1514 // CHECK9-NEXT: [[TMP53:%.*]] = bitcast i8** [[TMP52]] to i64*
1515 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP53]], align 8
1516 // CHECK9-NEXT: [[TMP54:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 1
1517 // CHECK9-NEXT: store i8* null, i8** [[TMP54]], align 8
1518 // CHECK9-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 2
1519 // CHECK9-NEXT: [[TMP56:%.*]] = bitcast i8** [[TMP55]] to i32**
1520 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP56]], align 8
1521 // CHECK9-NEXT: [[TMP57:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 2
1522 // CHECK9-NEXT: [[TMP58:%.*]] = bitcast i8** [[TMP57]] to i32**
1523 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP58]], align 8
1524 // CHECK9-NEXT: [[TMP59:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 2
1525 // CHECK9-NEXT: store i64 [[TMP43]], i64* [[TMP59]], align 8
1526 // CHECK9-NEXT: [[TMP60:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS7]], i64 0, i64 2
1527 // CHECK9-NEXT: store i8* null, i8** [[TMP60]], align 8
1528 // CHECK9-NEXT: [[TMP61:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS5]], i32 0, i32 0
1529 // CHECK9-NEXT: [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS6]], i32 0, i32 0
1530 // CHECK9-NEXT: [[TMP63:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES8]], i32 0, i32 0
1531 // CHECK9-NEXT: [[TMP64:%.*]] = load i32, i32* [[N]], align 4
1532 // CHECK9-NEXT: store i32 [[TMP64]], i32* [[DOTCAPTURE_EXPR_10]], align 4
1533 // CHECK9-NEXT: [[TMP65:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
1534 // CHECK9-NEXT: [[SUB12:%.*]] = sub nsw i32 [[TMP65]], 0
1535 // CHECK9-NEXT: [[DIV13:%.*]] = sdiv i32 [[SUB12]], 1
1536 // CHECK9-NEXT: [[SUB14:%.*]] = sub nsw i32 [[DIV13]], 1
1537 // CHECK9-NEXT: store i32 [[SUB14]], i32* [[DOTCAPTURE_EXPR_11]], align 4
1538 // CHECK9-NEXT: [[TMP66:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_11]], align 4
1539 // CHECK9-NEXT: [[ADD15:%.*]] = add nsw i32 [[TMP66]], 1
1540 // CHECK9-NEXT: [[TMP67:%.*]] = zext i32 [[ADD15]] to i64
1541 // CHECK9-NEXT: [[KERNEL_ARGS16:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1542 // CHECK9-NEXT: [[TMP68:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 0
1543 // CHECK9-NEXT: store i32 1, i32* [[TMP68]], align 4
1544 // CHECK9-NEXT: [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 1
1545 // CHECK9-NEXT: store i32 3, i32* [[TMP69]], align 4
1546 // CHECK9-NEXT: [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 2
1547 // CHECK9-NEXT: store i8** [[TMP61]], i8*** [[TMP70]], align 8
1548 // CHECK9-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 3
1549 // CHECK9-NEXT: store i8** [[TMP62]], i8*** [[TMP71]], align 8
1550 // CHECK9-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 4
1551 // CHECK9-NEXT: store i64* [[TMP63]], i64** [[TMP72]], align 8
1552 // CHECK9-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 5
1553 // CHECK9-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP73]], align 8
1554 // CHECK9-NEXT: [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 6
1555 // CHECK9-NEXT: store i8** null, i8*** [[TMP74]], align 8
1556 // CHECK9-NEXT: [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 7
1557 // CHECK9-NEXT: store i8** null, i8*** [[TMP75]], align 8
1558 // CHECK9-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]], i32 0, i32 8
1559 // CHECK9-NEXT: store i64 [[TMP67]], i64* [[TMP76]], align 8
1560 // CHECK9-NEXT: [[TMP77:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l111.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS16]])
1561 // CHECK9-NEXT: [[TMP78:%.*]] = icmp ne i32 [[TMP77]], 0
1562 // CHECK9-NEXT: br i1 [[TMP78]], label [[OMP_OFFLOAD_FAILED17:%.*]], label [[OMP_OFFLOAD_CONT18:%.*]]
1563 // CHECK9: omp_offload.failed17:
1564 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l111(i64 [[TMP42]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
1565 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT18]]
1566 // CHECK9: omp_offload.cont18:
1567 // CHECK9-NEXT: [[TMP79:%.*]] = load i32, i32* [[M]], align 4
1568 // CHECK9-NEXT: [[CONV19:%.*]] = bitcast i64* [[M_CASTED]] to i32*
1569 // CHECK9-NEXT: store i32 [[TMP79]], i32* [[CONV19]], align 4
1570 // CHECK9-NEXT: [[TMP80:%.*]] = load i64, i64* [[M_CASTED]], align 8
1571 // CHECK9-NEXT: [[TMP81:%.*]] = load i32, i32* [[N]], align 4
1572 // CHECK9-NEXT: [[CONV21:%.*]] = bitcast i64* [[N_CASTED20]] to i32*
1573 // CHECK9-NEXT: store i32 [[TMP81]], i32* [[CONV21]], align 4
1574 // CHECK9-NEXT: [[TMP82:%.*]] = load i64, i64* [[N_CASTED20]], align 8
1575 // CHECK9-NEXT: [[TMP83:%.*]] = mul nuw i64 [[TMP1]], 4
1576 // CHECK9-NEXT: [[TMP84:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES25]] to i8*
1577 // CHECK9-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP84]], i8* align 8 bitcast ([4 x i64]* @.offload_sizes.8 to i8*), i64 32, i1 false)
1578 // CHECK9-NEXT: [[TMP85:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
1579 // CHECK9-NEXT: [[TMP86:%.*]] = bitcast i8** [[TMP85]] to i64*
1580 // CHECK9-NEXT: store i64 [[TMP80]], i64* [[TMP86]], align 8
1581 // CHECK9-NEXT: [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
1582 // CHECK9-NEXT: [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i64*
1583 // CHECK9-NEXT: store i64 [[TMP80]], i64* [[TMP88]], align 8
1584 // CHECK9-NEXT: [[TMP89:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 0
1585 // CHECK9-NEXT: store i8* null, i8** [[TMP89]], align 8
1586 // CHECK9-NEXT: [[TMP90:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 1
1587 // CHECK9-NEXT: [[TMP91:%.*]] = bitcast i8** [[TMP90]] to i64*
1588 // CHECK9-NEXT: store i64 [[TMP82]], i64* [[TMP91]], align 8
1589 // CHECK9-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 1
1590 // CHECK9-NEXT: [[TMP93:%.*]] = bitcast i8** [[TMP92]] to i64*
1591 // CHECK9-NEXT: store i64 [[TMP82]], i64* [[TMP93]], align 8
1592 // CHECK9-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 1
1593 // CHECK9-NEXT: store i8* null, i8** [[TMP94]], align 8
1594 // CHECK9-NEXT: [[TMP95:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 2
1595 // CHECK9-NEXT: [[TMP96:%.*]] = bitcast i8** [[TMP95]] to i64*
1596 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP96]], align 8
1597 // CHECK9-NEXT: [[TMP97:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 2
1598 // CHECK9-NEXT: [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i64*
1599 // CHECK9-NEXT: store i64 [[TMP1]], i64* [[TMP98]], align 8
1600 // CHECK9-NEXT: [[TMP99:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 2
1601 // CHECK9-NEXT: store i8* null, i8** [[TMP99]], align 8
1602 // CHECK9-NEXT: [[TMP100:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 3
1603 // CHECK9-NEXT: [[TMP101:%.*]] = bitcast i8** [[TMP100]] to i32**
1604 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP101]], align 8
1605 // CHECK9-NEXT: [[TMP102:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 3
1606 // CHECK9-NEXT: [[TMP103:%.*]] = bitcast i8** [[TMP102]] to i32**
1607 // CHECK9-NEXT: store i32* [[VLA]], i32** [[TMP103]], align 8
1608 // CHECK9-NEXT: [[TMP104:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES25]], i32 0, i32 3
1609 // CHECK9-NEXT: store i64 [[TMP83]], i64* [[TMP104]], align 8
1610 // CHECK9-NEXT: [[TMP105:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS24]], i64 0, i64 3
1611 // CHECK9-NEXT: store i8* null, i8** [[TMP105]], align 8
1612 // CHECK9-NEXT: [[TMP106:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS22]], i32 0, i32 0
1613 // CHECK9-NEXT: [[TMP107:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS23]], i32 0, i32 0
1614 // CHECK9-NEXT: [[TMP108:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES25]], i32 0, i32 0
1615 // CHECK9-NEXT: [[TMP109:%.*]] = load i32, i32* [[N]], align 4
1616 // CHECK9-NEXT: store i32 [[TMP109]], i32* [[DOTCAPTURE_EXPR_27]], align 4
1617 // CHECK9-NEXT: [[TMP110:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_27]], align 4
1618 // CHECK9-NEXT: [[SUB29:%.*]] = sub nsw i32 [[TMP110]], 0
1619 // CHECK9-NEXT: [[DIV30:%.*]] = sdiv i32 [[SUB29]], 1
1620 // CHECK9-NEXT: [[SUB31:%.*]] = sub nsw i32 [[DIV30]], 1
1621 // CHECK9-NEXT: store i32 [[SUB31]], i32* [[DOTCAPTURE_EXPR_28]], align 4
1622 // CHECK9-NEXT: [[TMP111:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_28]], align 4
1623 // CHECK9-NEXT: [[ADD32:%.*]] = add nsw i32 [[TMP111]], 1
1624 // CHECK9-NEXT: [[TMP112:%.*]] = zext i32 [[ADD32]] to i64
1625 // CHECK9-NEXT: [[KERNEL_ARGS33:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
1626 // CHECK9-NEXT: [[TMP113:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 0
1627 // CHECK9-NEXT: store i32 1, i32* [[TMP113]], align 4
1628 // CHECK9-NEXT: [[TMP114:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 1
1629 // CHECK9-NEXT: store i32 4, i32* [[TMP114]], align 4
1630 // CHECK9-NEXT: [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 2
1631 // CHECK9-NEXT: store i8** [[TMP106]], i8*** [[TMP115]], align 8
1632 // CHECK9-NEXT: [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 3
1633 // CHECK9-NEXT: store i8** [[TMP107]], i8*** [[TMP116]], align 8
1634 // CHECK9-NEXT: [[TMP117:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 4
1635 // CHECK9-NEXT: store i64* [[TMP108]], i64** [[TMP117]], align 8
1636 // CHECK9-NEXT: [[TMP118:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 5
1637 // CHECK9-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP118]], align 8
1638 // CHECK9-NEXT: [[TMP119:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 6
1639 // CHECK9-NEXT: store i8** null, i8*** [[TMP119]], align 8
1640 // CHECK9-NEXT: [[TMP120:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 7
1641 // CHECK9-NEXT: store i8** null, i8*** [[TMP120]], align 8
1642 // CHECK9-NEXT: [[TMP121:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]], i32 0, i32 8
1643 // CHECK9-NEXT: store i64 [[TMP112]], i64* [[TMP121]], align 8
1644 // CHECK9-NEXT: [[TMP122:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS33]])
1645 // CHECK9-NEXT: [[TMP123:%.*]] = icmp ne i32 [[TMP122]], 0
1646 // CHECK9-NEXT: br i1 [[TMP123]], label [[OMP_OFFLOAD_FAILED34:%.*]], label [[OMP_OFFLOAD_CONT35:%.*]]
1647 // CHECK9: omp_offload.failed34:
1648 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116(i64 [[TMP80]], i64 [[TMP82]], i64 [[TMP1]], i32* [[VLA]]) #[[ATTR3]]
1649 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT35]]
1650 // CHECK9: omp_offload.cont35:
1651 // CHECK9-NEXT: [[TMP124:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
1652 // CHECK9-NEXT: [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiLi10EEiT_(i32 noundef signext [[TMP124]])
1653 // CHECK9-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
1654 // CHECK9-NEXT: [[TMP125:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1655 // CHECK9-NEXT: call void @llvm.stackrestore(i8* [[TMP125]])
1656 // CHECK9-NEXT: [[TMP126:%.*]] = load i32, i32* [[RETVAL]], align 4
1657 // CHECK9-NEXT: ret i32 [[TMP126]]
1658 //
1659 //
1660 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l106
1661 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
1662 // CHECK9-NEXT: entry:
1663 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
1664 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1665 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1666 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
1667 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1668 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1669 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
1670 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1671 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1672 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
1673 // CHECK9-NEXT: ret void
1674 //
1675 //
1676 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
1677 // 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]] {
1678 // CHECK9-NEXT: entry:
1679 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1680 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1681 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
1682 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1683 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1684 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1685 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1686 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1687 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1688 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1689 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1690 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1691 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1692 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1693 // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4
1694 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1695 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1696 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
1697 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1698 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1699 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
1700 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1701 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1702 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1703 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1704 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1705 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1706 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1707 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1708 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1709 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
1710 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1711 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1712 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1713 // CHECK9: omp.precond.then:
1714 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1715 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1716 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
1717 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1718 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1719 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1720 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
1721 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1722 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1723 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1724 // CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1725 // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1726 // CHECK9: cond.true:
1727 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1728 // CHECK9-NEXT: br label [[COND_END:%.*]]
1729 // CHECK9: cond.false:
1730 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1731 // CHECK9-NEXT: br label [[COND_END]]
1732 // CHECK9: cond.end:
1733 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1734 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1735 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1736 // CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1737 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1738 // CHECK9: omp.inner.for.cond:
1739 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1740 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1741 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1742 // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1743 // CHECK9: omp.inner.for.body:
1744 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1745 // CHECK9-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
1746 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1747 // CHECK9-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
1748 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
1749 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1750 // CHECK9: omp.inner.for.inc:
1751 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1752 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1753 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
1754 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1755 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
1756 // CHECK9: omp.inner.for.end:
1757 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1758 // CHECK9: omp.loop.exit:
1759 // CHECK9-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1760 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
1761 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
1762 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
1763 // CHECK9: omp.precond.end:
1764 // CHECK9-NEXT: ret void
1765 //
1766 //
1767 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
1768 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1769 // CHECK9-NEXT: entry:
1770 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1771 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1772 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1773 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1774 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
1775 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1776 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1777 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1778 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1779 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1780 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1781 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1782 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1783 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1784 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1785 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1786 // CHECK9-NEXT: [[I4:%.*]] = alloca i32, align 4
1787 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1788 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1789 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1790 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1791 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
1792 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1793 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1794 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
1795 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1796 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1797 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1798 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1799 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1800 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1801 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1802 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1803 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1804 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
1805 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1806 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1807 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1808 // CHECK9: omp.precond.then:
1809 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
1810 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1811 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
1812 // CHECK9-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1813 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
1814 // CHECK9-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1815 // CHECK9-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
1816 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1817 // CHECK9-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
1818 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1819 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1820 // CHECK9-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1821 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
1822 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1823 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1824 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1825 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
1826 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1827 // CHECK9: cond.true:
1828 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1829 // CHECK9-NEXT: br label [[COND_END:%.*]]
1830 // CHECK9: cond.false:
1831 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1832 // CHECK9-NEXT: br label [[COND_END]]
1833 // CHECK9: cond.end:
1834 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
1835 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1836 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1837 // CHECK9-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
1838 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1839 // CHECK9: omp.inner.for.cond:
1840 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1841 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1842 // CHECK9-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
1843 // CHECK9-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1844 // CHECK9: omp.inner.for.body:
1845 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1846 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
1847 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1848 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
1849 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
1850 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
1851 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
1852 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
1853 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
1854 // CHECK9: omp.body.continue:
1855 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1856 // CHECK9: omp.inner.for.inc:
1857 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1858 // CHECK9-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
1859 // CHECK9-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
1860 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
1861 // CHECK9: omp.inner.for.end:
1862 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1863 // CHECK9: omp.loop.exit:
1864 // CHECK9-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1865 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
1866 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
1867 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
1868 // CHECK9: omp.precond.end:
1869 // CHECK9-NEXT: ret void
1870 //
1871 //
1872 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l111
1873 // CHECK9-SAME: (i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1874 // CHECK9-NEXT: entry:
1875 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
1876 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1877 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1878 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
1879 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1880 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1881 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[N_ADDR]] to i32*
