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 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
4 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
5 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
6 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
7 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
8 
9 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK5
10 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
11 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK5
12 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK7
13 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
14 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK7
15 
16 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
17 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple aarch64-unknown-unknown -emit-pch -o %t %s
18 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
19 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK11
20 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple aarch64-unknown-unknown -emit-pch -o %t %s
21 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK11
22 
23 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK13
24 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple aarch64-unknown-unknown -emit-pch -o %t %s
25 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK13
26 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK15
27 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple aarch64-unknown-unknown -emit-pch -o %t %s
28 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp-simd -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple aarch64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK15
29 // expected-no-diagnostics
30 #ifndef HEADER
31 #define HEADER
32 
33 void fn1();
34 void fn2();
35 void fn3();
36 void fn4();
37 void fn5();
38 void fn6();
39 
40 int Arg;
41 
42 void gtid_test() {
43 #pragma omp target
44 #pragma omp teams
45 #pragma omp distribute parallel for simd
46   for(int i = 0 ; i < 100; i++) {}
47 
48 #pragma omp target
49 #pragma omp teams
50 #pragma omp distribute parallel for simd if (parallel: false)
51   for(int i = 0 ; i < 100; i++) {
52     gtid_test();
53   }
54 }
55 
56 
57 template <typename T>
58 int tmain(T Arg) {
59 #pragma omp target
60 #pragma omp teams
61 #pragma omp distribute parallel for simd if (true)
62   for(int i = 0 ; i < 100; i++) {
63     fn1();
64   }
65 #pragma omp target
66 #pragma omp teams
67 #pragma omp distribute parallel for simd if (false)
68   for(int i = 0 ; i < 100; i++) {
69     fn2();
70   }
71 #pragma omp target
72 #pragma omp teams
73 #pragma omp distribute parallel for simd if (parallel: Arg)
74   for(int i = 0 ; i < 100; i++) {
75     fn3();
76   }
77   return 0;
78 }
79 
80 int main() {
81 #pragma omp target
82 #pragma omp teams
83 #pragma omp distribute parallel for simd if (true)
84   for(int i = 0 ; i < 100; i++) {
85 
86 
87     fn4();
88   }
89 
90 #pragma omp target
91 #pragma omp teams
92 #pragma omp distribute parallel for simd if (false)
93   for(int i = 0 ; i < 100; i++) {
94 
95 
96     fn5();
97   }
98 
99 #pragma omp target
100 #pragma omp teams
101 #pragma omp distribute parallel for simd if (Arg)
102   for(int i = 0 ; i < 100; i++) {
103 
104 
105     fn6();
106   }
107 
108   return tmain(Arg);
109 }
110 
111 
112 
113 
114 
115 
116 // call void [[T_OUTLINE_FUN_3:@.+]](
117 
118 
119 #endif
120 // CHECK1-LABEL: define {{[^@]+}}@_Z9gtid_testv
121 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
122 // CHECK1-NEXT:  entry:
123 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
124 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
125 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
126 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
127 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
128 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
129 // CHECK1:       omp_offload.failed:
130 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43() #[[ATTR2:[0-9]+]]
131 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
132 // CHECK1:       omp_offload.cont:
133 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
134 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
135 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
136 // CHECK1-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
137 // CHECK1:       omp_offload.failed2:
138 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48() #[[ATTR2]]
139 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
140 // CHECK1:       omp_offload.cont3:
141 // CHECK1-NEXT:    ret void
142 //
143 //
144 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43
145 // CHECK1-SAME: () #[[ATTR1:[0-9]+]] {
146 // CHECK1-NEXT:  entry:
147 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
148 // CHECK1-NEXT:    ret void
149 //
150 //
151 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
152 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
153 // CHECK1-NEXT:  entry:
154 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
155 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
156 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
157 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
158 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
159 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
160 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
161 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
162 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
163 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
164 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
165 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
166 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
167 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
168 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
169 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
170 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
171 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
172 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
173 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
174 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
175 // CHECK1:       cond.true:
176 // CHECK1-NEXT:    br label [[COND_END:%.*]]
177 // CHECK1:       cond.false:
178 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
179 // CHECK1-NEXT:    br label [[COND_END]]
180 // CHECK1:       cond.end:
181 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
182 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
183 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
184 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
185 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
186 // CHECK1:       omp.inner.for.cond:
187 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
188 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !11
189 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
190 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
191 // CHECK1:       omp.inner.for.body:
192 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !11
193 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
194 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !11
195 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
196 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group !11
197 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
198 // CHECK1:       omp.inner.for.inc:
199 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
200 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !11
201 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
202 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
203 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
204 // CHECK1:       omp.inner.for.end:
205 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
206 // CHECK1:       omp.loop.exit:
207 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
208 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
209 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
210 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
211 // CHECK1:       .omp.final.then:
212 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
213 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
214 // CHECK1:       .omp.final.done:
215 // CHECK1-NEXT:    ret void
216 //
217 //
218 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
219 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
220 // CHECK1-NEXT:  entry:
221 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
222 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
223 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
224 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
225 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
226 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
227 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
228 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
229 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
230 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
231 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
232 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
233 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
234 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
235 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
236 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
237 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
238 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
239 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
240 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
241 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
242 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
243 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
244 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
245 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
246 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
247 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
248 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
249 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
250 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
251 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
252 // CHECK1:       cond.true:
253 // CHECK1-NEXT:    br label [[COND_END:%.*]]
254 // CHECK1:       cond.false:
255 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
256 // CHECK1-NEXT:    br label [[COND_END]]
257 // CHECK1:       cond.end:
258 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
259 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
260 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
261 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
262 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
263 // CHECK1:       omp.inner.for.cond:
264 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
265 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !15
266 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
267 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
268 // CHECK1:       omp.inner.for.body:
269 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
270 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
271 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
272 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !15
273 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
274 // CHECK1:       omp.body.continue:
275 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
276 // CHECK1:       omp.inner.for.inc:
277 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
278 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
279 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
280 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
281 // CHECK1:       omp.inner.for.end:
282 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
283 // CHECK1:       omp.loop.exit:
284 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
285 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
286 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
287 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
288 // CHECK1:       .omp.final.then:
289 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
290 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
291 // CHECK1:       .omp.final.done:
292 // CHECK1-NEXT:    ret void
293 //
294 //
295 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48
296 // CHECK1-SAME: () #[[ATTR1]] {
297 // CHECK1-NEXT:  entry:
298 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
299 // CHECK1-NEXT:    ret void
300 //
301 //
302 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
303 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
304 // CHECK1-NEXT:  entry:
305 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
306 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
307 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
308 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
309 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
310 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
311 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
312 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
313 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
314 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
315 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
316 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
317 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
318 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
319 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
320 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
321 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
322 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
323 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
324 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
325 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
326 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
327 // CHECK1:       cond.true:
328 // CHECK1-NEXT:    br label [[COND_END:%.*]]
329 // CHECK1:       cond.false:
330 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
331 // CHECK1-NEXT:    br label [[COND_END]]
332 // CHECK1:       cond.end:
333 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
334 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
335 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
336 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
337 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
338 // CHECK1:       omp.inner.for.cond:
339 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
340 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !20
341 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
342 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
343 // CHECK1:       omp.inner.for.body:
344 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !20
345 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
346 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !20
347 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
348 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !20
349 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !20
350 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !20
351 // CHECK1-NEXT:    call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !20
352 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !20
353 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
354 // CHECK1:       omp.inner.for.inc:
355 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
356 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !20
357 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
358 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
359 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
360 // CHECK1:       omp.inner.for.end:
361 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
362 // CHECK1:       omp.loop.exit:
363 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
364 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
365 // CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
366 // CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
367 // CHECK1:       .omp.final.then:
368 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
369 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
370 // CHECK1:       .omp.final.done:
371 // CHECK1-NEXT:    ret void
372 //
373 //
374 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
375 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
376 // CHECK1-NEXT:  entry:
377 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
378 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
379 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
380 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
381 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
382 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
383 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
384 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
385 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
386 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
387 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
388 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
389 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
390 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
391 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
392 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
393 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
394 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
395 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
396 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
397 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
398 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
399 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
400 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
401 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
402 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
403 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
404 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
405 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
406 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
407 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
408 // CHECK1:       cond.true:
409 // CHECK1-NEXT:    br label [[COND_END:%.*]]
410 // CHECK1:       cond.false:
411 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
412 // CHECK1-NEXT:    br label [[COND_END]]
413 // CHECK1:       cond.end:
414 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
415 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
416 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
417 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
418 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
419 // CHECK1:       omp.inner.for.cond:
420 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
421 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !23
422 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
423 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
424 // CHECK1:       omp.inner.for.body:
425 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
426 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
427 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
428 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !23
429 // CHECK1-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !23
430 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
431 // CHECK1:       omp.body.continue:
432 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
433 // CHECK1:       omp.inner.for.inc:
434 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
435 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
436 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
437 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
438 // CHECK1:       omp.inner.for.end:
439 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
440 // CHECK1:       omp.loop.exit:
441 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
442 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
443 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
444 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
445 // CHECK1:       .omp.final.then:
446 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
447 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
448 // CHECK1:       .omp.final.done:
449 // CHECK1-NEXT:    ret void
450 //
451 //
452 // CHECK1-LABEL: define {{[^@]+}}@main
453 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
454 // CHECK1-NEXT:  entry:
455 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
456 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
457 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
458 // CHECK1-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
459 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
460 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
461 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
462 // CHECK1-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
463 // CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
464 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
465 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
466 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
467 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
468 // CHECK1:       omp_offload.failed:
469 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81() #[[ATTR2]]
470 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
471 // CHECK1:       omp_offload.cont:
472 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
473 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
474 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
475 // CHECK1-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
476 // CHECK1:       omp_offload.failed2:
477 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90() #[[ATTR2]]
478 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
479 // CHECK1:       omp_offload.cont3:
480 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* @Arg, align 4
481 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
482 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
483 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
484 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
485 // CHECK1-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
486 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
487 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
488 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
489 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
490 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
491 // CHECK1-NEXT:    store i8* null, i8** [[TMP10]], align 8
492 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
493 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
494 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
495 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 1)
496 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
497 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
498 // CHECK1:       omp_offload.failed5:
499 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99(i64 [[TMP5]]) #[[ATTR2]]
500 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
501 // CHECK1:       omp_offload.cont6:
502 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
503 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
504 // CHECK1-NEXT:    ret i32 [[CALL]]
505 //
506 //
507 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
508 // CHECK1-SAME: () #[[ATTR1]] {
509 // CHECK1-NEXT:  entry:
510 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
511 // CHECK1-NEXT:    ret void
512 //
513 //
514 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
515 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
516 // CHECK1-NEXT:  entry:
517 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
518 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
519 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
520 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
521 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
522 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
523 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
524 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
525 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
526 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
527 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
528 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
529 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
530 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
531 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
532 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
533 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
534 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
535 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
536 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
537 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
538 // CHECK1:       cond.true:
539 // CHECK1-NEXT:    br label [[COND_END:%.*]]
540 // CHECK1:       cond.false:
541 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
542 // CHECK1-NEXT:    br label [[COND_END]]
543 // CHECK1:       cond.end:
544 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
545 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
546 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
547 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
548 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
549 // CHECK1:       omp.inner.for.cond:
550 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
551 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !26
552 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
553 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
554 // CHECK1:       omp.inner.for.body:
555 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !26
556 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
557 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !26
558 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
559 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group !26
560 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
561 // CHECK1:       omp.inner.for.inc:
562 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
563 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !26
564 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
565 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
566 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
567 // CHECK1:       omp.inner.for.end:
568 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
569 // CHECK1:       omp.loop.exit:
570 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
571 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
572 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
573 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
574 // CHECK1:       .omp.final.then:
575 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
576 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
577 // CHECK1:       .omp.final.done:
578 // CHECK1-NEXT:    ret void
579 //
580 //
581 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..5
582 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
583 // CHECK1-NEXT:  entry:
584 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
585 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
586 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
587 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
588 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
589 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
590 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
591 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
592 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
593 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
594 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
595 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
596 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
597 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
598 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
599 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
600 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
601 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
602 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
603 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
604 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
605 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
606 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
607 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
608 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
609 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
610 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
611 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
612 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
613 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
614 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
615 // CHECK1:       cond.true:
616 // CHECK1-NEXT:    br label [[COND_END:%.*]]
617 // CHECK1:       cond.false:
618 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
619 // CHECK1-NEXT:    br label [[COND_END]]
620 // CHECK1:       cond.end:
621 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
622 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
623 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
624 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
625 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
626 // CHECK1:       omp.inner.for.cond:
627 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
628 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !29
629 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
630 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
631 // CHECK1:       omp.inner.for.body:
632 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
633 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
634 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
635 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !29
636 // CHECK1-NEXT:    call void @_Z3fn4v(), !llvm.access.group !29
637 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
638 // CHECK1:       omp.body.continue:
639 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
640 // CHECK1:       omp.inner.for.inc:
641 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
642 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
643 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
644 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
645 // CHECK1:       omp.inner.for.end:
646 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
647 // CHECK1:       omp.loop.exit:
648 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
649 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
650 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
651 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
652 // CHECK1:       .omp.final.then:
653 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
654 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
655 // CHECK1:       .omp.final.done:
656 // CHECK1-NEXT:    ret void
657 //
658 //
659 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90
660 // CHECK1-SAME: () #[[ATTR1]] {
661 // CHECK1-NEXT:  entry:
662 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..6 to void (i32*, i32*, ...)*))
663 // CHECK1-NEXT:    ret void
664 //
665 //
666 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
667 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
668 // CHECK1-NEXT:  entry:
669 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
670 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
671 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
672 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
673 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
674 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
675 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
676 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
677 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
678 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
679 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
680 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
681 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
682 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
683 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
684 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
685 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
686 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
687 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
688 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
689 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
690 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
691 // CHECK1:       cond.true:
692 // CHECK1-NEXT:    br label [[COND_END:%.*]]
693 // CHECK1:       cond.false:
694 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
695 // CHECK1-NEXT:    br label [[COND_END]]
696 // CHECK1:       cond.end:
697 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
698 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
699 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
700 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
701 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
702 // CHECK1:       omp.inner.for.cond:
703 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
704 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !32
705 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
706 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
707 // CHECK1:       omp.inner.for.body:
708 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !32
709 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
710 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !32
711 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
712 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !32
713 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !32
714 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !32
715 // CHECK1-NEXT:    call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !32
716 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !32
717 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
718 // CHECK1:       omp.inner.for.inc:
719 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
720 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !32
721 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
722 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
723 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
724 // CHECK1:       omp.inner.for.end:
725 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
726 // CHECK1:       omp.loop.exit:
727 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
728 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
729 // CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
730 // CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
731 // CHECK1:       .omp.final.then:
732 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
733 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
734 // CHECK1:       .omp.final.done:
735 // CHECK1-NEXT:    ret void
736 //
737 //
738 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
739 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
740 // CHECK1-NEXT:  entry:
741 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
742 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
743 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
744 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
745 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
746 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
747 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
748 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
749 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
750 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
751 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
752 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
753 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
754 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
755 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
756 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
757 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
758 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
759 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
760 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
761 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
762 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
763 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
764 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
765 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
766 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
767 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
768 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
769 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
770 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
771 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
772 // CHECK1:       cond.true:
773 // CHECK1-NEXT:    br label [[COND_END:%.*]]
774 // CHECK1:       cond.false:
775 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
776 // CHECK1-NEXT:    br label [[COND_END]]
777 // CHECK1:       cond.end:
778 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
779 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
780 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
781 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
782 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
783 // CHECK1:       omp.inner.for.cond:
784 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
785 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !35
786 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
787 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
788 // CHECK1:       omp.inner.for.body:
789 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
790 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
791 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
792 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !35
793 // CHECK1-NEXT:    call void @_Z3fn5v(), !llvm.access.group !35
794 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
795 // CHECK1:       omp.body.continue:
796 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
797 // CHECK1:       omp.inner.for.inc:
798 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
799 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
800 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
801 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
802 // CHECK1:       omp.inner.for.end:
803 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
804 // CHECK1:       omp.loop.exit:
805 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
806 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
807 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
808 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
809 // CHECK1:       .omp.final.then:
810 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
811 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
812 // CHECK1:       .omp.final.done:
813 // CHECK1-NEXT:    ret void
814 //
815 //
816 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99
817 // CHECK1-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
818 // CHECK1-NEXT:  entry:
819 // CHECK1-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
820 // CHECK1-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
821 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
822 // 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*, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]])
823 // CHECK1-NEXT:    ret void
824 //
825 //
826 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..8
827 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
828 // CHECK1-NEXT:  entry:
829 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
830 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
831 // CHECK1-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
832 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
833 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
834 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
835 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
836 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
837 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
838 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
839 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
840 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
841 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
842 // CHECK1-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
843 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
844 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
845 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
846 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
847 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
848 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