1882 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1883 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1884 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[CONV]], i64 [[TMP0]], i32* [[TMP1]])
1885 // CHECK9-NEXT: ret void
1886 //
1887 //
1888 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
1889 // 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]] {
1890 // CHECK9-NEXT: entry:
1891 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1892 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1893 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
1894 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1895 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1896 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1897 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1898 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1899 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1900 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1901 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1902 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1903 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1904 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1905 // CHECK9-NEXT: [[I3:%.*]] = alloca i32, align 4
1906 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1907 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1908 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
1909 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
1910 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
1911 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
1912 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
1913 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
1914 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
1915 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
1916 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1917 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
1918 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
1919 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
1920 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
1921 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
1922 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
1923 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
1924 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1925 // CHECK9: omp.precond.then:
1926 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1927 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1928 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
1929 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1930 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1931 // CHECK9-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1932 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
1933 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1934 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1935 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1936 // CHECK9-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
1937 // CHECK9-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1938 // CHECK9: cond.true:
1939 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
1940 // CHECK9-NEXT: br label [[COND_END:%.*]]
1941 // CHECK9: cond.false:
1942 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1943 // CHECK9-NEXT: br label [[COND_END]]
1944 // CHECK9: cond.end:
1945 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1946 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1947 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1948 // CHECK9-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1949 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
1950 // CHECK9: omp.inner.for.cond:
1951 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1952 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1953 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1954 // CHECK9-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1955 // CHECK9: omp.inner.for.body:
1956 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1957 // CHECK9-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
1958 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1959 // CHECK9-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
1960 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP17]], i64 [[TMP19]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]])
1961 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
1962 // CHECK9: omp.inner.for.inc:
1963 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1964 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1965 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
1966 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1967 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
1968 // CHECK9: omp.inner.for.end:
1969 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
1970 // CHECK9: omp.loop.exit:
1971 // CHECK9-NEXT: [[TMP22:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1972 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[TMP22]], align 4
1973 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP23]])
1974 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
1975 // CHECK9: omp.precond.end:
1976 // CHECK9-NEXT: ret void
1977 //
1978 //
1979 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
1980 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
1981 // CHECK9-NEXT: entry:
1982 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1983 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1984 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1985 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1986 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
1987 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
1988 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
1989 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
1990 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
1991 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
1992 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
1993 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
1994 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
1995 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
1996 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1997 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1998 // CHECK9-NEXT: [[I4:%.*]] = alloca i32, align 4
1999 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2000 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2001 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2002 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2003 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
2004 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2005 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2006 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
2007 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2008 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2009 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2010 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
2011 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2012 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2013 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2014 // CHECK9-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
2015 // CHECK9-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2016 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
2017 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2018 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2019 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2020 // CHECK9: omp.precond.then:
2021 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2022 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2023 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
2024 // CHECK9-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2025 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
2026 // CHECK9-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2027 // CHECK9-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
2028 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2029 // CHECK9-NEXT: store i32 [[CONV3]], i32* [[DOTOMP_UB]], align 4
2030 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2031 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2032 // CHECK9-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2033 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
2034 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2035 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2036 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2037 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
2038 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2039 // CHECK9: cond.true:
2040 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2041 // CHECK9-NEXT: br label [[COND_END:%.*]]
2042 // CHECK9: cond.false:
2043 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2044 // CHECK9-NEXT: br label [[COND_END]]
2045 // CHECK9: cond.end:
2046 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
2047 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2048 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2049 // CHECK9-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
2050 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2051 // CHECK9: omp.inner.for.cond:
2052 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2053 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2054 // CHECK9-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
2055 // CHECK9-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2056 // CHECK9: omp.inner.for.body:
2057 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2058 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
2059 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2060 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
2061 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
2062 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
2063 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
2064 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2065 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2066 // CHECK9: omp.body.continue:
2067 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2068 // CHECK9: omp.inner.for.inc:
2069 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2070 // CHECK9-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
2071 // CHECK9-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
2072 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
2073 // CHECK9: omp.inner.for.end:
2074 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2075 // CHECK9: omp.loop.exit:
2076 // CHECK9-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2077 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
2078 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
2079 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
2080 // CHECK9: omp.precond.end:
2081 // CHECK9-NEXT: ret void
2082 //
2083 //
2084 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116
2085 // CHECK9-SAME: (i64 noundef [[M:%.*]], i64 noundef [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
2086 // CHECK9-NEXT: entry:
2087 // CHECK9-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
2088 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8
2089 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
2090 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
2091 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2092 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
2093 // CHECK9-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
2094 // CHECK9-NEXT: store i64 [[N]], i64* [[N_ADDR]], align 8
2095 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2096 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2097 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
2098 // CHECK9-NEXT: [[CONV1:%.*]] = bitcast i64* [[N_ADDR]] to i32*
2099 // CHECK9-NEXT: [[TMP0:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2100 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2101 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[CONV]], align 4
2102 // CHECK9-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
2103 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2104 // CHECK9-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
2105 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[CONV2]], align 4
2106 // CHECK9-NEXT: [[TMP4:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
2107 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i64, i32*, i64)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32* [[CONV1]], i64 [[TMP0]], i32* [[TMP1]], i64 [[TMP4]])
2108 // CHECK9-NEXT: ret void
2109 //
2110 //
2111 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..6
2112 // 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]] {
2113 // CHECK9-NEXT: entry:
2114 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2115 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2116 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
2117 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
2118 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
2119 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2120 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2121 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2122 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2123 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
2124 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2125 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2126 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2127 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2128 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2129 // CHECK9-NEXT: [[I4:%.*]] = alloca i32, align 4
2130 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
2131 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2132 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2133 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
2134 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2135 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2136 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2137 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
2138 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2139 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2140 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2141 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2142 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2143 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2144 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2145 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2146 // CHECK9-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
2147 // CHECK9-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
2148 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
2149 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2150 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2151 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2152 // CHECK9: omp.precond.then:
2153 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2154 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2155 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
2156 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2157 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2158 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[CONV]], align 4
2159 // CHECK9-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2160 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
2161 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
2162 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2163 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2164 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
2165 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2166 // CHECK9: cond.true:
2167 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2168 // CHECK9-NEXT: br label [[COND_END:%.*]]
2169 // CHECK9: cond.false:
2170 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2171 // CHECK9-NEXT: br label [[COND_END]]
2172 // CHECK9: cond.end:
2173 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
2174 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2175 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2176 // CHECK9-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
2177 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2178 // CHECK9: omp.inner.for.cond:
2179 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2180 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2181 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
2182 // CHECK9-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
2183 // CHECK9-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2184 // CHECK9: omp.inner.for.body:
2185 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2186 // CHECK9-NEXT: [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
2187 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2188 // CHECK9-NEXT: [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
2189 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[CONV]], align 4
2190 // CHECK9-NEXT: [[CONV7:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
2191 // CHECK9-NEXT: store i32 [[TMP21]], i32* [[CONV7]], align 4
2192 // CHECK9-NEXT: [[TMP22:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
2193 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, i32*, i64, i32*, i64)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i64 [[TMP18]], i64 [[TMP20]], i32* [[TMP0]], i64 [[TMP1]], i32* [[TMP2]], i64 [[TMP22]])
2194 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2195 // CHECK9: omp.inner.for.inc:
2196 // CHECK9-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2197 // CHECK9-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2198 // CHECK9-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
2199 // CHECK9-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_IV]], align 4
2200 // CHECK9-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2201 // CHECK9-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2202 // CHECK9-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
2203 // CHECK9-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_LB]], align 4
2204 // CHECK9-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2205 // CHECK9-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2206 // CHECK9-NEXT: [[ADD10:%.*]] = add nsw i32 [[TMP27]], [[TMP28]]
2207 // CHECK9-NEXT: store i32 [[ADD10]], i32* [[DOTOMP_COMB_UB]], align 4
2208 // CHECK9-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2209 // CHECK9-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2210 // CHECK9-NEXT: [[CMP11:%.*]] = icmp sgt i32 [[TMP29]], [[TMP30]]
2211 // CHECK9-NEXT: br i1 [[CMP11]], label [[COND_TRUE12:%.*]], label [[COND_FALSE13:%.*]]
2212 // CHECK9: cond.true12:
2213 // CHECK9-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2214 // CHECK9-NEXT: br label [[COND_END14:%.*]]
2215 // CHECK9: cond.false13:
2216 // CHECK9-NEXT: [[TMP32:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2217 // CHECK9-NEXT: br label [[COND_END14]]
2218 // CHECK9: cond.end14:
2219 // CHECK9-NEXT: [[COND15:%.*]] = phi i32 [ [[TMP31]], [[COND_TRUE12]] ], [ [[TMP32]], [[COND_FALSE13]] ]
2220 // CHECK9-NEXT: store i32 [[COND15]], i32* [[DOTOMP_COMB_UB]], align 4
2221 // CHECK9-NEXT: [[TMP33:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2222 // CHECK9-NEXT: store i32 [[TMP33]], i32* [[DOTOMP_IV]], align 4
2223 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
2224 // CHECK9: omp.inner.for.end:
2225 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2226 // CHECK9: omp.loop.exit:
2227 // CHECK9-NEXT: [[TMP34:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2228 // CHECK9-NEXT: [[TMP35:%.*]] = load i32, i32* [[TMP34]], align 4
2229 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP35]])
2230 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
2231 // CHECK9: omp.precond.end:
2232 // CHECK9-NEXT: ret void
2233 //
2234 //
2235 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..7
2236 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i64 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
2237 // CHECK9-NEXT: entry:
2238 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2239 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2240 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2241 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2242 // CHECK9-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 8
2243 // CHECK9-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8
2244 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 8
2245 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2246 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2247 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2248 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
2249 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
2250 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2251 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2252 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2253 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2254 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2255 // CHECK9-NEXT: [[I6:%.*]] = alloca i32, align 4
2256 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2257 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2258 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2259 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2260 // CHECK9-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 8
2261 // CHECK9-NEXT: store i64 [[VLA]], i64* [[VLA_ADDR]], align 8
2262 // CHECK9-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 8
2263 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2264 // CHECK9-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 8
2265 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[VLA_ADDR]], align 8
2266 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 8
2267 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2268 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
2269 // CHECK9-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
2270 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2271 // CHECK9-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
2272 // CHECK9-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
2273 // CHECK9-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
2274 // CHECK9-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
2275 // CHECK9-NEXT: store i32 0, i32* [[I]], align 4
2276 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
2277 // CHECK9-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
2278 // CHECK9-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2279 // CHECK9: omp.precond.then:
2280 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2281 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2282 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
2283 // CHECK9-NEXT: [[TMP7:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2284 // CHECK9-NEXT: [[CONV4:%.*]] = trunc i64 [[TMP7]] to i32
2285 // CHECK9-NEXT: [[TMP8:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2286 // CHECK9-NEXT: [[CONV5:%.*]] = trunc i64 [[TMP8]] to i32
2287 // CHECK9-NEXT: store i32 [[CONV4]], i32* [[DOTOMP_LB]], align 4
2288 // CHECK9-NEXT: store i32 [[CONV5]], i32* [[DOTOMP_UB]], align 4
2289 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2290 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2291 // CHECK9-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2292 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
2293 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2294 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2295 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2296 // CHECK9-NEXT: [[CMP7:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
2297 // CHECK9-NEXT: br i1 [[CMP7]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2298 // CHECK9: cond.true:
2299 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
2300 // CHECK9-NEXT: br label [[COND_END:%.*]]
2301 // CHECK9: cond.false:
2302 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2303 // CHECK9-NEXT: br label [[COND_END]]
2304 // CHECK9: cond.end:
2305 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
2306 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2307 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2308 // CHECK9-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
2309 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2310 // CHECK9: omp.inner.for.cond:
2311 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2312 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2313 // CHECK9-NEXT: [[CMP8:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
2314 // CHECK9-NEXT: br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2315 // CHECK9: omp.inner.for.body:
2316 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2317 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
2318 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2319 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I6]], align 4
2320 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[I6]], align 4
2321 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP19]] to i64
2322 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i64 [[IDXPROM]]
2323 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2324 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2325 // CHECK9: omp.body.continue:
2326 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2327 // CHECK9: omp.inner.for.inc:
2328 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2329 // CHECK9-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP20]], 1
2330 // CHECK9-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
2331 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
2332 // CHECK9: omp.inner.for.end:
2333 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2334 // CHECK9: omp.loop.exit:
2335 // CHECK9-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2336 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
2337 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
2338 // CHECK9-NEXT: br label [[OMP_PRECOND_END]]
2339 // CHECK9: omp.precond.end:
2340 // CHECK9-NEXT: ret void
2341 //
2342 //
2343 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
2344 // CHECK9-SAME: (i32 noundef signext [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
2345 // CHECK9-NEXT: entry:
2346 // CHECK9-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
2347 // CHECK9-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
2348 // CHECK9-NEXT: [[M:%.*]] = alloca i32, align 4
2349 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
2350 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
2351 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
2352 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2353 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 8
2354 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 8
2355 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 8
2356 // CHECK9-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
2357 // CHECK9-NEXT: [[M_CASTED:%.*]] = alloca i64, align 8
2358 // CHECK9-NEXT: [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x i8*], align 8
2359 // CHECK9-NEXT: [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 8
2360 // CHECK9-NEXT: [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 8
2361 // CHECK9-NEXT: [[_TMP11:%.*]] = alloca i32, align 4
2362 // CHECK9-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
2363 // CHECK9-NEXT: store i32 10, i32* [[M]], align 4
2364 // CHECK9-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2365 // CHECK9-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
2366 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 8
2367 // CHECK9-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2368 // CHECK9-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
2369 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 8
2370 // CHECK9-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2371 // CHECK9-NEXT: store i8* null, i8** [[TMP4]], align 8
2372 // CHECK9-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2373 // CHECK9-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2374 // CHECK9-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2375 // CHECK9-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
2376 // CHECK9-NEXT: store i32 1, i32* [[TMP7]], align 4
2377 // CHECK9-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
2378 // CHECK9-NEXT: store i32 1, i32* [[TMP8]], align 4
2379 // CHECK9-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
2380 // CHECK9-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 8
2381 // CHECK9-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
2382 // CHECK9-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 8
2383 // CHECK9-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
2384 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP11]], align 8
2385 // CHECK9-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
2386 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP12]], align 8
2387 // CHECK9-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
2388 // CHECK9-NEXT: store i8** null, i8*** [[TMP13]], align 8
2389 // CHECK9-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
2390 // CHECK9-NEXT: store i8** null, i8*** [[TMP14]], align 8
2391 // CHECK9-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
2392 // CHECK9-NEXT: store i64 10, i64* [[TMP15]], align 8
2393 // CHECK9-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
2394 // CHECK9-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
2395 // CHECK9-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2396 // CHECK9: omp_offload.failed:
2397 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84([10 x i32]* [[A]]) #[[ATTR3]]
2398 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT]]
2399 // CHECK9: omp_offload.cont:
2400 // CHECK9-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
2401 // CHECK9-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to [10 x i32]**
2402 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP19]], align 8
2403 // CHECK9-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
2404 // CHECK9-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [10 x i32]**
2405 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP21]], align 8
2406 // CHECK9-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i64 0, i64 0
2407 // CHECK9-NEXT: store i8* null, i8** [[TMP22]], align 8
2408 // CHECK9-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
2409 // CHECK9-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
2410 // CHECK9-NEXT: [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2411 // CHECK9-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 0
2412 // CHECK9-NEXT: store i32 1, i32* [[TMP25]], align 4
2413 // CHECK9-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 1
2414 // CHECK9-NEXT: store i32 1, i32* [[TMP26]], align 4
2415 // CHECK9-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 2
2416 // CHECK9-NEXT: store i8** [[TMP23]], i8*** [[TMP27]], align 8
2417 // CHECK9-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 3
2418 // CHECK9-NEXT: store i8** [[TMP24]], i8*** [[TMP28]], align 8
2419 // CHECK9-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 4
2420 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64** [[TMP29]], align 8
2421 // CHECK9-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 5
2422 // CHECK9-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP30]], align 8
2423 // CHECK9-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 6
2424 // CHECK9-NEXT: store i8** null, i8*** [[TMP31]], align 8
2425 // CHECK9-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 7
2426 // CHECK9-NEXT: store i8** null, i8*** [[TMP32]], align 8
2427 // CHECK9-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 8
2428 // CHECK9-NEXT: store i64 10, i64* [[TMP33]], align 8
2429 // CHECK9-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]])
2430 // CHECK9-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
2431 // CHECK9-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
2432 // CHECK9: omp_offload.failed6:
2433 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89([10 x i32]* [[A]]) #[[ATTR3]]
2434 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT7]]
2435 // CHECK9: omp_offload.cont7:
2436 // CHECK9-NEXT: [[TMP36:%.*]] = load i32, i32* [[M]], align 4
2437 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[M_CASTED]] to i32*
2438 // CHECK9-NEXT: store i32 [[TMP36]], i32* [[CONV]], align 4
2439 // CHECK9-NEXT: [[TMP37:%.*]] = load i64, i64* [[M_CASTED]], align 8
2440 // CHECK9-NEXT: [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
2441 // CHECK9-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i64*
2442 // CHECK9-NEXT: store i64 [[TMP37]], i64* [[TMP39]], align 8
2443 // CHECK9-NEXT: [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
2444 // CHECK9-NEXT: [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i64*
2445 // CHECK9-NEXT: store i64 [[TMP37]], i64* [[TMP41]], align 8
2446 // CHECK9-NEXT: [[TMP42:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 0
2447 // CHECK9-NEXT: store i8* null, i8** [[TMP42]], align 8
2448 // CHECK9-NEXT: [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
2449 // CHECK9-NEXT: [[TMP44:%.*]] = bitcast i8** [[TMP43]] to [10 x i32]**
2450 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP44]], align 8
2451 // CHECK9-NEXT: [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
2452 // CHECK9-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to [10 x i32]**
2453 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP46]], align 8
2454 // CHECK9-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i64 0, i64 1
2455 // CHECK9-NEXT: store i8* null, i8** [[TMP47]], align 8
2456 // CHECK9-NEXT: [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
2457 // CHECK9-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
2458 // CHECK9-NEXT: [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
2459 // CHECK9-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 0
2460 // CHECK9-NEXT: store i32 1, i32* [[TMP50]], align 4
2461 // CHECK9-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 1
2462 // CHECK9-NEXT: store i32 2, i32* [[TMP51]], align 4
2463 // CHECK9-NEXT: [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 2
2464 // CHECK9-NEXT: store i8** [[TMP48]], i8*** [[TMP52]], align 8
2465 // CHECK9-NEXT: [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 3
2466 // CHECK9-NEXT: store i8** [[TMP49]], i8*** [[TMP53]], align 8
2467 // CHECK9-NEXT: [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 4
2468 // CHECK9-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.20, i32 0, i32 0), i64** [[TMP54]], align 8
2469 // CHECK9-NEXT: [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 5
2470 // CHECK9-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.21, i32 0, i32 0), i64** [[TMP55]], align 8
2471 // CHECK9-NEXT: [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 6
2472 // CHECK9-NEXT: store i8** null, i8*** [[TMP56]], align 8
2473 // CHECK9-NEXT: [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 7
2474 // CHECK9-NEXT: store i8** null, i8*** [[TMP57]], align 8
2475 // CHECK9-NEXT: [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 8
2476 // CHECK9-NEXT: store i64 10, i64* [[TMP58]], align 8
2477 // CHECK9-NEXT: [[TMP59:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l94.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]])
2478 // CHECK9-NEXT: [[TMP60:%.*]] = icmp ne i32 [[TMP59]], 0
2479 // CHECK9-NEXT: br i1 [[TMP60]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
2480 // CHECK9: omp_offload.failed13:
2481 // CHECK9-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l94(i64 [[TMP37]], [10 x i32]* [[A]]) #[[ATTR3]]
2482 // CHECK9-NEXT: br label [[OMP_OFFLOAD_CONT14]]
2483 // CHECK9: omp_offload.cont14:
2484 // CHECK9-NEXT: ret i32 0
2485 //
2486 //
2487 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84
2488 // CHECK9-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2489 // CHECK9-NEXT: entry:
2490 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2491 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2492 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2493 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
2494 // CHECK9-NEXT: ret void
2495 //
2496 //
2497 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..10
2498 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2499 // CHECK9-NEXT: entry:
2500 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2501 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2502 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2503 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2504 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2505 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2506 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2507 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2508 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2509 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2510 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2511 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2512 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2513 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2514 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2515 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
2516 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2517 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2518 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2519 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2520 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2521 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2522 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2523 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2524 // CHECK9: cond.true:
2525 // CHECK9-NEXT: br label [[COND_END:%.*]]
2526 // CHECK9: cond.false:
2527 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2528 // CHECK9-NEXT: br label [[COND_END]]
2529 // CHECK9: cond.end:
2530 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2531 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2532 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2533 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2534 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2535 // CHECK9: omp.inner.for.cond:
2536 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2537 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2538 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2539 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2540 // CHECK9: omp.inner.for.body:
2541 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2542 // CHECK9-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
2543 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2544 // CHECK9-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
2545 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
2546 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2547 // CHECK9: omp.inner.for.inc:
2548 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2549 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2550 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2551 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2552 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
2553 // CHECK9: omp.inner.for.end:
2554 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2555 // CHECK9: omp.loop.exit:
2556 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2557 // CHECK9-NEXT: ret void
2558 //
2559 //
2560 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..11
2561 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2562 // CHECK9-NEXT: entry:
2563 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2564 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2565 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2566 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2567 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2568 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2569 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2570 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2571 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2572 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2573 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2574 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2575 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2576 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2577 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2578 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2579 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2580 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2581 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2582 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
2583 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2584 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
2585 // CHECK9-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2586 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
2587 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2588 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2589 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2590 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2591 // CHECK9-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2592 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2593 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2594 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2595 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
2596 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2597 // CHECK9: cond.true:
2598 // CHECK9-NEXT: br label [[COND_END:%.*]]
2599 // CHECK9: cond.false:
2600 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2601 // CHECK9-NEXT: br label [[COND_END]]
2602 // CHECK9: cond.end:
2603 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2604 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2605 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2606 // CHECK9-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
2607 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2608 // CHECK9: omp.inner.for.cond:
2609 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2610 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2611 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2612 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2613 // CHECK9: omp.inner.for.body:
2614 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2615 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2616 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2617 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2618 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
2619 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
2620 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
2621 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2622 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2623 // CHECK9: omp.body.continue:
2624 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2625 // CHECK9: omp.inner.for.inc:
2626 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2627 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
2628 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
2629 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
2630 // CHECK9: omp.inner.for.end:
2631 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2632 // CHECK9: omp.loop.exit:
2633 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
2634 // CHECK9-NEXT: ret void
2635 //
2636 //
2637 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89
2638 // CHECK9-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2639 // CHECK9-NEXT: entry:
2640 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2641 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2642 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2643 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
2644 // CHECK9-NEXT: ret void
2645 //
2646 //
2647 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..14
2648 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2649 // CHECK9-NEXT: entry:
2650 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2651 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2652 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2653 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2654 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2655 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2656 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2657 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2658 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2659 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2660 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2661 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2662 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2663 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2664 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2665 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