849 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
850 // 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)
851 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
852 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
853 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
854 // CHECK1:       cond.true:
855 // CHECK1-NEXT:    br label [[COND_END:%.*]]
856 // CHECK1:       cond.false:
857 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
858 // CHECK1-NEXT:    br label [[COND_END]]
859 // CHECK1:       cond.end:
860 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
861 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
862 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
863 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
864 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
865 // CHECK1:       omp.inner.for.cond:
866 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
867 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !38
868 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
869 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
870 // CHECK1:       omp.inner.for.body:
871 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !38
872 // CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
873 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !38
874 // CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
875 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !38
876 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
877 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
878 // CHECK1:       omp_if.then:
879 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]]), !llvm.access.group !38
880 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
881 // CHECK1:       omp_if.else:
882 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !38
883 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !38
884 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !38
885 // CHECK1-NEXT:    call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !38
886 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !38
887 // CHECK1-NEXT:    br label [[OMP_IF_END]]
888 // CHECK1:       omp_if.end:
889 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
890 // CHECK1:       omp.inner.for.inc:
891 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
892 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !38
893 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
894 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
895 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
896 // CHECK1:       omp.inner.for.end:
897 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
898 // CHECK1:       omp.loop.exit:
899 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
900 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
901 // CHECK1-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
902 // CHECK1-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
903 // CHECK1:       .omp.final.then:
904 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
905 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
906 // CHECK1:       .omp.final.done:
907 // CHECK1-NEXT:    ret void
908 //
909 //
910 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..9
911 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
912 // CHECK1-NEXT:  entry:
913 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
914 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
915 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
916 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
917 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
918 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
919 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
920 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
921 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
922 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
923 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
924 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
925 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
926 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
927 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
928 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
929 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
930 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
931 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
932 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
933 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
934 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
935 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
936 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
937 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
938 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
939 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
940 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
941 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
942 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
943 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
944 // CHECK1:       cond.true:
945 // CHECK1-NEXT:    br label [[COND_END:%.*]]
946 // CHECK1:       cond.false:
947 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
948 // CHECK1-NEXT:    br label [[COND_END]]
949 // CHECK1:       cond.end:
950 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
951 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
952 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
953 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
954 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
955 // CHECK1:       omp.inner.for.cond:
956 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
957 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !41
958 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
959 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
960 // CHECK1:       omp.inner.for.body:
961 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
962 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
963 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
964 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !41
965 // CHECK1-NEXT:    call void @_Z3fn6v(), !llvm.access.group !41
966 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
967 // CHECK1:       omp.body.continue:
968 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
969 // CHECK1:       omp.inner.for.inc:
970 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
971 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
972 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
973 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP42:![0-9]+]]
974 // CHECK1:       omp.inner.for.end:
975 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
976 // CHECK1:       omp.loop.exit:
977 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
978 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
979 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
980 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
981 // CHECK1:       .omp.final.then:
982 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
983 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
984 // CHECK1:       .omp.final.done:
985 // CHECK1-NEXT:    ret void
986 //
987 //
988 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
989 // CHECK1-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
990 // CHECK1-NEXT:  entry:
991 // CHECK1-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
992 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
993 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
994 // CHECK1-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
995 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
996 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
997 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
998 // CHECK1-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
999 // CHECK1-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
1000 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1001 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
1002 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
1003 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1004 // CHECK1:       omp_offload.failed:
1005 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59() #[[ATTR2]]
1006 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1007 // CHECK1:       omp_offload.cont:
1008 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1009 // CHECK1-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
1010 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
1011 // CHECK1-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
1012 // CHECK1:       omp_offload.failed2:
1013 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65() #[[ATTR2]]
1014 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
1015 // CHECK1:       omp_offload.cont3:
1016 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
1017 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
1018 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
1019 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
1020 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1021 // CHECK1-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
1022 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
1023 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1024 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
1025 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
1026 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1027 // CHECK1-NEXT:    store i8* null, i8** [[TMP10]], align 8
1028 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1029 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1030 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1031 // CHECK1-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 1)
1032 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1033 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
1034 // CHECK1:       omp_offload.failed5:
1035 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71(i64 [[TMP5]]) #[[ATTR2]]
1036 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
1037 // CHECK1:       omp_offload.cont6:
1038 // CHECK1-NEXT:    ret i32 0
1039 //
1040 //
1041 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59
1042 // CHECK1-SAME: () #[[ATTR1]] {
1043 // CHECK1-NEXT:  entry:
1044 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*))
1045 // CHECK1-NEXT:    ret void
1046 //
1047 //
1048 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..10
1049 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1050 // CHECK1-NEXT:  entry:
1051 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1052 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1053 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1054 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1055 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1056 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1057 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1058 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1059 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1060 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1061 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1062 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1063 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1064 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1065 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1066 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1067 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1068 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1069 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1070 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1071 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1072 // CHECK1:       cond.true:
1073 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1074 // CHECK1:       cond.false:
1075 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1076 // CHECK1-NEXT:    br label [[COND_END]]
1077 // CHECK1:       cond.end:
1078 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1079 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1080 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1081 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1082 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1083 // CHECK1:       omp.inner.for.cond:
1084 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
1085 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !44
1086 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1087 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1088 // CHECK1:       omp.inner.for.body:
1089 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !44
1090 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1091 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !44
1092 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1093 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group !44
1094 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1095 // CHECK1:       omp.inner.for.inc:
1096 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
1097 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !44
1098 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1099 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
1100 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP45:![0-9]+]]
1101 // CHECK1:       omp.inner.for.end:
1102 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1103 // CHECK1:       omp.loop.exit:
1104 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1105 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1106 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1107 // CHECK1-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1108 // CHECK1:       .omp.final.then:
1109 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
1110 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1111 // CHECK1:       .omp.final.done:
1112 // CHECK1-NEXT:    ret void
1113 //
1114 //
1115 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
1116 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1117 // CHECK1-NEXT:  entry:
1118 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1119 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1120 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1121 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1122 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1123 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1124 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1125 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1126 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1127 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1128 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1129 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1130 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1131 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1132 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1133 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1134 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
1135 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1136 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1137 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1138 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1139 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1140 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1141 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1142 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1143 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1144 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1145 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1146 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1147 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1148 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1149 // CHECK1:       cond.true:
1150 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1151 // CHECK1:       cond.false:
1152 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1153 // CHECK1-NEXT:    br label [[COND_END]]
1154 // CHECK1:       cond.end:
1155 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1156 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1157 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1158 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1159 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1160 // CHECK1:       omp.inner.for.cond:
1161 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
1162 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !47
1163 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1164 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1165 // CHECK1:       omp.inner.for.body:
1166 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
1167 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1168 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1169 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !47
1170 // CHECK1-NEXT:    call void @_Z3fn1v(), !llvm.access.group !47
1171 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1172 // CHECK1:       omp.body.continue:
1173 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1174 // CHECK1:       omp.inner.for.inc:
1175 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
1176 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1177 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
1178 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
1179 // CHECK1:       omp.inner.for.end:
1180 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1181 // CHECK1:       omp.loop.exit:
1182 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1183 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1184 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1185 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1186 // CHECK1:       .omp.final.then:
1187 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
1188 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1189 // CHECK1:       .omp.final.done:
1190 // CHECK1-NEXT:    ret void
1191 //
1192 //
1193 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65
1194 // CHECK1-SAME: () #[[ATTR1]] {
1195 // CHECK1-NEXT:  entry:
1196 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..12 to void (i32*, i32*, ...)*))
1197 // CHECK1-NEXT:    ret void
1198 //
1199 //
1200 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..12
1201 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1202 // CHECK1-NEXT:  entry:
1203 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1204 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1205 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1206 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1207 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1208 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1209 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1210 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1211 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1212 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1213 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1214 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1215 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1216 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1217 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1218 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1219 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1220 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1221 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1222 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1223 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1224 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1225 // CHECK1:       cond.true:
1226 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1227 // CHECK1:       cond.false:
1228 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1229 // CHECK1-NEXT:    br label [[COND_END]]
1230 // CHECK1:       cond.end:
1231 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1232 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1233 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1234 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1235 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1236 // CHECK1:       omp.inner.for.cond:
1237 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
1238 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !50
1239 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1240 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1241 // CHECK1:       omp.inner.for.body:
1242 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !50
1243 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1244 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !50
1245 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1246 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !50
1247 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !50
1248 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !50
1249 // CHECK1-NEXT:    call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !50
1250 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !50
1251 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1252 // CHECK1:       omp.inner.for.inc:
1253 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
1254 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !50
1255 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1256 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
1257 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
1258 // CHECK1:       omp.inner.for.end:
1259 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1260 // CHECK1:       omp.loop.exit:
1261 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1262 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1263 // CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
1264 // CHECK1-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1265 // CHECK1:       .omp.final.then:
1266 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
1267 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1268 // CHECK1:       .omp.final.done:
1269 // CHECK1-NEXT:    ret void
1270 //
1271 //
1272 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..13
1273 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1274 // CHECK1-NEXT:  entry:
1275 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1276 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1277 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1278 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1279 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1280 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1281 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1282 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1283 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1284 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1285 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1286 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1287 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1288 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1289 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1290 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1291 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
1292 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1293 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1294 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1295 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1296 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1297 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1298 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1299 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1300 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1301 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1302 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1303 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1304 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1305 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1306 // CHECK1:       cond.true:
1307 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1308 // CHECK1:       cond.false:
1309 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1310 // CHECK1-NEXT:    br label [[COND_END]]
1311 // CHECK1:       cond.end:
1312 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1313 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1314 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1315 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1316 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1317 // CHECK1:       omp.inner.for.cond:
1318 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
1319 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !53
1320 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1321 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1322 // CHECK1:       omp.inner.for.body:
1323 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
1324 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1325 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1326 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !53
1327 // CHECK1-NEXT:    call void @_Z3fn2v(), !llvm.access.group !53
1328 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1329 // CHECK1:       omp.body.continue:
1330 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1331 // CHECK1:       omp.inner.for.inc:
1332 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
1333 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1334 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
1335 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP54:![0-9]+]]
1336 // CHECK1:       omp.inner.for.end:
1337 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1338 // CHECK1:       omp.loop.exit:
1339 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1340 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1341 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1342 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1343 // CHECK1:       .omp.final.then:
1344 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
1345 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1346 // CHECK1:       .omp.final.done:
1347 // CHECK1-NEXT:    ret void
1348 //
1349 //
1350 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71
1351 // CHECK1-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
1352 // CHECK1-NEXT:  entry:
1353 // CHECK1-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
1354 // CHECK1-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
1355 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
1356 // 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*, i32*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[CONV]])
1357 // CHECK1-NEXT:    ret void
1358 //
1359 //
1360 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..14
1361 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
1362 // CHECK1-NEXT:  entry:
1363 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1364 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1365 // CHECK1-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
1366 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1367 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1368 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1369 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1370 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1371 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1372 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1373 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1374 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1375 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1376 // CHECK1-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
1377 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
1378 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1379 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1380 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1381 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1382 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1383 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1384 // 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)
1385 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1386 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
1387 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1388 // CHECK1:       cond.true:
1389 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1390 // CHECK1:       cond.false:
1391 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1392 // CHECK1-NEXT:    br label [[COND_END]]
1393 // CHECK1:       cond.end:
1394 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1395 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1396 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1397 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1398 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1399 // CHECK1:       omp.inner.for.cond:
1400 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
1401 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !56
1402 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1403 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1404 // CHECK1:       omp.inner.for.body:
1405 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !56
1406 // CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
1407 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !56
1408 // CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
1409 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !56
1410 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
1411 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1412 // CHECK1:       omp_if.then:
1413 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]]), !llvm.access.group !56
1414 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1415 // CHECK1:       omp_if.else:
1416 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !56
1417 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !56
1418 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !56
1419 // CHECK1-NEXT:    call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !56
1420 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !56
1421 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1422 // CHECK1:       omp_if.end:
1423 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1424 // CHECK1:       omp.inner.for.inc:
1425 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
1426 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !56
1427 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
1428 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
1429 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP57:![0-9]+]]
1430 // CHECK1:       omp.inner.for.end:
1431 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1432 // CHECK1:       omp.loop.exit:
1433 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1434 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1435 // CHECK1-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
1436 // CHECK1-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1437 // CHECK1:       .omp.final.then:
1438 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
1439 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1440 // CHECK1:       .omp.final.done:
1441 // CHECK1-NEXT:    ret void
1442 //
1443 //
1444 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..15
1445 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1446 // CHECK1-NEXT:  entry:
1447 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1448 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1449 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1450 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1451 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1452 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1453 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1454 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1455 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1456 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1457 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1458 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1459 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1460 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1461 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1462 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1463 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
1464 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1465 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1466 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1467 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1468 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1469 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1470 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1471 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1472 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1473 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1474 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1475 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1476 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1477 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1478 // CHECK1:       cond.true:
1479 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1480 // CHECK1:       cond.false:
1481 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1482 // CHECK1-NEXT:    br label [[COND_END]]
1483 // CHECK1:       cond.end:
1484 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1485 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1486 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1487 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1488 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1489 // CHECK1:       omp.inner.for.cond:
1490 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
1491 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !59
1492 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1493 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1494 // CHECK1:       omp.inner.for.body:
1495 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
1496 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1497 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1498 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !59
1499 // CHECK1-NEXT:    call void @_Z3fn3v(), !llvm.access.group !59
1500 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1501 // CHECK1:       omp.body.continue:
1502 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1503 // CHECK1:       omp.inner.for.inc:
1504 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
1505 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1506 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
1507 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP60:![0-9]+]]
1508 // CHECK1:       omp.inner.for.end:
1509 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1510 // CHECK1:       omp.loop.exit:
1511 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1512 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1513 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1514 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1515 // CHECK1:       .omp.final.then:
1516 // CHECK1-NEXT:    store i32 100, i32* [[I]], align 4
1517 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1518 // CHECK1:       .omp.final.done:
1519 // CHECK1-NEXT:    ret void
1520 //
1521 //
1522 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1523 // CHECK1-SAME: () #[[ATTR5:[0-9]+]] {
1524 // CHECK1-NEXT:  entry:
1525 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
1526 // CHECK1-NEXT:    ret void
1527 //
1528 //
1529 // CHECK3-LABEL: define {{[^@]+}}@_Z9gtid_testv
1530 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
1531 // CHECK3-NEXT:  entry:
1532 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1533 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1534 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
1535 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
1536 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
1537 // CHECK3-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1538 // CHECK3:       omp_offload.failed:
1539 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43() #[[ATTR2:[0-9]+]]
1540 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1541 // CHECK3:       omp_offload.cont:
1542 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1543 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
1544 // CHECK3-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
1545 // CHECK3-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
1546 // CHECK3:       omp_offload.failed2:
1547 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48() #[[ATTR2]]
1548 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
1549 // CHECK3:       omp_offload.cont3:
1550 // CHECK3-NEXT:    ret void
1551 //
1552 //
1553 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43
1554 // CHECK3-SAME: () #[[ATTR1:[0-9]+]] {
1555 // CHECK3-NEXT:  entry:
1556 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
1557 // CHECK3-NEXT:    ret void
1558 //
1559 //
1560 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
1561 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1562 // CHECK3-NEXT:  entry:
1563 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1564 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1565 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1566 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1567 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1568 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1569 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1570 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1571 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1572 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1573 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1574 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1575 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1576 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1577 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1578 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1579 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1580 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1581 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1582 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1583 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1584 // CHECK3:       cond.true:
1585 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1586 // CHECK3:       cond.false:
1587 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1588 // CHECK3-NEXT:    br label [[COND_END]]
1589 // CHECK3:       cond.end:
1590 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1591 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1592 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1593 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1594 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1595 // CHECK3:       omp.inner.for.cond:
1596 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1597 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !11
1598 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1599 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1600 // CHECK3:       omp.inner.for.body:
1601 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !11
1602 // CHECK3-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1603 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !11
1604 // CHECK3-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1605 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group !11
1606 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1607 // CHECK3:       omp.inner.for.inc:
1608 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1609 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !11
1610 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1611 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1612 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
1613 // CHECK3:       omp.inner.for.end:
1614 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1615 // CHECK3:       omp.loop.exit:
1616 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1617 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1618 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1619 // CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1620 // CHECK3:       .omp.final.then:
1621 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
1622 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1623 // CHECK3:       .omp.final.done:
1624 // CHECK3-NEXT:    ret void
1625 //
1626 //
1627 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
1628 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1629 // CHECK3-NEXT:  entry:
1630 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1631 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1632 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1633 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1634 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1635 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1636 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1637 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1638 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1639 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1640 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1641 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1642 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1643 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1644 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1645 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1646 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
1647 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1648 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1649 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1650 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1651 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1652 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1653 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1654 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1655 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1656 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1657 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1658 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1659 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1660 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1661 // CHECK3:       cond.true:
1662 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1663 // CHECK3:       cond.false:
1664 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1665 // CHECK3-NEXT:    br label [[COND_END]]
1666 // CHECK3:       cond.end:
1667 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1668 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1669 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1670 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1671 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1672 // CHECK3:       omp.inner.for.cond:
1673 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
1674 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !15
1675 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1676 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1677 // CHECK3:       omp.inner.for.body:
1678 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
1679 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1680 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1681 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !15
1682 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1683 // CHECK3:       omp.body.continue:
1684 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1685 // CHECK3:       omp.inner.for.inc:
1686 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
1687 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1688 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
1689 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
1690 // CHECK3:       omp.inner.for.end:
1691 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1692 // CHECK3:       omp.loop.exit:
1693 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1694 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1695 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1696 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1697 // CHECK3:       .omp.final.then:
1698 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
1699 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1700 // CHECK3:       .omp.final.done:
1701 // CHECK3-NEXT:    ret void
1702 //
1703 //
1704 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48
1705 // CHECK3-SAME: () #[[ATTR1]] {
1706 // CHECK3-NEXT:  entry:
1707 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
1708 // CHECK3-NEXT:    ret void
1709 //
1710 //
1711 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
1712 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1713 // CHECK3-NEXT:  entry:
1714 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1715 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1716 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1717 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1718 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1719 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1720 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1721 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1722 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1723 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1724 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1725 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1726 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1727 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1728 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1729 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1730 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1731 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1732 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1733 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1734 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1735 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1736 // CHECK3:       cond.true:
1737 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1738 // CHECK3:       cond.false:
1739 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1740 // CHECK3-NEXT:    br label [[COND_END]]
1741 // CHECK3:       cond.end:
1742 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1743 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1744 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1745 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1746 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1747 // CHECK3:       omp.inner.for.cond:
1748 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
1749 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !20
1750 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1751 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1752 // CHECK3:       omp.inner.for.body:
1753 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !20
1754 // CHECK3-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1755 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !20
1756 // CHECK3-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1757 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !20
1758 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !20
1759 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !20
1760 // CHECK3-NEXT:    call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !20
1761 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !20
1762 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1763 // CHECK3:       omp.inner.for.inc:
1764 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
1765 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !20
1766 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1767 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
1768 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
1769 // CHECK3:       omp.inner.for.end:
1770 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1771 // CHECK3:       omp.loop.exit:
1772 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1773 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1774 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
1775 // CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1776 // CHECK3:       .omp.final.then:
1777 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
1778 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1779 // CHECK3:       .omp.final.done:
1780 // CHECK3-NEXT:    ret void
1781 //
1782 //
1783 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
1784 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1785 // CHECK3-NEXT:  entry:
1786 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1787 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1788 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1789 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1790 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1791 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1792 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1793 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1794 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1795 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1796 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1797 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1798 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1799 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1800 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1801 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1802 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
1803 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1804 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1805 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1806 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1807 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1808 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1809 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1810 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1811 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1812 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1813 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1814 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1815 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1816 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1817 // CHECK3:       cond.true:
1818 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1819 // CHECK3:       cond.false:
1820 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1821 // CHECK3-NEXT:    br label [[COND_END]]
1822 // CHECK3:       cond.end:
1823 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1824 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1825 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1826 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1827 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1828 // CHECK3:       omp.inner.for.cond:
1829 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
1830 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !23
1831 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1832 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1833 // CHECK3:       omp.inner.for.body:
1834 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
1835 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1836 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1837 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !23
1838 // CHECK3-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !23
1839 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1840 // CHECK3:       omp.body.continue:
1841 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1842 // CHECK3:       omp.inner.for.inc:
1843 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
1844 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1845 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
1846 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
1847 // CHECK3:       omp.inner.for.end:
1848 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1849 // CHECK3:       omp.loop.exit:
1850 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1851 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1852 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1853 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1854 // CHECK3:       .omp.final.then:
1855 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
1856 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1857 // CHECK3:       .omp.final.done:
1858 // CHECK3-NEXT:    ret void
1859 //
1860 //
1861 // CHECK3-LABEL: define {{[^@]+}}@main
1862 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
1863 // CHECK3-NEXT:  entry:
1864 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1865 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1866 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1867 // CHECK3-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
1868 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
1869 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
1870 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
1871 // CHECK3-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
1872 // CHECK3-NEXT:    store i32 0, i32* [[RETVAL]], align 4
1873 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1874 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
1875 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
1876 // CHECK3-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1877 // CHECK3:       omp_offload.failed:
1878 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81() #[[ATTR2]]
1879 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1880 // CHECK3:       omp_offload.cont:
1881 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1882 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
1883 // CHECK3-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
1884 // CHECK3-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
1885 // CHECK3:       omp_offload.failed2:
1886 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90() #[[ATTR2]]
1887 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
1888 // CHECK3:       omp_offload.cont3:
1889 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* @Arg, align 4
1890 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
1891 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
1892 // CHECK3-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
1893 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1894 // CHECK3-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
1895 // CHECK3-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
1896 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1897 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
1898 // CHECK3-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
1899 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1900 // CHECK3-NEXT:    store i8* null, i8** [[TMP10]], align 8
1901 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1902 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1903 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1904 // CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 1)
1905 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1906 // CHECK3-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
1907 // CHECK3:       omp_offload.failed5:
1908 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99(i64 [[TMP5]]) #[[ATTR2]]
1909 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
1910 // CHECK3:       omp_offload.cont6:
1911 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
1912 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
1913 // CHECK3-NEXT:    ret i32 [[CALL]]
1914 //
1915 //
1916 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
1917 // CHECK3-SAME: () #[[ATTR1]] {
1918 // CHECK3-NEXT:  entry:
1919 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
1920 // CHECK3-NEXT:    ret void
1921 //
1922 //
1923 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
1924 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1925 // CHECK3-NEXT:  entry:
1926 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1927 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1928 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1929 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1930 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1931 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1932 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1933 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1934 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
1935 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1936 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1937 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1938 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1939 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1940 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1941 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1942 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1943 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1944 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1945 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1946 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1947 // CHECK3:       cond.true:
1948 // CHECK3-NEXT:    br label [[COND_END:%.*]]
1949 // CHECK3:       cond.false:
1950 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1951 // CHECK3-NEXT:    br label [[COND_END]]
1952 // CHECK3:       cond.end:
1953 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1954 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1955 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1956 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1957 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1958 // CHECK3:       omp.inner.for.cond:
1959 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
1960 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !26
1961 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1962 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1963 // CHECK3:       omp.inner.for.body:
1964 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !26
1965 // CHECK3-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1966 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !26
1967 // CHECK3-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1968 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group !26
1969 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1970 // CHECK3:       omp.inner.for.inc:
1971 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
1972 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !26
1973 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1974 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
1975 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
1976 // CHECK3:       omp.inner.for.end:
1977 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1978 // CHECK3:       omp.loop.exit:
1979 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1980 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1981 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1982 // CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1983 // CHECK3:       .omp.final.then:
1984 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
1985 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1986 // CHECK3:       .omp.final.done:
1987 // CHECK3-NEXT:    ret void
1988 //
1989 //
1990 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..5
1991 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1992 // CHECK3-NEXT:  entry:
1993 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1994 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1995 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1996 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1997 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1998 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1999 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2000 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2001 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2002 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2003 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2004 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2005 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2006 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2007 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2008 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2009 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
2010 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2011 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2012 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2013 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2014 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2015 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2016 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2017 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2018 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2019 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2020 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2021 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2022 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2023 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2024 // CHECK3:       cond.true:
2025 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2026 // CHECK3:       cond.false:
2027 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2028 // CHECK3-NEXT:    br label [[COND_END]]
2029 // CHECK3:       cond.end:
2030 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2031 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2032 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2033 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2034 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2035 // CHECK3:       omp.inner.for.cond:
2036 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
2037 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !29
2038 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2039 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2040 // CHECK3:       omp.inner.for.body:
2041 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
2042 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2043 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2044 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !29
2045 // CHECK3-NEXT:    call void @_Z3fn4v(), !llvm.access.group !29
2046 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2047 // CHECK3:       omp.body.continue:
2048 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2049 // CHECK3:       omp.inner.for.inc:
2050 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
2051 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2052 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
2053 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
2054 // CHECK3:       omp.inner.for.end:
2055 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2056 // CHECK3:       omp.loop.exit:
2057 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2058 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2059 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2060 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2061 // CHECK3:       .omp.final.then:
2062 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
2063 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2064 // CHECK3:       .omp.final.done:
2065 // CHECK3-NEXT:    ret void
2066 //
2067 //
2068 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90
2069 // CHECK3-SAME: () #[[ATTR1]] {
2070 // CHECK3-NEXT:  entry:
2071 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..6 to void (i32*, i32*, ...)*))
2072 // CHECK3-NEXT:    ret void
2073 //
2074 //
2075 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..6
2076 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2077 // CHECK3-NEXT:  entry:
2078 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2079 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2080 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2081 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2082 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2083 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2084 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2085 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2086 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2087 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
2088 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2089 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2090 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2091 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2092 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2093 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2094 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2095 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2096 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2097 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2098 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2099 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2100 // CHECK3:       cond.true:
2101 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2102 // CHECK3:       cond.false:
2103 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2104 // CHECK3-NEXT:    br label [[COND_END]]
2105 // CHECK3:       cond.end:
2106 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2107 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2108 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2109 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2110 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2111 // CHECK3:       omp.inner.for.cond:
2112 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2113 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2114 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2115 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2116 // CHECK3:       omp.inner.for.body:
2117 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2118 // CHECK3-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2119 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2120 // CHECK3-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2121 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
2122 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2123 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
2124 // CHECK3-NEXT:    call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
2125 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
2126 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2127 // CHECK3:       omp.inner.for.inc:
2128 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2129 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2130 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2131 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2132 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
2133 // CHECK3:       omp.inner.for.end:
2134 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2135 // CHECK3:       omp.loop.exit:
2136 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2137 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2138 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
2139 // CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2140 // CHECK3:       .omp.final.then:
2141 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
2142 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2143 // CHECK3:       .omp.final.done:
2144 // CHECK3-NEXT:    ret void
2145 //
2146 //
2147 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
2148 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2149 // CHECK3-NEXT:  entry:
2150 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2151 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2152 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2153 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2154 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2155 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2156 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2157 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2158 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2159 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2160 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2161 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2162 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2163 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2164 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2165 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2166 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
2167 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2168 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2169 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2170 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2171 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2172 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2173 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2174 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2175 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2176 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2177 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2178 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2179 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2180 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2181 // CHECK3:       cond.true:
2182 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2183 // CHECK3:       cond.false:
2184 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2185 // CHECK3-NEXT:    br label [[COND_END]]
2186 // CHECK3:       cond.end:
2187 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2188 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2189 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2190 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2191 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2192 // CHECK3:       omp.inner.for.cond:
2193 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2194 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2195 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2196 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2197 // CHECK3:       omp.inner.for.body:
2198 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2199 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2200 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2201 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
2202 // CHECK3-NEXT:    call void @_Z3fn5v()
2203 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2204 // CHECK3:       omp.body.continue:
2205 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2206 // CHECK3:       omp.inner.for.inc:
2207 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2208 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2209 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
2210 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
2211 // CHECK3:       omp.inner.for.end:
2212 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2213 // CHECK3:       omp.loop.exit:
2214 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2215 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2216 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2217 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2218 // CHECK3:       .omp.final.then:
2219 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
2220 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2221 // CHECK3:       .omp.final.done:
2222 // CHECK3-NEXT:    ret void
2223 //
2224 //
2225 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99
2226 // CHECK3-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
2227 // CHECK3-NEXT:  entry:
2228 // CHECK3-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
2229 // CHECK3-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
2230 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
2231 // 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*, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]])
2232 // CHECK3-NEXT:    ret void
2233 //
2234 //
2235 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..8
2236 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
2237 // CHECK3-NEXT:  entry:
2238 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2239 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2240 // CHECK3-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
2241 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
2242 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2243 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2244 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2245 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2246 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2247 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2248 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2249 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
2250 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
2251 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED12:%.*]] = alloca i64, align 8
2252 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR18:%.*]] = alloca i32, align 4
2253 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2254 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2255 // CHECK3-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
2256 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
2257 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2258 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
2259 // CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
2260 // CHECK3-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
2261 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2262 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2263 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2264 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2265 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2266 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2267 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2268 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2269 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2270 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2271 // CHECK3:       cond.true:
2272 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2273 // CHECK3:       cond.false:
2274 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2275 // CHECK3-NEXT:    br label [[COND_END]]
2276 // CHECK3:       cond.end:
2277 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2278 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2279 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2280 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2281 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
2282 // CHECK3-NEXT:    [[TOBOOL1:%.*]] = trunc i8 [[TMP7]] to i1
2283 // CHECK3-NEXT:    br i1 [[TOBOOL1]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE7:%.*]]
2284 // CHECK3:       omp_if.then:
2285 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2286 // CHECK3:       omp.inner.for.cond:
2287 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
2288 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
2289 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2290 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2291 // CHECK3:       omp.inner.for.body:
2292 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
2293 // CHECK3-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
2294 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
2295 // CHECK3-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
2296 // CHECK3-NEXT:    [[TMP14:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !35
2297 // CHECK3-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP14]] to i1
2298 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
2299 // CHECK3-NEXT:    [[FROMBOOL4:%.*]] = zext i1 [[TOBOOL3]] to i8
2300 // CHECK3-NEXT:    store i8 [[FROMBOOL4]], i8* [[CONV]], align 1, !llvm.access.group !35
2301 // CHECK3-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !35
2302 // CHECK3-NEXT:    [[TMP16:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !35
2303 // CHECK3-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP16]] to i1
2304 // CHECK3-NEXT:    br i1 [[TOBOOL5]], label [[OMP_IF_THEN6:%.*]], label [[OMP_IF_ELSE:%.*]]
2305 // CHECK3:       omp_if.then6:
2306 // 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*, i64, i64, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]]), !llvm.access.group !35
2307 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
2308 // CHECK3:       omp_if.else:
2309 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]]), !llvm.access.group !35
2310 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !35
2311 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !35
2312 // CHECK3-NEXT:    call void @.omp_outlined..9(i32* [[TMP17]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]]) #[[ATTR2]], !llvm.access.group !35
2313 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]]), !llvm.access.group !35
2314 // CHECK3-NEXT:    br label [[OMP_IF_END]]
2315 // CHECK3:       omp_if.end:
2316 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2317 // CHECK3:       omp.inner.for.inc:
2318 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
2319 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
2320 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
2321 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
2322 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
2323 // CHECK3:       omp.inner.for.end:
2324 // CHECK3-NEXT:    br label [[OMP_IF_END23:%.*]]
2325 // CHECK3:       omp_if.else7:
2326 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
2327 // CHECK3:       omp.inner.for.cond8:
2328 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2329 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2330 // CHECK3-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
2331 // CHECK3-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END22:%.*]]
2332 // CHECK3:       omp.inner.for.body10:
2333 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2334 // CHECK3-NEXT:    [[TMP23:%.*]] = zext i32 [[TMP22]] to i64
2335 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2336 // CHECK3-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP24]] to i64
2337 // CHECK3-NEXT:    [[TMP26:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
2338 // CHECK3-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP26]] to i1
2339 // CHECK3-NEXT:    [[CONV13:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED12]] to i8*
2340 // CHECK3-NEXT:    [[FROMBOOL14:%.*]] = zext i1 [[TOBOOL11]] to i8
2341 // CHECK3-NEXT:    store i8 [[FROMBOOL14]], i8* [[CONV13]], align 1
2342 // CHECK3-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED12]], align 8
2343 // CHECK3-NEXT:    [[TMP28:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
2344 // CHECK3-NEXT:    [[TOBOOL15:%.*]] = trunc i8 [[TMP28]] to i1
2345 // CHECK3-NEXT:    br i1 [[TOBOOL15]], label [[OMP_IF_THEN16:%.*]], label [[OMP_IF_ELSE17:%.*]]
2346 // CHECK3:       omp_if.then16:
2347 // 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*, i64, i64, i64)* @.omp_outlined..10 to void (i32*, i32*, ...)*), i64 [[TMP23]], i64 [[TMP25]], i64 [[TMP27]])
2348 // CHECK3-NEXT:    br label [[OMP_IF_END19:%.*]]
2349 // CHECK3:       omp_if.else17:
2350 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
2351 // CHECK3-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2352 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR18]], align 4
2353 // CHECK3-NEXT:    call void @.omp_outlined..10(i32* [[TMP29]], i32* [[DOTBOUND_ZERO_ADDR18]], i64 [[TMP23]], i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR2]]
2354 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
2355 // CHECK3-NEXT:    br label [[OMP_IF_END19]]
2356 // CHECK3:       omp_if.end19:
2357 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC20:%.*]]
2358 // CHECK3:       omp.inner.for.inc20:
2359 // CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2360 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2361 // CHECK3-NEXT:    [[ADD21:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
2362 // CHECK3-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
2363 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP38:![0-9]+]]
2364 // CHECK3:       omp.inner.for.end22:
2365 // CHECK3-NEXT:    br label [[OMP_IF_END23]]
2366 // CHECK3:       omp_if.end23:
2367 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2368 // CHECK3:       omp.loop.exit:
2369 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2370 // CHECK3-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2371 // CHECK3-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
2372 // CHECK3-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2373 // CHECK3:       .omp.final.then:
2374 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
2375 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2376 // CHECK3:       .omp.final.done:
2377 // CHECK3-NEXT:    ret void
2378 //
2379 //
2380 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..9
2381 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
2382 // CHECK3-NEXT:  entry:
2383 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2384 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2385 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2386 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2387 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2388 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2389 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2390 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2391 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2392 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2393 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2394 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2395 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2396 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2397 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2398 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2399 // CHECK3-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2400 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
2401 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2402 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
2403 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2404 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP0]] to i32
2405 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2406 // CHECK3-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
2407 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
2408 // CHECK3-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
2409 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2410 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2411 // CHECK3-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
2412 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
2413 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2414 // CHECK3:       omp_if.then:
2415 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2416 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2417 // 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)
2418 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2419 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
2420 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2421 // CHECK3:       cond.true:
2422 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2423 // CHECK3:       cond.false:
2424 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2425 // CHECK3-NEXT:    br label [[COND_END]]
2426 // CHECK3:       cond.end:
2427 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2428 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2429 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2430 // CHECK3-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
2431 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2432 // CHECK3:       omp.inner.for.cond:
2433 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
2434 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
2435 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2436 // CHECK3-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2437 // CHECK3:       omp.inner.for.body:
2438 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
2439 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2440 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2441 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
2442 // CHECK3-NEXT:    call void @_Z3fn6v(), !llvm.access.group !39
2443 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2444 // CHECK3:       omp.body.continue:
2445 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2446 // CHECK3:       omp.inner.for.inc:
2447 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
2448 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
2449 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
2450 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
2451 // CHECK3:       omp.inner.for.end:
2452 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
2453 // CHECK3:       omp_if.else:
2454 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2455 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
2456 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2457 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2458 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP14]], 99
2459 // CHECK3-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
2460 // CHECK3:       cond.true6:
2461 // CHECK3-NEXT:    br label [[COND_END8:%.*]]
2462 // CHECK3:       cond.false7:
2463 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2464 // CHECK3-NEXT:    br label [[COND_END8]]
2465 // CHECK3:       cond.end8:
2466 // CHECK3-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP15]], [[COND_FALSE7]] ]
2467 // CHECK3-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_UB]], align 4
2468 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2469 // CHECK3-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
2470 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND10:%.*]]
2471 // CHECK3:       omp.inner.for.cond10:
2472 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2473 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2474 // CHECK3-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
2475 // CHECK3-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY12:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
2476 // CHECK3:       omp.inner.for.body12:
2477 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2478 // CHECK3-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP19]], 1
2479 // CHECK3-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
2480 // CHECK3-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
2481 // CHECK3-NEXT:    call void @_Z3fn6v()
2482 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
2483 // CHECK3:       omp.body.continue15:
2484 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
2485 // CHECK3:       omp.inner.for.inc16:
2486 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2487 // CHECK3-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
2488 // CHECK3-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
2489 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND10]], !llvm.loop [[LOOP42:![0-9]+]]
2490 // CHECK3:       omp.inner.for.end18:
2491 // CHECK3-NEXT:    br label [[OMP_IF_END]]
2492 // CHECK3:       omp_if.end:
2493 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2494 // CHECK3:       omp.loop.exit:
2495 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2496 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
2497 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
2498 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2499 // CHECK3-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
2500 // CHECK3-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2501 // CHECK3:       .omp.final.then:
2502 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
2503 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2504 // CHECK3:       .omp.final.done:
2505 // CHECK3-NEXT:    ret void
2506 //
2507 //
2508 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..10
2509 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
2510 // CHECK3-NEXT:  entry:
2511 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2512 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2513 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2514 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2515 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
2516 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2517 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2518 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2519 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2520 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2521 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2522 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2523 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2524 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2525 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2526 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2527 // CHECK3-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
2528 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
2529 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2530 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
2531 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2532 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP0]] to i32
2533 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2534 // CHECK3-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
2535 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
2536 // CHECK3-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
2537 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2538 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2539 // CHECK3-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
2540 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
2541 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2542 // CHECK3:       omp_if.then:
2543 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2544 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
2545 // 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)
2546 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2547 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
2548 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2549 // CHECK3:       cond.true:
2550 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2551 // CHECK3:       cond.false:
2552 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2553 // CHECK3-NEXT:    br label [[COND_END]]
2554 // CHECK3:       cond.end:
2555 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
2556 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2557 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2558 // CHECK3-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
2559 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2560 // CHECK3:       omp.inner.for.cond:
2561 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
2562 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !43
2563 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
2564 // CHECK3-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2565 // CHECK3:       omp.inner.for.body:
2566 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
2567 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
2568 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2569 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !43
2570 // CHECK3-NEXT:    call void @_Z3fn6v(), !llvm.access.group !43
2571 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2572 // CHECK3:       omp.body.continue:
2573 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2574 // CHECK3:       omp.inner.for.inc:
2575 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
2576 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
2577 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
2578 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
2579 // CHECK3:       omp.inner.for.end:
2580 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
2581 // CHECK3:       omp_if.else:
2582 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2583 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
2584 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2585 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2586 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP14]], 99
2587 // CHECK3-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
2588 // CHECK3:       cond.true6:
2589 // CHECK3-NEXT:    br label [[COND_END8:%.*]]
2590 // CHECK3:       cond.false7:
2591 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2592 // CHECK3-NEXT:    br label [[COND_END8]]
2593 // CHECK3:       cond.end8:
2594 // CHECK3-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP15]], [[COND_FALSE7]] ]
2595 // CHECK3-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_UB]], align 4
2596 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2597 // CHECK3-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
2598 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND10:%.*]]
2599 // CHECK3:       omp.inner.for.cond10:
2600 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2601 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2602 // CHECK3-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
2603 // CHECK3-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY12:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
2604 // CHECK3:       omp.inner.for.body12:
2605 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2606 // CHECK3-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP19]], 1
2607 // CHECK3-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
2608 // CHECK3-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
2609 // CHECK3-NEXT:    call void @_Z3fn6v()
2610 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
2611 // CHECK3:       omp.body.continue15:
2612 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
2613 // CHECK3:       omp.inner.for.inc16:
2614 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2615 // CHECK3-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
2616 // CHECK3-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
2617 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND10]], !llvm.loop [[LOOP46:![0-9]+]]
2618 // CHECK3:       omp.inner.for.end18:
2619 // CHECK3-NEXT:    br label [[OMP_IF_END]]
2620 // CHECK3:       omp_if.end:
2621 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2622 // CHECK3:       omp.loop.exit:
2623 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2624 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
2625 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
2626 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2627 // CHECK3-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
2628 // CHECK3-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2629 // CHECK3:       .omp.final.then:
2630 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
2631 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2632 // CHECK3:       .omp.final.done:
2633 // CHECK3-NEXT:    ret void
2634 //
2635 //
2636 // CHECK3-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
2637 // CHECK3-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
2638 // CHECK3-NEXT:  entry:
2639 // CHECK3-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
2640 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2641 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
2642 // CHECK3-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
2643 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
2644 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
2645 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
2646 // CHECK3-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
2647 // CHECK3-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
2648 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
2649 // CHECK3-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
2650 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
2651 // CHECK3-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2652 // CHECK3:       omp_offload.failed:
2653 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59() #[[ATTR2]]
2654 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2655 // CHECK3:       omp_offload.cont:
2656 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
2657 // CHECK3-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
2658 // CHECK3-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
2659 // CHECK3-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
2660 // CHECK3:       omp_offload.failed2:
2661 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65() #[[ATTR2]]
2662 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
2663 // CHECK3:       omp_offload.cont3:
2664 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
2665 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
2666 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
2667 // CHECK3-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
2668 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2669 // CHECK3-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
2670 // CHECK3-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
2671 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2672 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
2673 // CHECK3-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
2674 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2675 // CHECK3-NEXT:    store i8* null, i8** [[TMP10]], align 8
2676 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2677 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2678 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
2679 // CHECK3-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.17, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.18, i32 0, i32 0), i8** null, i8** null, i32 0, i32 1)
2680 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
2681 // CHECK3-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
2682 // CHECK3:       omp_offload.failed5:
2683 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71(i64 [[TMP5]]) #[[ATTR2]]
2684 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
2685 // CHECK3:       omp_offload.cont6:
2686 // CHECK3-NEXT:    ret i32 0
2687 //
2688 //
2689 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59
2690 // CHECK3-SAME: () #[[ATTR1]] {
2691 // CHECK3-NEXT:  entry:
2692 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..11 to void (i32*, i32*, ...)*))
2693 // CHECK3-NEXT:    ret void
2694 //
2695 //
2696 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
2697 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2698 // CHECK3-NEXT:  entry:
2699 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2700 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2701 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2702 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2703 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2704 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2705 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2706 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2707 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2708 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2709 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2710 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2711 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2712 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2713 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2714 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2715 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2716 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2717 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2718 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2719 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2720 // CHECK3:       cond.true:
2721 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2722 // CHECK3:       cond.false:
2723 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2724 // CHECK3-NEXT:    br label [[COND_END]]
2725 // CHECK3:       cond.end:
2726 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2727 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2728 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2729 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2730 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2731 // CHECK3:       omp.inner.for.cond:
2732 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
2733 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
2734 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2735 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2736 // CHECK3:       omp.inner.for.body:
2737 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !47
2738 // CHECK3-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2739 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
2740 // CHECK3-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2741 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group !47
2742 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2743 // CHECK3:       omp.inner.for.inc:
2744 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
2745 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !47
2746 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
2747 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
2748 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
2749 // CHECK3:       omp.inner.for.end:
2750 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2751 // CHECK3:       omp.loop.exit:
2752 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2753 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2754 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
2755 // CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2756 // CHECK3:       .omp.final.then:
2757 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
2758 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2759 // CHECK3:       .omp.final.done:
2760 // CHECK3-NEXT:    ret void