2666 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2667 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2668 // CHECK9-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2669 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2670 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2671 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2672 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
2673 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2674 // CHECK9: cond.true:
2675 // CHECK9-NEXT: br label [[COND_END:%.*]]
2676 // CHECK9: cond.false:
2677 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2678 // CHECK9-NEXT: br label [[COND_END]]
2679 // CHECK9: cond.end:
2680 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2681 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2682 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2683 // CHECK9-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2684 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2685 // CHECK9: omp.inner.for.cond:
2686 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2687 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2688 // CHECK9-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2689 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2690 // CHECK9: omp.inner.for.body:
2691 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2692 // CHECK9-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
2693 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2694 // CHECK9-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
2695 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]])
2696 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2697 // CHECK9: omp.inner.for.inc:
2698 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2699 // CHECK9-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2700 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2701 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2702 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
2703 // CHECK9: omp.inner.for.end:
2704 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2705 // CHECK9: omp.loop.exit:
2706 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2707 // CHECK9-NEXT: ret void
2708 //
2709 //
2710 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..15
2711 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2712 // CHECK9-NEXT: entry:
2713 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2714 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2715 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2716 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2717 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2718 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2719 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2720 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2721 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2722 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2723 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2724 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2725 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2726 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2727 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2728 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2729 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2730 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2731 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2732 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
2733 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2734 // CHECK9-NEXT: [[CONV:%.*]] = trunc i64 [[TMP1]] to i32
2735 // CHECK9-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2736 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP2]] to i32
2737 // CHECK9-NEXT: store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2738 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2739 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2740 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2741 // CHECK9-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2742 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2743 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2744 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2745 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
2746 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2747 // CHECK9: cond.true:
2748 // CHECK9-NEXT: br label [[COND_END:%.*]]
2749 // CHECK9: cond.false:
2750 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2751 // CHECK9-NEXT: br label [[COND_END]]
2752 // CHECK9: cond.end:
2753 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2754 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2755 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2756 // CHECK9-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
2757 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2758 // CHECK9: omp.inner.for.cond:
2759 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2760 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2761 // CHECK9-NEXT: [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2762 // CHECK9-NEXT: br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2763 // CHECK9: omp.inner.for.body:
2764 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2765 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2766 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2767 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2768 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
2769 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
2770 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
2771 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2772 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2773 // CHECK9: omp.body.continue:
2774 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2775 // CHECK9: omp.inner.for.inc:
2776 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2777 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP12]], 1
2778 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
2779 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
2780 // CHECK9: omp.inner.for.end:
2781 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2782 // CHECK9: omp.loop.exit:
2783 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
2784 // CHECK9-NEXT: ret void
2785 //
2786 //
2787 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l94
2788 // CHECK9-SAME: (i64 noundef [[M:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
2789 // CHECK9-NEXT: entry:
2790 // CHECK9-NEXT: [[M_ADDR:%.*]] = alloca i64, align 8
2791 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2792 // CHECK9-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
2793 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
2794 // CHECK9-NEXT: store i64 [[M]], i64* [[M_ADDR]], align 8
2795 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2796 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[M_ADDR]] to i32*
2797 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2798 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
2799 // CHECK9-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
2800 // CHECK9-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
2801 // CHECK9-NEXT: [[CONV1:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
2802 // CHECK9-NEXT: store i32 [[TMP2]], i32* [[CONV1]], align 4
2803 // CHECK9-NEXT: [[TMP3:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
2804 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i64)* @.omp_outlined..18 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i64 [[TMP3]])
2805 // CHECK9-NEXT: ret void
2806 //
2807 //
2808 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..18
2809 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
2810 // CHECK9-NEXT: entry:
2811 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2812 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2813 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2814 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2815 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2816 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2817 // CHECK9-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2818 // CHECK9-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2819 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2820 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2821 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2822 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
2823 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2824 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2825 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2826 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2827 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2828 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2829 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2830 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
2831 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2832 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2833 // CHECK9-NEXT: [[TMP1:%.*]] = load i32, i32* [[CONV]], align 4
2834 // CHECK9-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2835 // CHECK9-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2836 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP1]])
2837 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2838 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
2839 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2840 // CHECK9: cond.true:
2841 // CHECK9-NEXT: br label [[COND_END:%.*]]
2842 // CHECK9: cond.false:
2843 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2844 // CHECK9-NEXT: br label [[COND_END]]
2845 // CHECK9: cond.end:
2846 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2847 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2848 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2849 // CHECK9-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2850 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2851 // CHECK9: omp.inner.for.cond:
2852 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2853 // CHECK9-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP7]], 10
2854 // CHECK9-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2855 // CHECK9: omp.inner.for.body:
2856 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2857 // CHECK9-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
2858 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2859 // CHECK9-NEXT: [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
2860 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[CONV]], align 4
2861 // CHECK9-NEXT: [[CONV2:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i32*
2862 // CHECK9-NEXT: store i32 [[TMP12]], i32* [[CONV2]], align 4
2863 // CHECK9-NEXT: [[TMP13:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8
2864 // CHECK9-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [10 x i32]*, i64)* @.omp_outlined..19 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]], [10 x i32]* [[TMP0]], i64 [[TMP13]])
2865 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2866 // CHECK9: omp.inner.for.inc:
2867 // CHECK9-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2868 // CHECK9-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2869 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
2870 // CHECK9-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2871 // CHECK9-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2872 // CHECK9-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2873 // CHECK9-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
2874 // CHECK9-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_LB]], align 4
2875 // CHECK9-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2876 // CHECK9-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2877 // CHECK9-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
2878 // CHECK9-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_COMB_UB]], align 4
2879 // CHECK9-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2880 // CHECK9-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP20]], 9
2881 // CHECK9-NEXT: br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
2882 // CHECK9: cond.true6:
2883 // CHECK9-NEXT: br label [[COND_END8:%.*]]
2884 // CHECK9: cond.false7:
2885 // CHECK9-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2886 // CHECK9-NEXT: br label [[COND_END8]]
2887 // CHECK9: cond.end8:
2888 // CHECK9-NEXT: [[COND9:%.*]] = phi i32 [ 9, [[COND_TRUE6]] ], [ [[TMP21]], [[COND_FALSE7]] ]
2889 // CHECK9-NEXT: store i32 [[COND9]], i32* [[DOTOMP_COMB_UB]], align 4
2890 // CHECK9-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2891 // CHECK9-NEXT: store i32 [[TMP22]], i32* [[DOTOMP_IV]], align 4
2892 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
2893 // CHECK9: omp.inner.for.end:
2894 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2895 // CHECK9: omp.loop.exit:
2896 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2897 // CHECK9-NEXT: ret void
2898 //
2899 //
2900 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..19
2901 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
2902 // CHECK9-NEXT: entry:
2903 // CHECK9-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2904 // CHECK9-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2905 // CHECK9-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2906 // CHECK9-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2907 // CHECK9-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 8
2908 // CHECK9-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2909 // CHECK9-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
2910 // CHECK9-NEXT: [[TMP:%.*]] = alloca i32, align 4
2911 // CHECK9-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
2912 // CHECK9-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
2913 // CHECK9-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2914 // CHECK9-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2915 // CHECK9-NEXT: [[I:%.*]] = alloca i32, align 4
2916 // CHECK9-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2917 // CHECK9-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2918 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2919 // CHECK9-NEXT: store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2920 // CHECK9-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 8
2921 // CHECK9-NEXT: store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2922 // CHECK9-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 8
2923 // CHECK9-NEXT: [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i32*
2924 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
2925 // CHECK9-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
2926 // CHECK9-NEXT: [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2927 // CHECK9-NEXT: [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2928 // CHECK9-NEXT: [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2929 // CHECK9-NEXT: [[CONV2:%.*]] = trunc i64 [[TMP2]] to i32
2930 // CHECK9-NEXT: store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
2931 // CHECK9-NEXT: store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
2932 // CHECK9-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2933 // CHECK9-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2934 // CHECK9-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2935 // CHECK9-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2936 // CHECK9-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2937 // CHECK9-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2938 // CHECK9-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
2939 // CHECK9-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2940 // CHECK9: cond.true:
2941 // CHECK9-NEXT: br label [[COND_END:%.*]]
2942 // CHECK9: cond.false:
2943 // CHECK9-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2944 // CHECK9-NEXT: br label [[COND_END]]
2945 // CHECK9: cond.end:
2946 // CHECK9-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2947 // CHECK9-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2948 // CHECK9-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2949 // CHECK9-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
2950 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
2951 // CHECK9: omp.inner.for.cond:
2952 // CHECK9-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2953 // CHECK9-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2954 // CHECK9-NEXT: [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2955 // CHECK9-NEXT: br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2956 // CHECK9: omp.inner.for.body:
2957 // CHECK9-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2958 // CHECK9-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2959 // CHECK9-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2960 // CHECK9-NEXT: store i32 [[ADD]], i32* [[I]], align 4
2961 // CHECK9-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
2962 // CHECK9-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP11]] to i64
2963 // CHECK9-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
2964 // CHECK9-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
2965 // CHECK9-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
2966 // CHECK9: omp.body.continue:
2967 // CHECK9-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
2968 // CHECK9: omp.inner.for.inc:
2969 // CHECK9-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2970 // CHECK9-NEXT: [[ADD4:%.*]] = add nsw i32 [[TMP12]], 1
2971 // CHECK9-NEXT: store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4
2972 // CHECK9-NEXT: br label [[OMP_INNER_FOR_COND]]
2973 // CHECK9: omp.inner.for.end:
2974 // CHECK9-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
2975 // CHECK9: omp.loop.exit:
2976 // CHECK9-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
2977 // CHECK9-NEXT: ret void
2978 //
2979 //
2980 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2981 // CHECK9-SAME: () #[[ATTR6:[0-9]+]] {
2982 // CHECK9-NEXT: entry:
2983 // CHECK9-NEXT: call void @__tgt_register_requires(i64 1)
2984 // CHECK9-NEXT: ret void
2985 //
2986 //
2987 // CHECK11-LABEL: define {{[^@]+}}@main
2988 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]], i8** noundef [[ARGV:%.*]]) #[[ATTR0:[0-9]+]] {
2989 // CHECK11-NEXT: entry:
2990 // CHECK11-NEXT: [[RETVAL:%.*]] = alloca i32, align 4
2991 // CHECK11-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
2992 // CHECK11-NEXT: [[ARGV_ADDR:%.*]] = alloca i8**, align 4
2993 // CHECK11-NEXT: [[N:%.*]] = alloca i32, align 4
2994 // CHECK11-NEXT: [[SAVED_STACK:%.*]] = alloca i8*, align 4
2995 // CHECK11-NEXT: [[__VLA_EXPR0:%.*]] = alloca i32, align 4
2996 // CHECK11-NEXT: [[M:%.*]] = alloca i32, align 4
2997 // CHECK11-NEXT: [[N_CASTED:%.*]] = alloca i32, align 4
2998 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [3 x i8*], align 4
2999 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [3 x i8*], align 4
3000 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [3 x i8*], align 4
3001 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES:%.*]] = alloca [3 x i64], align 4
3002 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
3003 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3004 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3005 // CHECK11-NEXT: [[N_CASTED3:%.*]] = alloca i32, align 4
3006 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS4:%.*]] = alloca [3 x i8*], align 4
3007 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS5:%.*]] = alloca [3 x i8*], align 4
3008 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS6:%.*]] = alloca [3 x i8*], align 4
3009 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES7:%.*]] = alloca [3 x i64], align 4
3010 // CHECK11-NEXT: [[_TMP8:%.*]] = alloca i32, align 4
3011 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32, align 4
3012 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_10:%.*]] = alloca i32, align 4
3013 // CHECK11-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
3014 // CHECK11-NEXT: [[N_CASTED18:%.*]] = alloca i32, align 4
3015 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS19:%.*]] = alloca [4 x i8*], align 4
3016 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS20:%.*]] = alloca [4 x i8*], align 4
3017 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS21:%.*]] = alloca [4 x i8*], align 4
3018 // CHECK11-NEXT: [[DOTOFFLOAD_SIZES22:%.*]] = alloca [4 x i64], align 4
3019 // CHECK11-NEXT: [[_TMP23:%.*]] = alloca i32, align 4
3020 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_24:%.*]] = alloca i32, align 4
3021 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_25:%.*]] = alloca i32, align 4
3022 // CHECK11-NEXT: store i32 0, i32* [[RETVAL]], align 4
3023 // CHECK11-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3024 // CHECK11-NEXT: store i8** [[ARGV]], i8*** [[ARGV_ADDR]], align 4
3025 // CHECK11-NEXT: store i32 100, i32* [[N]], align 4
3026 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[N]], align 4
3027 // CHECK11-NEXT: [[TMP1:%.*]] = call i8* @llvm.stacksave()
3028 // CHECK11-NEXT: store i8* [[TMP1]], i8** [[SAVED_STACK]], align 4
3029 // CHECK11-NEXT: [[VLA:%.*]] = alloca i32, i32 [[TMP0]], align 4
3030 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[__VLA_EXPR0]], align 4
3031 // CHECK11-NEXT: store i32 10, i32* [[M]], align 4
3032 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[N]], align 4
3033 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[N_CASTED]], align 4
3034 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[N_CASTED]], align 4
3035 // CHECK11-NEXT: [[TMP4:%.*]] = mul nuw i32 [[TMP0]], 4
3036 // CHECK11-NEXT: [[TMP5:%.*]] = sext i32 [[TMP4]] to i64
3037 // CHECK11-NEXT: [[TMP6:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES]] to i8*
3038 // 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)
3039 // CHECK11-NEXT: [[TMP7:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3040 // CHECK11-NEXT: [[TMP8:%.*]] = bitcast i8** [[TMP7]] to i32*
3041 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP8]], align 4
3042 // CHECK11-NEXT: [[TMP9:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3043 // CHECK11-NEXT: [[TMP10:%.*]] = bitcast i8** [[TMP9]] to i32*
3044 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[TMP10]], align 4
3045 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3046 // CHECK11-NEXT: store i8* null, i8** [[TMP11]], align 4
3047 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
3048 // CHECK11-NEXT: [[TMP13:%.*]] = bitcast i8** [[TMP12]] to i32*
3049 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP13]], align 4
3050 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
3051 // CHECK11-NEXT: [[TMP15:%.*]] = bitcast i8** [[TMP14]] to i32*
3052 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP15]], align 4
3053 // CHECK11-NEXT: [[TMP16:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
3054 // CHECK11-NEXT: store i8* null, i8** [[TMP16]], align 4
3055 // CHECK11-NEXT: [[TMP17:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
3056 // CHECK11-NEXT: [[TMP18:%.*]] = bitcast i8** [[TMP17]] to i32**
3057 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP18]], align 4
3058 // CHECK11-NEXT: [[TMP19:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 2
3059 // CHECK11-NEXT: [[TMP20:%.*]] = bitcast i8** [[TMP19]] to i32**
3060 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP20]], align 4
3061 // CHECK11-NEXT: [[TMP21:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 2
3062 // CHECK11-NEXT: store i64 [[TMP5]], i64* [[TMP21]], align 4
3063 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
3064 // CHECK11-NEXT: store i8* null, i8** [[TMP22]], align 4
3065 // CHECK11-NEXT: [[TMP23:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3066 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3067 // CHECK11-NEXT: [[TMP25:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES]], i32 0, i32 0
3068 // CHECK11-NEXT: [[TMP26:%.*]] = load i32, i32* [[N]], align 4
3069 // CHECK11-NEXT: store i32 [[TMP26]], i32* [[DOTCAPTURE_EXPR_]], align 4
3070 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3071 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP27]], 0
3072 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3073 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3074 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3075 // CHECK11-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3076 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
3077 // CHECK11-NEXT: [[TMP29:%.*]] = zext i32 [[ADD]] to i64
3078 // CHECK11-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
3079 // CHECK11-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
3080 // CHECK11-NEXT: store i32 1, i32* [[TMP30]], align 4
3081 // CHECK11-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
3082 // CHECK11-NEXT: store i32 3, i32* [[TMP31]], align 4
3083 // CHECK11-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
3084 // CHECK11-NEXT: store i8** [[TMP23]], i8*** [[TMP32]], align 4
3085 // CHECK11-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
3086 // CHECK11-NEXT: store i8** [[TMP24]], i8*** [[TMP33]], align 4
3087 // CHECK11-NEXT: [[TMP34:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
3088 // CHECK11-NEXT: store i64* [[TMP25]], i64** [[TMP34]], align 4
3089 // CHECK11-NEXT: [[TMP35:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
3090 // CHECK11-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes, i32 0, i32 0), i64** [[TMP35]], align 4
3091 // CHECK11-NEXT: [[TMP36:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
3092 // CHECK11-NEXT: store i8** null, i8*** [[TMP36]], align 4
3093 // CHECK11-NEXT: [[TMP37:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
3094 // CHECK11-NEXT: store i8** null, i8*** [[TMP37]], align 4
3095 // CHECK11-NEXT: [[TMP38:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
3096 // CHECK11-NEXT: store i64 [[TMP29]], i64* [[TMP38]], align 8
3097 // CHECK11-NEXT: [[TMP39:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l106.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
3098 // CHECK11-NEXT: [[TMP40:%.*]] = icmp ne i32 [[TMP39]], 0
3099 // CHECK11-NEXT: br i1 [[TMP40]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3100 // CHECK11: omp_offload.failed:
3101 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l106(i32 [[TMP3]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3:[0-9]+]]
3102 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
3103 // CHECK11: omp_offload.cont:
3104 // CHECK11-NEXT: [[TMP41:%.*]] = load i32, i32* [[N]], align 4
3105 // CHECK11-NEXT: store i32 [[TMP41]], i32* [[N_CASTED3]], align 4
3106 // CHECK11-NEXT: [[TMP42:%.*]] = load i32, i32* [[N_CASTED3]], align 4
3107 // CHECK11-NEXT: [[TMP43:%.*]] = mul nuw i32 [[TMP0]], 4
3108 // CHECK11-NEXT: [[TMP44:%.*]] = sext i32 [[TMP43]] to i64
3109 // CHECK11-NEXT: [[TMP45:%.*]] = bitcast [3 x i64]* [[DOTOFFLOAD_SIZES7]] to i8*
3110 // CHECK11-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP45]], i8* align 4 bitcast ([3 x i64]* @.offload_sizes.4 to i8*), i32 24, i1 false)
3111 // CHECK11-NEXT: [[TMP46:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
3112 // CHECK11-NEXT: [[TMP47:%.*]] = bitcast i8** [[TMP46]] to i32*
3113 // CHECK11-NEXT: store i32 [[TMP42]], i32* [[TMP47]], align 4
3114 // CHECK11-NEXT: [[TMP48:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
3115 // CHECK11-NEXT: [[TMP49:%.*]] = bitcast i8** [[TMP48]] to i32*
3116 // CHECK11-NEXT: store i32 [[TMP42]], i32* [[TMP49]], align 4
3117 // CHECK11-NEXT: [[TMP50:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 0
3118 // CHECK11-NEXT: store i8* null, i8** [[TMP50]], align 4
3119 // CHECK11-NEXT: [[TMP51:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 1
3120 // CHECK11-NEXT: [[TMP52:%.*]] = bitcast i8** [[TMP51]] to i32*
3121 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP52]], align 4
3122 // CHECK11-NEXT: [[TMP53:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 1
3123 // CHECK11-NEXT: [[TMP54:%.*]] = bitcast i8** [[TMP53]] to i32*
3124 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP54]], align 4
3125 // CHECK11-NEXT: [[TMP55:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 1
3126 // CHECK11-NEXT: store i8* null, i8** [[TMP55]], align 4
3127 // CHECK11-NEXT: [[TMP56:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 2
3128 // CHECK11-NEXT: [[TMP57:%.*]] = bitcast i8** [[TMP56]] to i32**
3129 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP57]], align 4
3130 // CHECK11-NEXT: [[TMP58:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 2
3131 // CHECK11-NEXT: [[TMP59:%.*]] = bitcast i8** [[TMP58]] to i32**
3132 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP59]], align 4
3133 // CHECK11-NEXT: [[TMP60:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 2
3134 // CHECK11-NEXT: store i64 [[TMP44]], i64* [[TMP60]], align 4
3135 // CHECK11-NEXT: [[TMP61:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_MAPPERS6]], i32 0, i32 2
3136 // CHECK11-NEXT: store i8* null, i8** [[TMP61]], align 4
3137 // CHECK11-NEXT: [[TMP62:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_BASEPTRS4]], i32 0, i32 0
3138 // CHECK11-NEXT: [[TMP63:%.*]] = getelementptr inbounds [3 x i8*], [3 x i8*]* [[DOTOFFLOAD_PTRS5]], i32 0, i32 0
3139 // CHECK11-NEXT: [[TMP64:%.*]] = getelementptr inbounds [3 x i64], [3 x i64]* [[DOTOFFLOAD_SIZES7]], i32 0, i32 0
3140 // CHECK11-NEXT: [[TMP65:%.*]] = load i32, i32* [[N]], align 4
3141 // CHECK11-NEXT: store i32 [[TMP65]], i32* [[DOTCAPTURE_EXPR_9]], align 4
3142 // CHECK11-NEXT: [[TMP66:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_9]], align 4
3143 // CHECK11-NEXT: [[SUB11:%.*]] = sub nsw i32 [[TMP66]], 0
3144 // CHECK11-NEXT: [[DIV12:%.*]] = sdiv i32 [[SUB11]], 1
3145 // CHECK11-NEXT: [[SUB13:%.*]] = sub nsw i32 [[DIV12]], 1
3146 // CHECK11-NEXT: store i32 [[SUB13]], i32* [[DOTCAPTURE_EXPR_10]], align 4
3147 // CHECK11-NEXT: [[TMP67:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_10]], align 4
3148 // CHECK11-NEXT: [[ADD14:%.*]] = add nsw i32 [[TMP67]], 1
3149 // CHECK11-NEXT: [[TMP68:%.*]] = zext i32 [[ADD14]] to i64
3150 // CHECK11-NEXT: [[KERNEL_ARGS15:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
3151 // CHECK11-NEXT: [[TMP69:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 0
3152 // CHECK11-NEXT: store i32 1, i32* [[TMP69]], align 4
3153 // CHECK11-NEXT: [[TMP70:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 1
3154 // CHECK11-NEXT: store i32 3, i32* [[TMP70]], align 4
3155 // CHECK11-NEXT: [[TMP71:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 2
3156 // CHECK11-NEXT: store i8** [[TMP62]], i8*** [[TMP71]], align 4
3157 // CHECK11-NEXT: [[TMP72:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 3
3158 // CHECK11-NEXT: store i8** [[TMP63]], i8*** [[TMP72]], align 4
3159 // CHECK11-NEXT: [[TMP73:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 4
3160 // CHECK11-NEXT: store i64* [[TMP64]], i64** [[TMP73]], align 4
3161 // CHECK11-NEXT: [[TMP74:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 5
3162 // CHECK11-NEXT: store i64* getelementptr inbounds ([3 x i64], [3 x i64]* @.offload_maptypes.5, i32 0, i32 0), i64** [[TMP74]], align 4
3163 // CHECK11-NEXT: [[TMP75:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 6
3164 // CHECK11-NEXT: store i8** null, i8*** [[TMP75]], align 4
3165 // CHECK11-NEXT: [[TMP76:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 7
3166 // CHECK11-NEXT: store i8** null, i8*** [[TMP76]], align 4
3167 // CHECK11-NEXT: [[TMP77:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]], i32 0, i32 8
3168 // CHECK11-NEXT: store i64 [[TMP68]], i64* [[TMP77]], align 8
3169 // CHECK11-NEXT: [[TMP78:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l111.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS15]])
3170 // CHECK11-NEXT: [[TMP79:%.*]] = icmp ne i32 [[TMP78]], 0
3171 // CHECK11-NEXT: br i1 [[TMP79]], label [[OMP_OFFLOAD_FAILED16:%.*]], label [[OMP_OFFLOAD_CONT17:%.*]]
3172 // CHECK11: omp_offload.failed16:
3173 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l111(i32 [[TMP42]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
3174 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT17]]
3175 // CHECK11: omp_offload.cont17:
3176 // CHECK11-NEXT: [[TMP80:%.*]] = load i32, i32* [[M]], align 4
3177 // CHECK11-NEXT: store i32 [[TMP80]], i32* [[M_CASTED]], align 4
3178 // CHECK11-NEXT: [[TMP81:%.*]] = load i32, i32* [[M_CASTED]], align 4
3179 // CHECK11-NEXT: [[TMP82:%.*]] = load i32, i32* [[N]], align 4
3180 // CHECK11-NEXT: store i32 [[TMP82]], i32* [[N_CASTED18]], align 4
3181 // CHECK11-NEXT: [[TMP83:%.*]] = load i32, i32* [[N_CASTED18]], align 4
3182 // CHECK11-NEXT: [[TMP84:%.*]] = mul nuw i32 [[TMP0]], 4
3183 // CHECK11-NEXT: [[TMP85:%.*]] = sext i32 [[TMP84]] to i64
3184 // CHECK11-NEXT: [[TMP86:%.*]] = bitcast [4 x i64]* [[DOTOFFLOAD_SIZES22]] to i8*
3185 // CHECK11-NEXT: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 [[TMP86]], i8* align 4 bitcast ([4 x i64]* @.offload_sizes.8 to i8*), i32 32, i1 false)
3186 // CHECK11-NEXT: [[TMP87:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
3187 // CHECK11-NEXT: [[TMP88:%.*]] = bitcast i8** [[TMP87]] to i32*
3188 // CHECK11-NEXT: store i32 [[TMP81]], i32* [[TMP88]], align 4
3189 // CHECK11-NEXT: [[TMP89:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
3190 // CHECK11-NEXT: [[TMP90:%.*]] = bitcast i8** [[TMP89]] to i32*
3191 // CHECK11-NEXT: store i32 [[TMP81]], i32* [[TMP90]], align 4
3192 // CHECK11-NEXT: [[TMP91:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 0
3193 // CHECK11-NEXT: store i8* null, i8** [[TMP91]], align 4
3194 // CHECK11-NEXT: [[TMP92:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 1
3195 // CHECK11-NEXT: [[TMP93:%.*]] = bitcast i8** [[TMP92]] to i32*
3196 // CHECK11-NEXT: store i32 [[TMP83]], i32* [[TMP93]], align 4
3197 // CHECK11-NEXT: [[TMP94:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 1
3198 // CHECK11-NEXT: [[TMP95:%.*]] = bitcast i8** [[TMP94]] to i32*
3199 // CHECK11-NEXT: store i32 [[TMP83]], i32* [[TMP95]], align 4
3200 // CHECK11-NEXT: [[TMP96:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 1
3201 // CHECK11-NEXT: store i8* null, i8** [[TMP96]], align 4
3202 // CHECK11-NEXT: [[TMP97:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 2
3203 // CHECK11-NEXT: [[TMP98:%.*]] = bitcast i8** [[TMP97]] to i32*
3204 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP98]], align 4
3205 // CHECK11-NEXT: [[TMP99:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 2
3206 // CHECK11-NEXT: [[TMP100:%.*]] = bitcast i8** [[TMP99]] to i32*
3207 // CHECK11-NEXT: store i32 [[TMP0]], i32* [[TMP100]], align 4
3208 // CHECK11-NEXT: [[TMP101:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 2
3209 // CHECK11-NEXT: store i8* null, i8** [[TMP101]], align 4
3210 // CHECK11-NEXT: [[TMP102:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 3
3211 // CHECK11-NEXT: [[TMP103:%.*]] = bitcast i8** [[TMP102]] to i32**
3212 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP103]], align 4
3213 // CHECK11-NEXT: [[TMP104:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 3
3214 // CHECK11-NEXT: [[TMP105:%.*]] = bitcast i8** [[TMP104]] to i32**
3215 // CHECK11-NEXT: store i32* [[VLA]], i32** [[TMP105]], align 4
3216 // CHECK11-NEXT: [[TMP106:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES22]], i32 0, i32 3
3217 // CHECK11-NEXT: store i64 [[TMP85]], i64* [[TMP106]], align 4
3218 // CHECK11-NEXT: [[TMP107:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_MAPPERS21]], i32 0, i32 3
3219 // CHECK11-NEXT: store i8* null, i8** [[TMP107]], align 4
3220 // CHECK11-NEXT: [[TMP108:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_BASEPTRS19]], i32 0, i32 0
3221 // CHECK11-NEXT: [[TMP109:%.*]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[DOTOFFLOAD_PTRS20]], i32 0, i32 0
3222 // CHECK11-NEXT: [[TMP110:%.*]] = getelementptr inbounds [4 x i64], [4 x i64]* [[DOTOFFLOAD_SIZES22]], i32 0, i32 0
3223 // CHECK11-NEXT: [[TMP111:%.*]] = load i32, i32* [[N]], align 4
3224 // CHECK11-NEXT: store i32 [[TMP111]], i32* [[DOTCAPTURE_EXPR_24]], align 4
3225 // CHECK11-NEXT: [[TMP112:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_24]], align 4
3226 // CHECK11-NEXT: [[SUB26:%.*]] = sub nsw i32 [[TMP112]], 0
3227 // CHECK11-NEXT: [[DIV27:%.*]] = sdiv i32 [[SUB26]], 1
3228 // CHECK11-NEXT: [[SUB28:%.*]] = sub nsw i32 [[DIV27]], 1
3229 // CHECK11-NEXT: store i32 [[SUB28]], i32* [[DOTCAPTURE_EXPR_25]], align 4
3230 // CHECK11-NEXT: [[TMP113:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_25]], align 4
3231 // CHECK11-NEXT: [[ADD29:%.*]] = add nsw i32 [[TMP113]], 1
3232 // CHECK11-NEXT: [[TMP114:%.*]] = zext i32 [[ADD29]] to i64
3233 // CHECK11-NEXT: [[KERNEL_ARGS30:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
3234 // CHECK11-NEXT: [[TMP115:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 0
3235 // CHECK11-NEXT: store i32 1, i32* [[TMP115]], align 4
3236 // CHECK11-NEXT: [[TMP116:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 1
3237 // CHECK11-NEXT: store i32 4, i32* [[TMP116]], align 4
3238 // CHECK11-NEXT: [[TMP117:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 2
3239 // CHECK11-NEXT: store i8** [[TMP108]], i8*** [[TMP117]], align 4
3240 // CHECK11-NEXT: [[TMP118:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 3
3241 // CHECK11-NEXT: store i8** [[TMP109]], i8*** [[TMP118]], align 4
3242 // CHECK11-NEXT: [[TMP119:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 4
3243 // CHECK11-NEXT: store i64* [[TMP110]], i64** [[TMP119]], align 4
3244 // CHECK11-NEXT: [[TMP120:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 5
3245 // CHECK11-NEXT: store i64* getelementptr inbounds ([4 x i64], [4 x i64]* @.offload_maptypes.9, i32 0, i32 0), i64** [[TMP120]], align 4
3246 // CHECK11-NEXT: [[TMP121:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 6
3247 // CHECK11-NEXT: store i8** null, i8*** [[TMP121]], align 4
3248 // CHECK11-NEXT: [[TMP122:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 7
3249 // CHECK11-NEXT: store i8** null, i8*** [[TMP122]], align 4
3250 // CHECK11-NEXT: [[TMP123:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]], i32 0, i32 8
3251 // CHECK11-NEXT: store i64 [[TMP114]], i64* [[TMP123]], align 8
3252 // CHECK11-NEXT: [[TMP124:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS30]])
3253 // CHECK11-NEXT: [[TMP125:%.*]] = icmp ne i32 [[TMP124]], 0
3254 // CHECK11-NEXT: br i1 [[TMP125]], label [[OMP_OFFLOAD_FAILED31:%.*]], label [[OMP_OFFLOAD_CONT32:%.*]]
3255 // CHECK11: omp_offload.failed31:
3256 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116(i32 [[TMP81]], i32 [[TMP83]], i32 [[TMP0]], i32* [[VLA]]) #[[ATTR3]]
3257 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT32]]
3258 // CHECK11: omp_offload.cont32:
3259 // CHECK11-NEXT: [[TMP126:%.*]] = load i32, i32* [[ARGC_ADDR]], align 4
3260 // CHECK11-NEXT: [[CALL:%.*]] = call noundef i32 @_Z5tmainIiLi10EEiT_(i32 noundef [[TMP126]])
3261 // CHECK11-NEXT: store i32 [[CALL]], i32* [[RETVAL]], align 4
3262 // CHECK11-NEXT: [[TMP127:%.*]] = load i8*, i8** [[SAVED_STACK]], align 4
3263 // CHECK11-NEXT: call void @llvm.stackrestore(i8* [[TMP127]])
3264 // CHECK11-NEXT: [[TMP128:%.*]] = load i32, i32* [[RETVAL]], align 4
3265 // CHECK11-NEXT: ret i32 [[TMP128]]
3266 //
3267 //
3268 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l106
3269 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
3270 // CHECK11-NEXT: entry:
3271 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
3272 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3273 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3274 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
3275 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3276 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3277 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3278 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3279 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
3280 // CHECK11-NEXT: ret void
3281 //
3282 //
3283 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
3284 // 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]] {
3285 // CHECK11-NEXT: entry:
3286 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3287 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3288 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
3289 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3290 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3291 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3292 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
3293 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3294 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3295 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
3296 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3297 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3298 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3299 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3300 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
3301 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3302 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3303 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
3304 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3305 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3306 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
3307 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3308 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3309 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3310 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3311 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3312 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3313 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3314 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3315 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3316 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
3317 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3318 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3319 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3320 // CHECK11: omp.precond.then:
3321 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3322 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3323 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
3324 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3325 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3326 // CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3327 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3328 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3329 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3330 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3331 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
3332 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3333 // CHECK11: cond.true:
3334 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3335 // CHECK11-NEXT: br label [[COND_END:%.*]]
3336 // CHECK11: cond.false:
3337 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3338 // CHECK11-NEXT: br label [[COND_END]]
3339 // CHECK11: cond.end:
3340 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3341 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3342 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3343 // CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
3344 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3345 // CHECK11: omp.inner.for.cond:
3346 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3347 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3348 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
3349 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3350 // CHECK11: omp.inner.for.body:
3351 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3352 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3353 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
3354 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3355 // CHECK11: omp.inner.for.inc:
3356 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3357 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3358 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
3359 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3360 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
3361 // CHECK11: omp.inner.for.end:
3362 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3363 // CHECK11: omp.loop.exit:
3364 // CHECK11-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3365 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
3366 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
3367 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
3368 // CHECK11: omp.precond.end:
3369 // CHECK11-NEXT: ret void
3370 //
3371 //
3372 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
3373 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
3374 // CHECK11-NEXT: entry:
3375 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3376 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3377 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
3378 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
3379 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
3380 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3381 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3382 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3383 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
3384 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3385 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3386 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
3387 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3388 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3389 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3390 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3391 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
3392 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3393 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3394 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3395 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3396 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
3397 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3398 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3399 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
3400 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3401 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3402 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3403 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3404 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3405 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3406 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3407 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3408 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3409 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
3410 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3411 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3412 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3413 // CHECK11: omp.precond.then:
3414 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3415 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3416 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
3417 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3418 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3419 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
3420 // CHECK11-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
3421 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3422 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3423 // CHECK11-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3424 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3425 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3426 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3427 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3428 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
3429 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3430 // CHECK11: cond.true:
3431 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3432 // CHECK11-NEXT: br label [[COND_END:%.*]]
3433 // CHECK11: cond.false:
3434 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3435 // CHECK11-NEXT: br label [[COND_END]]
3436 // CHECK11: cond.end:
3437 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
3438 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3439 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3440 // CHECK11-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
3441 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3442 // CHECK11: omp.inner.for.cond:
3443 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3444 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3445 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
3446 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3447 // CHECK11: omp.inner.for.body:
3448 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3449 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
3450 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3451 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
3452 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
3453 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
3454 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3455 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3456 // CHECK11: omp.body.continue:
3457 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3458 // CHECK11: omp.inner.for.inc:
3459 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3460 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
3461 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
3462 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
3463 // CHECK11: omp.inner.for.end:
3464 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3465 // CHECK11: omp.loop.exit:
3466 // CHECK11-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3467 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
3468 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
3469 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
3470 // CHECK11: omp.precond.end:
3471 // CHECK11-NEXT: ret void
3472 //
3473 //
3474 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l111
3475 // CHECK11-SAME: (i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
3476 // CHECK11-NEXT: entry:
3477 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
3478 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3479 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3480 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
3481 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3482 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3483 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3484 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3485 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]])
3486 // CHECK11-NEXT: ret void
3487 //
3488 //
3489 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
3490 // 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]] {
3491 // CHECK11-NEXT: entry:
3492 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3493 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3494 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
3495 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3496 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3497 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3498 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
3499 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3500 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3501 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
3502 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3503 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3504 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3505 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3506 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
3507 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3508 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3509 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
3510 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3511 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3512 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
3513 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3514 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3515 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3516 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3517 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3518 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3519 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3520 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3521 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3522 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
3523 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3524 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3525 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3526 // CHECK11: omp.precond.then:
3527 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3528 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3529 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
3530 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3531 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3532 // CHECK11-NEXT: [[TMP7:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3533 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[TMP7]], align 4
3534 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP8]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3535 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3536 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3537 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
3538 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3539 // CHECK11: cond.true:
3540 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3541 // CHECK11-NEXT: br label [[COND_END:%.*]]
3542 // CHECK11: cond.false:
3543 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3544 // CHECK11-NEXT: br label [[COND_END]]
3545 // CHECK11: cond.end:
3546 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3547 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3548 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3549 // CHECK11-NEXT: store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
3550 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3551 // CHECK11: omp.inner.for.cond:
3552 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3553 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3554 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
3555 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3556 // CHECK11: omp.inner.for.body:
3557 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3558 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3559 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 5, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP16]], i32 [[TMP17]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]])
3560 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3561 // CHECK11: omp.inner.for.inc:
3562 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3563 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3564 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
3565 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3566 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
3567 // CHECK11: omp.inner.for.end:
3568 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3569 // CHECK11: omp.loop.exit:
3570 // CHECK11-NEXT: [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3571 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
3572 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]])
3573 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
3574 // CHECK11: omp.precond.end:
3575 // CHECK11-NEXT: ret void
3576 //
3577 //
3578 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
3579 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
3580 // CHECK11-NEXT: entry:
3581 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3582 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3583 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
3584 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
3585 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
3586 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3587 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3588 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3589 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
3590 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3591 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3592 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
3593 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3594 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3595 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3596 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3597 // CHECK11-NEXT: [[I3:%.*]] = alloca i32, align 4
3598 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3599 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3600 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3601 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3602 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
3603 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3604 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3605 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
3606 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3607 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3608 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3609 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_]], align 4
3610 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3611 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3612 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3613 // CHECK11-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
3614 // CHECK11-NEXT: store i32 [[SUB2]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3615 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
3616 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3617 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3618 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3619 // CHECK11: omp.precond.then:
3620 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3621 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3622 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
3623 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3624 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3625 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
3626 // CHECK11-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
3627 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3628 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3629 // CHECK11-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3630 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3631 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3632 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3633 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3634 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
3635 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3636 // CHECK11: cond.true:
3637 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3638 // CHECK11-NEXT: br label [[COND_END:%.*]]
3639 // CHECK11: cond.false:
3640 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3641 // CHECK11-NEXT: br label [[COND_END]]
3642 // CHECK11: cond.end:
3643 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
3644 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3645 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3646 // CHECK11-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
3647 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3648 // CHECK11: omp.inner.for.cond:
3649 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3650 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3651 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
3652 // CHECK11-NEXT: br i1 [[CMP5]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3653 // CHECK11: omp.inner.for.body:
3654 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3655 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
3656 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3657 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I3]], align 4
3658 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[I3]], align 4
3659 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
3660 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3661 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3662 // CHECK11: omp.body.continue:
3663 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3664 // CHECK11: omp.inner.for.inc:
3665 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3666 // CHECK11-NEXT: [[ADD6:%.*]] = add nsw i32 [[TMP20]], 1
3667 // CHECK11-NEXT: store i32 [[ADD6]], i32* [[DOTOMP_IV]], align 4
3668 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
3669 // CHECK11: omp.inner.for.end:
3670 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3671 // CHECK11: omp.loop.exit:
3672 // CHECK11-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3673 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
3674 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
3675 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
3676 // CHECK11: omp.precond.end:
3677 // CHECK11-NEXT: ret void
3678 //
3679 //
3680 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l116
3681 // CHECK11-SAME: (i32 noundef [[M:%.*]], i32 noundef [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]]) #[[ATTR2]] {
3682 // CHECK11-NEXT: entry:
3683 // CHECK11-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
3684 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32, align 4
3685 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3686 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3687 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
3688 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
3689 // CHECK11-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
3690 // CHECK11-NEXT: store i32 [[N]], i32* [[N_ADDR]], align 4
3691 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3692 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3693 // CHECK11-NEXT: [[TMP0:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3694 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3695 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[M_ADDR]], align 4
3696 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR_]], align 4
3697 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
3698 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
3699 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
3700 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32*, i32, i32*, i32)* @.omp_outlined..6 to void (i32*, i32*, ...)*), i32* [[N_ADDR]], i32 [[TMP0]], i32* [[TMP1]], i32 [[TMP4]])
3701 // CHECK11-NEXT: ret void
3702 //
3703 //
3704 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..6
3705 // 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]] {
3706 // CHECK11-NEXT: entry:
3707 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3708 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3709 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
3710 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3711 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3712 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
3713 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3714 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
3715 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3716 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
3717 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
3718 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3719 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3720 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3721 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3722 // CHECK11-NEXT: [[I4:%.*]] = alloca i32, align 4
3723 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
3724 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3725 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3726 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
3727 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3728 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3729 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
3730 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
3731 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3732 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3733 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3734 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3735 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3736 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3737 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3738 // CHECK11-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
3739 // CHECK11-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
3740 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
3741 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3742 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3743 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3744 // CHECK11: omp.precond.then:
3745 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3746 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3747 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_COMB_UB]], align 4
3748 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3749 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3750 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
3751 // CHECK11-NEXT: [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3752 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
3753 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP7]])
3754 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3755 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3756 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
3757 // CHECK11-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3758 // CHECK11: cond.true:
3759 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3760 // CHECK11-NEXT: br label [[COND_END:%.*]]
3761 // CHECK11: cond.false:
3762 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3763 // CHECK11-NEXT: br label [[COND_END]]
3764 // CHECK11: cond.end:
3765 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
3766 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3767 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3768 // CHECK11-NEXT: store i32 [[TMP14]], i32* [[DOTOMP_IV]], align 4
3769 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3770 // CHECK11: omp.inner.for.cond:
3771 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3772 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3773 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
3774 // CHECK11-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
3775 // CHECK11-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3776 // CHECK11: omp.inner.for.body:
3777 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3778 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3779 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
3780 // CHECK11-NEXT: store i32 [[TMP19]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
3781 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
3782 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 6, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, i32*, i32, i32*, i32)* @.omp_outlined..7 to void (i32*, i32*, ...)*), i32 [[TMP17]], i32 [[TMP18]], i32* [[TMP0]], i32 [[TMP1]], i32* [[TMP2]], i32 [[TMP20]])
3783 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3784 // CHECK11: omp.inner.for.inc:
3785 // CHECK11-NEXT: [[TMP21:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3786 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3787 // CHECK11-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP21]], [[TMP22]]
3788 // CHECK11-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
3789 // CHECK11-NEXT: [[TMP23:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3790 // CHECK11-NEXT: [[TMP24:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3791 // CHECK11-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP23]], [[TMP24]]
3792 // CHECK11-NEXT: store i32 [[ADD8]], i32* [[DOTOMP_COMB_LB]], align 4
3793 // CHECK11-NEXT: [[TMP25:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3794 // CHECK11-NEXT: [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3795 // CHECK11-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP25]], [[TMP26]]
3796 // CHECK11-NEXT: store i32 [[ADD9]], i32* [[DOTOMP_COMB_UB]], align 4
3797 // CHECK11-NEXT: [[TMP27:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3798 // CHECK11-NEXT: [[TMP28:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3799 // CHECK11-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP27]], [[TMP28]]
3800 // CHECK11-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
3801 // CHECK11: cond.true11:
3802 // CHECK11-NEXT: [[TMP29:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3803 // CHECK11-NEXT: br label [[COND_END13:%.*]]
3804 // CHECK11: cond.false12:
3805 // CHECK11-NEXT: [[TMP30:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3806 // CHECK11-NEXT: br label [[COND_END13]]
3807 // CHECK11: cond.end13:
3808 // CHECK11-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP29]], [[COND_TRUE11]] ], [ [[TMP30]], [[COND_FALSE12]] ]
3809 // CHECK11-NEXT: store i32 [[COND14]], i32* [[DOTOMP_COMB_UB]], align 4
3810 // CHECK11-NEXT: [[TMP31:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3811 // CHECK11-NEXT: store i32 [[TMP31]], i32* [[DOTOMP_IV]], align 4
3812 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
3813 // CHECK11: omp.inner.for.end:
3814 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3815 // CHECK11: omp.loop.exit:
3816 // CHECK11-NEXT: [[TMP32:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3817 // CHECK11-NEXT: [[TMP33:%.*]] = load i32, i32* [[TMP32]], align 4
3818 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP33]])
3819 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
3820 // CHECK11: omp.precond.end:
3821 // CHECK11-NEXT: ret void
3822 //
3823 //
3824 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..7
3825 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[N:%.*]], i32 noundef [[VLA:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
3826 // CHECK11-NEXT: entry:
3827 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
3828 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
3829 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
3830 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
3831 // CHECK11-NEXT: [[N_ADDR:%.*]] = alloca i32*, align 4
3832 // CHECK11-NEXT: [[VLA_ADDR:%.*]] = alloca i32, align 4
3833 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca i32*, align 4
3834 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
3835 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
3836 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
3837 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4
3838 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4
3839 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
3840 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
3841 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
3842 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3843 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3844 // CHECK11-NEXT: [[I4:%.*]] = alloca i32, align 4
3845 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
3846 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
3847 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3848 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3849 // CHECK11-NEXT: store i32* [[N]], i32** [[N_ADDR]], align 4
3850 // CHECK11-NEXT: store i32 [[VLA]], i32* [[VLA_ADDR]], align 4
3851 // CHECK11-NEXT: store i32* [[A]], i32** [[A_ADDR]], align 4
3852 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
3853 // CHECK11-NEXT: [[TMP0:%.*]] = load i32*, i32** [[N_ADDR]], align 4
3854 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[VLA_ADDR]], align 4
3855 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[A_ADDR]], align 4
3856 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP0]], align 4
3857 // CHECK11-NEXT: store i32 [[TMP3]], i32* [[DOTCAPTURE_EXPR_1]], align 4
3858 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3859 // CHECK11-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
3860 // CHECK11-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
3861 // CHECK11-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
3862 // CHECK11-NEXT: store i32 [[SUB3]], i32* [[DOTCAPTURE_EXPR_2]], align 4
3863 // CHECK11-NEXT: store i32 0, i32* [[I]], align 4
3864 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_1]], align 4
3865 // CHECK11-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
3866 // CHECK11-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3867 // CHECK11: omp.precond.then:
3868 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
3869 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3870 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_UB]], align 4
3871 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
3872 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
3873 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_LB]], align 4
3874 // CHECK11-NEXT: store i32 [[TMP8]], i32* [[DOTOMP_UB]], align 4
3875 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3876 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3877 // CHECK11-NEXT: [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3878 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3879 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3880 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3881 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3882 // CHECK11-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP11]], [[TMP12]]
3883 // CHECK11-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3884 // CHECK11: cond.true:
3885 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_2]], align 4
3886 // CHECK11-NEXT: br label [[COND_END:%.*]]
3887 // CHECK11: cond.false:
3888 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3889 // CHECK11-NEXT: br label [[COND_END]]
3890 // CHECK11: cond.end:
3891 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ [[TMP13]], [[COND_TRUE]] ], [ [[TMP14]], [[COND_FALSE]] ]
3892 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3893 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3894 // CHECK11-NEXT: store i32 [[TMP15]], i32* [[DOTOMP_IV]], align 4
3895 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
3896 // CHECK11: omp.inner.for.cond:
3897 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3898 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3899 // CHECK11-NEXT: [[CMP6:%.*]] = icmp sle i32 [[TMP16]], [[TMP17]]
3900 // CHECK11-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3901 // CHECK11: omp.inner.for.body:
3902 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3903 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP18]], 1
3904 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3905 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I4]], align 4
3906 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[I4]], align 4
3907 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, i32* [[TMP2]], i32 [[TMP19]]
3908 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
3909 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
3910 // CHECK11: omp.body.continue:
3911 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
3912 // CHECK11: omp.inner.for.inc:
3913 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3914 // CHECK11-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP20]], 1
3915 // CHECK11-NEXT: store i32 [[ADD7]], i32* [[DOTOMP_IV]], align 4
3916 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
3917 // CHECK11: omp.inner.for.end:
3918 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
3919 // CHECK11: omp.loop.exit:
3920 // CHECK11-NEXT: [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
3921 // CHECK11-NEXT: [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
3922 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
3923 // CHECK11-NEXT: br label [[OMP_PRECOND_END]]
3924 // CHECK11: omp.precond.end:
3925 // CHECK11-NEXT: ret void
3926 //
3927 //
3928 // CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiLi10EEiT_
3929 // CHECK11-SAME: (i32 noundef [[ARGC:%.*]]) #[[ATTR5:[0-9]+]] comdat {
3930 // CHECK11-NEXT: entry:
3931 // CHECK11-NEXT: [[ARGC_ADDR:%.*]] = alloca i32, align 4
3932 // CHECK11-NEXT: [[A:%.*]] = alloca [10 x i32], align 4
3933 // CHECK11-NEXT: [[M:%.*]] = alloca i32, align 4
3934 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 4
3935 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 4
3936 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 4
3937 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
3938 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS1:%.*]] = alloca [1 x i8*], align 4
3939 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS2:%.*]] = alloca [1 x i8*], align 4
3940 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS3:%.*]] = alloca [1 x i8*], align 4
3941 // CHECK11-NEXT: [[_TMP4:%.*]] = alloca i32, align 4
3942 // CHECK11-NEXT: [[M_CASTED:%.*]] = alloca i32, align 4
3943 // CHECK11-NEXT: [[DOTOFFLOAD_BASEPTRS8:%.*]] = alloca [2 x i8*], align 4
3944 // CHECK11-NEXT: [[DOTOFFLOAD_PTRS9:%.*]] = alloca [2 x i8*], align 4
3945 // CHECK11-NEXT: [[DOTOFFLOAD_MAPPERS10:%.*]] = alloca [2 x i8*], align 4
3946 // CHECK11-NEXT: [[_TMP11:%.*]] = alloca i32, align 4
3947 // CHECK11-NEXT: store i32 [[ARGC]], i32* [[ARGC_ADDR]], align 4
3948 // CHECK11-NEXT: store i32 10, i32* [[M]], align 4
3949 // CHECK11-NEXT: [[TMP0:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3950 // CHECK11-NEXT: [[TMP1:%.*]] = bitcast i8** [[TMP0]] to [10 x i32]**
3951 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP1]], align 4
3952 // CHECK11-NEXT: [[TMP2:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3953 // CHECK11-NEXT: [[TMP3:%.*]] = bitcast i8** [[TMP2]] to [10 x i32]**
3954 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP3]], align 4
3955 // CHECK11-NEXT: [[TMP4:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
3956 // CHECK11-NEXT: store i8* null, i8** [[TMP4]], align 4
3957 // CHECK11-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3958 // CHECK11-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3959 // CHECK11-NEXT: [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
3960 // CHECK11-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 0
3961 // CHECK11-NEXT: store i32 1, i32* [[TMP7]], align 4
3962 // CHECK11-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 1
3963 // CHECK11-NEXT: store i32 1, i32* [[TMP8]], align 4
3964 // CHECK11-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 2
3965 // CHECK11-NEXT: store i8** [[TMP5]], i8*** [[TMP9]], align 4
3966 // CHECK11-NEXT: [[TMP10:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 3
3967 // CHECK11-NEXT: store i8** [[TMP6]], i8*** [[TMP10]], align 4
3968 // CHECK11-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 4
3969 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.12, i32 0, i32 0), i64** [[TMP11]], align 4
3970 // CHECK11-NEXT: [[TMP12:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 5
3971 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.13, i32 0, i32 0), i64** [[TMP12]], align 4
3972 // CHECK11-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 6
3973 // CHECK11-NEXT: store i8** null, i8*** [[TMP13]], align 4
3974 // CHECK11-NEXT: [[TMP14:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 7
3975 // CHECK11-NEXT: store i8** null, i8*** [[TMP14]], align 4
3976 // CHECK11-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]], i32 0, i32 8
3977 // CHECK11-NEXT: store i64 10, i64* [[TMP15]], align 8
3978 // CHECK11-NEXT: [[TMP16:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS]])
3979 // CHECK11-NEXT: [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
3980 // CHECK11-NEXT: br i1 [[TMP17]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3981 // CHECK11: omp_offload.failed:
3982 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84([10 x i32]* [[A]]) #[[ATTR3]]
3983 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT]]
3984 // CHECK11: omp_offload.cont:
3985 // CHECK11-NEXT: [[TMP18:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
3986 // CHECK11-NEXT: [[TMP19:%.*]] = bitcast i8** [[TMP18]] to [10 x i32]**
3987 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP19]], align 4
3988 // CHECK11-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
3989 // CHECK11-NEXT: [[TMP21:%.*]] = bitcast i8** [[TMP20]] to [10 x i32]**
3990 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP21]], align 4
3991 // CHECK11-NEXT: [[TMP22:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS3]], i32 0, i32 0
3992 // CHECK11-NEXT: store i8* null, i8** [[TMP22]], align 4
3993 // CHECK11-NEXT: [[TMP23:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS1]], i32 0, i32 0
3994 // CHECK11-NEXT: [[TMP24:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS2]], i32 0, i32 0
3995 // CHECK11-NEXT: [[KERNEL_ARGS5:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
3996 // CHECK11-NEXT: [[TMP25:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 0
3997 // CHECK11-NEXT: store i32 1, i32* [[TMP25]], align 4
3998 // CHECK11-NEXT: [[TMP26:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 1
3999 // CHECK11-NEXT: store i32 1, i32* [[TMP26]], align 4
4000 // CHECK11-NEXT: [[TMP27:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 2
4001 // CHECK11-NEXT: store i8** [[TMP23]], i8*** [[TMP27]], align 4
4002 // CHECK11-NEXT: [[TMP28:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 3
4003 // CHECK11-NEXT: store i8** [[TMP24]], i8*** [[TMP28]], align 4
4004 // CHECK11-NEXT: [[TMP29:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 4
4005 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64** [[TMP29]], align 4
4006 // CHECK11-NEXT: [[TMP30:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 5
4007 // CHECK11-NEXT: store i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i64** [[TMP30]], align 4
4008 // CHECK11-NEXT: [[TMP31:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 6
4009 // CHECK11-NEXT: store i8** null, i8*** [[TMP31]], align 4
4010 // CHECK11-NEXT: [[TMP32:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 7
4011 // CHECK11-NEXT: store i8** null, i8*** [[TMP32]], align 4
4012 // CHECK11-NEXT: [[TMP33:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]], i32 0, i32 8
4013 // CHECK11-NEXT: store i64 10, i64* [[TMP33]], align 8
4014 // CHECK11-NEXT: [[TMP34:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS5]])
4015 // CHECK11-NEXT: [[TMP35:%.*]] = icmp ne i32 [[TMP34]], 0
4016 // CHECK11-NEXT: br i1 [[TMP35]], label [[OMP_OFFLOAD_FAILED6:%.*]], label [[OMP_OFFLOAD_CONT7:%.*]]
4017 // CHECK11: omp_offload.failed6:
4018 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89([10 x i32]* [[A]]) #[[ATTR3]]
4019 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT7]]
4020 // CHECK11: omp_offload.cont7:
4021 // CHECK11-NEXT: [[TMP36:%.*]] = load i32, i32* [[M]], align 4
4022 // CHECK11-NEXT: store i32 [[TMP36]], i32* [[M_CASTED]], align 4
4023 // CHECK11-NEXT: [[TMP37:%.*]] = load i32, i32* [[M_CASTED]], align 4
4024 // CHECK11-NEXT: [[TMP38:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
4025 // CHECK11-NEXT: [[TMP39:%.*]] = bitcast i8** [[TMP38]] to i32*
4026 // CHECK11-NEXT: store i32 [[TMP37]], i32* [[TMP39]], align 4
4027 // CHECK11-NEXT: [[TMP40:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
4028 // CHECK11-NEXT: [[TMP41:%.*]] = bitcast i8** [[TMP40]] to i32*
4029 // CHECK11-NEXT: store i32 [[TMP37]], i32* [[TMP41]], align 4
4030 // CHECK11-NEXT: [[TMP42:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 0
4031 // CHECK11-NEXT: store i8* null, i8** [[TMP42]], align 4
4032 // CHECK11-NEXT: [[TMP43:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 1
4033 // CHECK11-NEXT: [[TMP44:%.*]] = bitcast i8** [[TMP43]] to [10 x i32]**
4034 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP44]], align 4
4035 // CHECK11-NEXT: [[TMP45:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 1
4036 // CHECK11-NEXT: [[TMP46:%.*]] = bitcast i8** [[TMP45]] to [10 x i32]**
4037 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[TMP46]], align 4
4038 // CHECK11-NEXT: [[TMP47:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS10]], i32 0, i32 1
4039 // CHECK11-NEXT: store i8* null, i8** [[TMP47]], align 4
4040 // CHECK11-NEXT: [[TMP48:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS8]], i32 0, i32 0
4041 // CHECK11-NEXT: [[TMP49:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS9]], i32 0, i32 0
4042 // CHECK11-NEXT: [[KERNEL_ARGS12:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS]], align 8
4043 // CHECK11-NEXT: [[TMP50:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 0
4044 // CHECK11-NEXT: store i32 1, i32* [[TMP50]], align 4
4045 // CHECK11-NEXT: [[TMP51:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 1
4046 // CHECK11-NEXT: store i32 2, i32* [[TMP51]], align 4
4047 // CHECK11-NEXT: [[TMP52:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 2
4048 // CHECK11-NEXT: store i8** [[TMP48]], i8*** [[TMP52]], align 4
4049 // CHECK11-NEXT: [[TMP53:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 3
4050 // CHECK11-NEXT: store i8** [[TMP49]], i8*** [[TMP53]], align 4
4051 // CHECK11-NEXT: [[TMP54:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 4
4052 // CHECK11-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.20, i32 0, i32 0), i64** [[TMP54]], align 4
4053 // CHECK11-NEXT: [[TMP55:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 5
4054 // CHECK11-NEXT: store i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.21, i32 0, i32 0), i64** [[TMP55]], align 4
4055 // CHECK11-NEXT: [[TMP56:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 6
4056 // CHECK11-NEXT: store i8** null, i8*** [[TMP56]], align 4
4057 // CHECK11-NEXT: [[TMP57:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 7
4058 // CHECK11-NEXT: store i8** null, i8*** [[TMP57]], align 4
4059 // CHECK11-NEXT: [[TMP58:%.*]] = getelementptr inbounds [[STRUCT___TGT_KERNEL_ARGUMENTS]], %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]], i32 0, i32 8
4060 // CHECK11-NEXT: store i64 10, i64* [[TMP58]], align 8
4061 // CHECK11-NEXT: [[TMP59:%.*]] = call i32 @__tgt_target_kernel(%struct.ident_t* @[[GLOB3]], i64 -1, i32 0, i32 0, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l94.region_id, %struct.__tgt_kernel_arguments* [[KERNEL_ARGS12]])
4062 // CHECK11-NEXT: [[TMP60:%.*]] = icmp ne i32 [[TMP59]], 0
4063 // CHECK11-NEXT: br i1 [[TMP60]], label [[OMP_OFFLOAD_FAILED13:%.*]], label [[OMP_OFFLOAD_CONT14:%.*]]
4064 // CHECK11: omp_offload.failed13:
4065 // CHECK11-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l94(i32 [[TMP37]], [10 x i32]* [[A]]) #[[ATTR3]]
4066 // CHECK11-NEXT: br label [[OMP_OFFLOAD_CONT14]]
4067 // CHECK11: omp_offload.cont14:
4068 // CHECK11-NEXT: ret i32 0
4069 //
4070 //
4071 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l84
4072 // CHECK11-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4073 // CHECK11-NEXT: entry:
4074 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4075 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4076 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4077 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..10 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
4078 // CHECK11-NEXT: ret void
4079 //
4080 //
4081 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..10
4082 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4083 // CHECK11-NEXT: entry:
4084 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4085 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4086 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4087 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4088 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4089 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4090 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4091 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4092 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4093 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4094 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4095 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4096 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4097 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4098 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4099 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
4100 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4101 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4102 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4103 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4104 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4105 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4106 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
4107 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4108 // CHECK11: cond.true:
4109 // CHECK11-NEXT: br label [[COND_END:%.*]]
4110 // CHECK11: cond.false:
4111 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4112 // CHECK11-NEXT: br label [[COND_END]]
4113 // CHECK11: cond.end:
4114 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4115 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4116 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4117 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4118 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4119 // CHECK11: omp.inner.for.cond:
4120 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4121 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4122 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4123 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4124 // CHECK11: omp.inner.for.body:
4125 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4126 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4127 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
4128 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4129 // CHECK11: omp.inner.for.inc:
4130 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4131 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4132 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
4133 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4134 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
4135 // CHECK11: omp.inner.for.end:
4136 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4137 // CHECK11: omp.loop.exit:
4138 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4139 // CHECK11-NEXT: ret void
4140 //
4141 //
4142 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..11
4143 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4144 // CHECK11-NEXT: entry:
4145 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4146 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4147 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
4148 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
4149 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4150 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4151 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4152 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4153 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4154 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4155 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4156 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4157 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4158 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4159 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4160 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4161 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4162 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4163 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4164 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
4165 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4166 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4167 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
4168 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
4169 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4170 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4171 // CHECK11-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4172 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4173 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4174 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4175 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
4176 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4177 // CHECK11: cond.true:
4178 // CHECK11-NEXT: br label [[COND_END:%.*]]
4179 // CHECK11: cond.false:
4180 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4181 // CHECK11-NEXT: br label [[COND_END]]
4182 // CHECK11: cond.end:
4183 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
4184 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4185 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4186 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
4187 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4188 // CHECK11: omp.inner.for.cond:
4189 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4190 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4191 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
4192 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4193 // CHECK11: omp.inner.for.body:
4194 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4195 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
4196 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4197 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4198 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
4199 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
4200 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4201 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4202 // CHECK11: omp.body.continue:
4203 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4204 // CHECK11: omp.inner.for.inc:
4205 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4206 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
4207 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
4208 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
4209 // CHECK11: omp.inner.for.end:
4210 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4211 // CHECK11: omp.loop.exit:
4212 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
4213 // CHECK11-NEXT: ret void
4214 //
4215 //
4216 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l89
4217 // CHECK11-SAME: ([10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4218 // CHECK11-NEXT: entry:
4219 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4220 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4221 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4222 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]])
4223 // CHECK11-NEXT: ret void
4224 //
4225 //
4226 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..14
4227 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4228 // CHECK11-NEXT: entry:
4229 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4230 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4231 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4232 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4233 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4234 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4235 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4236 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4237 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4238 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4239 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4240 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4241 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4242 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4243 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4244 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
4245 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4246 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4247 // CHECK11-NEXT: [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4248 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4249 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4250 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4251 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 9
4252 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4253 // CHECK11: cond.true:
4254 // CHECK11-NEXT: br label [[COND_END:%.*]]
4255 // CHECK11: cond.false:
4256 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4257 // CHECK11-NEXT: br label [[COND_END]]
4258 // CHECK11: cond.end:
4259 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4260 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4261 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4262 // CHECK11-NEXT: store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4263 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4264 // CHECK11: omp.inner.for.cond:
4265 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4266 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4267 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4268 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4269 // CHECK11: omp.inner.for.body:
4270 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4271 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4272 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 3, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]])
4273 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4274 // CHECK11: omp.inner.for.inc:
4275 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4276 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4277 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP10]], [[TMP11]]
4278 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4279 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
4280 // CHECK11: omp.inner.for.end:
4281 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4282 // CHECK11: omp.loop.exit:
4283 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4284 // CHECK11-NEXT: ret void
4285 //
4286 //
4287 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..15
4288 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4289 // CHECK11-NEXT: entry:
4290 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4291 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4292 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
4293 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
4294 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4295 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4296 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4297 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4298 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4299 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4300 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4301 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4302 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4303 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4304 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4305 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4306 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4307 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4308 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4309 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
4310 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4311 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4312 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
4313 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
4314 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4315 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4316 // CHECK11-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4317 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4318 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4319 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4320 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
4321 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4322 // CHECK11: cond.true:
4323 // CHECK11-NEXT: br label [[COND_END:%.*]]
4324 // CHECK11: cond.false:
4325 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4326 // CHECK11-NEXT: br label [[COND_END]]
4327 // CHECK11: cond.end:
4328 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
4329 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4330 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4331 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
4332 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4333 // CHECK11: omp.inner.for.cond:
4334 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4335 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4336 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
4337 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4338 // CHECK11: omp.inner.for.body:
4339 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4340 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
4341 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4342 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4343 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
4344 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
4345 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4346 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4347 // CHECK11: omp.body.continue:
4348 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4349 // CHECK11: omp.inner.for.inc:
4350 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4351 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
4352 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
4353 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
4354 // CHECK11: omp.inner.for.end:
4355 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4356 // CHECK11: omp.loop.exit:
4357 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
4358 // CHECK11-NEXT: ret void
4359 //
4360 //
4361 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiLi10EEiT__l94
4362 // CHECK11-SAME: (i32 noundef [[M:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]]) #[[ATTR2]] {
4363 // CHECK11-NEXT: entry:
4364 // CHECK11-NEXT: [[M_ADDR:%.*]] = alloca i32, align 4
4365 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4366 // CHECK11-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4
4367 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4368 // CHECK11-NEXT: store i32 [[M]], i32* [[M_ADDR]], align 4
4369 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4370 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4371 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[M_ADDR]], align 4
4372 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[DOTCAPTURE_EXPR_]], align 4
4373 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]], align 4
4374 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4375 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4376 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [10 x i32]*, i32)* @.omp_outlined..18 to void (i32*, i32*, ...)*), [10 x i32]* [[TMP0]], i32 [[TMP3]])
4377 // CHECK11-NEXT: ret void
4378 //
4379 //
4380 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..18
4381 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
4382 // CHECK11-NEXT: entry:
4383 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4384 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4385 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4386 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4387 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4388 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4389 // CHECK11-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4390 // CHECK11-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4391 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4392 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4393 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4394 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i32, align 4
4395 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4396 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4397 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4398 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4399 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4400 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4401 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_COMB_UB]], align 4
4402 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4403 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4404 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4405 // CHECK11-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4406 // CHECK11-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4407 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 91, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 [[TMP1]])
4408 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4409 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
4410 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4411 // CHECK11: cond.true:
4412 // CHECK11-NEXT: br label [[COND_END:%.*]]
4413 // CHECK11: cond.false:
4414 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4415 // CHECK11-NEXT: br label [[COND_END]]
4416 // CHECK11: cond.end:
4417 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4418 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4419 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4420 // CHECK11-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4421 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4422 // CHECK11: omp.inner.for.cond:
4423 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4424 // CHECK11-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP7]], 10
4425 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4426 // CHECK11: omp.inner.for.body:
4427 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4428 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4429 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4430 // CHECK11-NEXT: store i32 [[TMP10]], i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4431 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR__CASTED]], align 4
4432 // CHECK11-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [10 x i32]*, i32)* @.omp_outlined..19 to void (i32*, i32*, ...)*), i32 [[TMP8]], i32 [[TMP9]], [10 x i32]* [[TMP0]], i32 [[TMP11]])
4433 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4434 // CHECK11: omp.inner.for.inc:
4435 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4436 // CHECK11-NEXT: [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4437 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4438 // CHECK11-NEXT: store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4439 // CHECK11-NEXT: [[TMP14:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4440 // CHECK11-NEXT: [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4441 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
4442 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_COMB_LB]], align 4
4443 // CHECK11-NEXT: [[TMP16:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4444 // CHECK11-NEXT: [[TMP17:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4445 // CHECK11-NEXT: [[ADD3:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
4446 // CHECK11-NEXT: store i32 [[ADD3]], i32* [[DOTOMP_COMB_UB]], align 4
4447 // CHECK11-NEXT: [[TMP18:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4448 // CHECK11-NEXT: [[CMP4:%.*]] = icmp sgt i32 [[TMP18]], 9
4449 // CHECK11-NEXT: br i1 [[CMP4]], label [[COND_TRUE5:%.*]], label [[COND_FALSE6:%.*]]
4450 // CHECK11: cond.true5:
4451 // CHECK11-NEXT: br label [[COND_END7:%.*]]
4452 // CHECK11: cond.false6:
4453 // CHECK11-NEXT: [[TMP19:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4454 // CHECK11-NEXT: br label [[COND_END7]]
4455 // CHECK11: cond.end7:
4456 // CHECK11-NEXT: [[COND8:%.*]] = phi i32 [ 9, [[COND_TRUE5]] ], [ [[TMP19]], [[COND_FALSE6]] ]
4457 // CHECK11-NEXT: store i32 [[COND8]], i32* [[DOTOMP_COMB_UB]], align 4
4458 // CHECK11-NEXT: [[TMP20:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4459 // CHECK11-NEXT: store i32 [[TMP20]], i32* [[DOTOMP_IV]], align 4
4460 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
4461 // CHECK11: omp.inner.for.end:
4462 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4463 // CHECK11: omp.loop.exit:
4464 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4465 // CHECK11-NEXT: ret void
4466 //
4467 //
4468 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..19
4469 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [10 x i32]* noundef nonnull align 4 dereferenceable(40) [[A:%.*]], i32 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR2]] {
4470 // CHECK11-NEXT: entry:
4471 // CHECK11-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
4472 // CHECK11-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
4473 // CHECK11-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
4474 // CHECK11-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
4475 // CHECK11-NEXT: [[A_ADDR:%.*]] = alloca [10 x i32]*, align 4
4476 // CHECK11-NEXT: [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i32, align 4
4477 // CHECK11-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
4478 // CHECK11-NEXT: [[TMP:%.*]] = alloca i32, align 4
4479 // CHECK11-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
4480 // CHECK11-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
4481 // CHECK11-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4482 // CHECK11-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4483 // CHECK11-NEXT: [[I:%.*]] = alloca i32, align 4
4484 // CHECK11-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
4485 // CHECK11-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
4486 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4487 // CHECK11-NEXT: store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4488 // CHECK11-NEXT: store [10 x i32]* [[A]], [10 x i32]** [[A_ADDR]], align 4
4489 // CHECK11-NEXT: store i32 [[DOTCAPTURE_EXPR_]], i32* [[DOTCAPTURE_EXPR__ADDR]], align 4
4490 // CHECK11-NEXT: [[TMP0:%.*]] = load [10 x i32]*, [10 x i32]** [[A_ADDR]], align 4
4491 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
4492 // CHECK11-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
4493 // CHECK11-NEXT: [[TMP1:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
4494 // CHECK11-NEXT: [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
4495 // CHECK11-NEXT: store i32 [[TMP1]], i32* [[DOTOMP_LB]], align 4
4496 // CHECK11-NEXT: store i32 [[TMP2]], i32* [[DOTOMP_UB]], align 4
4497 // CHECK11-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4498 // CHECK11-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4499 // CHECK11-NEXT: [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
4500 // CHECK11-NEXT: [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
4501 // CHECK11-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4502 // CHECK11-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4503 // CHECK11-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 9
4504 // CHECK11-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4505 // CHECK11: cond.true:
4506 // CHECK11-NEXT: br label [[COND_END:%.*]]
4507 // CHECK11: cond.false:
4508 // CHECK11-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4509 // CHECK11-NEXT: br label [[COND_END]]
4510 // CHECK11: cond.end:
4511 // CHECK11-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
4512 // CHECK11-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4513 // CHECK11-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4514 // CHECK11-NEXT: store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
4515 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
4516 // CHECK11: omp.inner.for.cond:
4517 // CHECK11-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4518 // CHECK11-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4519 // CHECK11-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
4520 // CHECK11-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4521 // CHECK11: omp.inner.for.body:
4522 // CHECK11-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4523 // CHECK11-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
4524 // CHECK11-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4525 // CHECK11-NEXT: store i32 [[ADD]], i32* [[I]], align 4
4526 // CHECK11-NEXT: [[TMP11:%.*]] = load i32, i32* [[I]], align 4
4527 // CHECK11-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP0]], i32 0, i32 [[TMP11]]
4528 // CHECK11-NEXT: store i32 0, i32* [[ARRAYIDX]], align 4
4529 // CHECK11-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
4530 // CHECK11: omp.body.continue:
4531 // CHECK11-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
4532 // CHECK11: omp.inner.for.inc:
4533 // CHECK11-NEXT: [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4534 // CHECK11-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP12]], 1
4535 // CHECK11-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
4536 // CHECK11-NEXT: br label [[OMP_INNER_FOR_COND]]
4537 // CHECK11: omp.inner.for.end:
4538 // CHECK11-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
4539 // CHECK11: omp.loop.exit:
4540 // CHECK11-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP4]])
4541 // CHECK11-NEXT: ret void
4542 //
4543 //
4544 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4545 // CHECK11-SAME: () #[[ATTR6:[0-9]+]] {
4546 // CHECK11-NEXT: entry:
4547 // CHECK11-NEXT: call void @__tgt_register_requires(i64 1)
4548 // CHECK11-NEXT: ret void
4549 //
4550