2761 //
2762 //
2763 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..12
2764 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2765 // CHECK3-NEXT:  entry:
2766 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2767 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2768 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2769 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2770 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2771 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2772 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2773 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2774 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2775 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2776 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2777 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2778 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2779 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2780 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2781 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2782 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
2783 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2784 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2785 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2786 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2787 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2788 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2789 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2790 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2791 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2792 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2793 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2794 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2795 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2796 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2797 // CHECK3:       cond.true:
2798 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2799 // CHECK3:       cond.false:
2800 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2801 // CHECK3-NEXT:    br label [[COND_END]]
2802 // CHECK3:       cond.end:
2803 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2804 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2805 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2806 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2807 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2808 // CHECK3:       omp.inner.for.cond:
2809 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
2810 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !50
2811 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2812 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2813 // CHECK3:       omp.inner.for.body:
2814 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
2815 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2816 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2817 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !50
2818 // CHECK3-NEXT:    call void @_Z3fn1v(), !llvm.access.group !50
2819 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2820 // CHECK3:       omp.body.continue:
2821 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2822 // CHECK3:       omp.inner.for.inc:
2823 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
2824 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2825 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
2826 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
2827 // CHECK3:       omp.inner.for.end:
2828 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2829 // CHECK3:       omp.loop.exit:
2830 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2831 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2832 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2833 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2834 // CHECK3:       .omp.final.then:
2835 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
2836 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2837 // CHECK3:       .omp.final.done:
2838 // CHECK3-NEXT:    ret void
2839 //
2840 //
2841 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65
2842 // CHECK3-SAME: () #[[ATTR1]] {
2843 // CHECK3-NEXT:  entry:
2844 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..13 to void (i32*, i32*, ...)*))
2845 // CHECK3-NEXT:    ret void
2846 //
2847 //
2848 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..13
2849 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2850 // CHECK3-NEXT:  entry:
2851 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2852 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2853 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2854 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2855 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2856 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2857 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2858 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2859 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2860 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
2861 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2862 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2863 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2864 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2865 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2866 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2867 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2868 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2869 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2870 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2871 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2872 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2873 // CHECK3:       cond.true:
2874 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2875 // CHECK3:       cond.false:
2876 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2877 // CHECK3-NEXT:    br label [[COND_END]]
2878 // CHECK3:       cond.end:
2879 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2880 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2881 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2882 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2883 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2884 // CHECK3:       omp.inner.for.cond:
2885 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2886 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2887 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2888 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2889 // CHECK3:       omp.inner.for.body:
2890 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2891 // CHECK3-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2892 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2893 // CHECK3-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2894 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
2895 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2896 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
2897 // CHECK3-NEXT:    call void @.omp_outlined..14(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
2898 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
2899 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2900 // CHECK3:       omp.inner.for.inc:
2901 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2902 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2903 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2904 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2905 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP53:![0-9]+]]
2906 // CHECK3:       omp.inner.for.end:
2907 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2908 // CHECK3:       omp.loop.exit:
2909 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2910 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2911 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
2912 // CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2913 // CHECK3:       .omp.final.then:
2914 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
2915 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2916 // CHECK3:       .omp.final.done:
2917 // CHECK3-NEXT:    ret void
2918 //
2919 //
2920 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..14
2921 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2922 // CHECK3-NEXT:  entry:
2923 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2924 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2925 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2926 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2927 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2928 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2929 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2930 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2931 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2932 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2933 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2934 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2935 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2936 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2937 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2938 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2939 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
2940 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2941 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2942 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2943 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2944 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2945 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2946 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2947 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2948 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2949 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2950 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2951 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2952 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2953 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2954 // CHECK3:       cond.true:
2955 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2956 // CHECK3:       cond.false:
2957 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2958 // CHECK3-NEXT:    br label [[COND_END]]
2959 // CHECK3:       cond.end:
2960 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2961 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2962 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2963 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2964 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2965 // CHECK3:       omp.inner.for.cond:
2966 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2967 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2968 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2969 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2970 // CHECK3:       omp.inner.for.body:
2971 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2972 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2973 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2974 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
2975 // CHECK3-NEXT:    call void @_Z3fn2v()
2976 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2977 // CHECK3:       omp.body.continue:
2978 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2979 // CHECK3:       omp.inner.for.inc:
2980 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2981 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2982 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
2983 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP54:![0-9]+]]
2984 // CHECK3:       omp.inner.for.end:
2985 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2986 // CHECK3:       omp.loop.exit:
2987 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2988 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2989 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2990 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2991 // CHECK3:       .omp.final.then:
2992 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
2993 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2994 // CHECK3:       .omp.final.done:
2995 // CHECK3-NEXT:    ret void
2996 //
2997 //
2998 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71
2999 // CHECK3-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
3000 // CHECK3-NEXT:  entry:
3001 // CHECK3-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
3002 // CHECK3-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
3003 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
3004 // 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*, i32*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32* [[CONV]])
3005 // CHECK3-NEXT:    ret void
3006 //
3007 //
3008 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..15
3009 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
3010 // CHECK3-NEXT:  entry:
3011 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3012 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3013 // CHECK3-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
3014 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3015 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3016 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3017 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3018 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3019 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3020 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3021 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
3022 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3023 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3024 // CHECK3-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
3025 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
3026 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3027 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3028 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3029 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3030 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3031 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
3032 // 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)
3033 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3034 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
3035 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3036 // CHECK3:       cond.true:
3037 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3038 // CHECK3:       cond.false:
3039 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3040 // CHECK3-NEXT:    br label [[COND_END]]
3041 // CHECK3:       cond.end:
3042 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
3043 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3044 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3045 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
3046 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3047 // CHECK3:       omp.inner.for.cond:
3048 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
3049 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !55
3050 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3051 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3052 // CHECK3:       omp.inner.for.body:
3053 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !55
3054 // CHECK3-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
3055 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !55
3056 // CHECK3-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
3057 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !55
3058 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
3059 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3060 // CHECK3:       omp_if.then:
3061 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]]), !llvm.access.group !55
3062 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
3063 // CHECK3:       omp_if.else:
3064 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !55
3065 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !55
3066 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !55
3067 // CHECK3-NEXT:    call void @.omp_outlined..16(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !55
3068 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !55
3069 // CHECK3-NEXT:    br label [[OMP_IF_END]]
3070 // CHECK3:       omp_if.end:
3071 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3072 // CHECK3:       omp.inner.for.inc:
3073 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
3074 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !55
3075 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
3076 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
3077 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP56:![0-9]+]]
3078 // CHECK3:       omp.inner.for.end:
3079 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3080 // CHECK3:       omp.loop.exit:
3081 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
3082 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3083 // CHECK3-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
3084 // CHECK3-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3085 // CHECK3:       .omp.final.then:
3086 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
3087 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3088 // CHECK3:       .omp.final.done:
3089 // CHECK3-NEXT:    ret void
3090 //
3091 //
3092 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..16
3093 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
3094 // CHECK3-NEXT:  entry:
3095 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3096 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3097 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3098 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3099 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3100 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3101 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3102 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3103 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3104 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3105 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3106 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3107 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3108 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3109 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3110 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3111 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3112 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3113 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
3114 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3115 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3116 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3117 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3118 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3119 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3120 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3121 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3122 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3123 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3124 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3125 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3126 // CHECK3:       cond.true:
3127 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3128 // CHECK3:       cond.false:
3129 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3130 // CHECK3-NEXT:    br label [[COND_END]]
3131 // CHECK3:       cond.end:
3132 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3133 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3134 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3135 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3136 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3137 // CHECK3:       omp.inner.for.cond:
3138 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
3139 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !58
3140 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3141 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3142 // CHECK3:       omp.inner.for.body:
3143 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
3144 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3145 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3146 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !58
3147 // CHECK3-NEXT:    call void @_Z3fn3v(), !llvm.access.group !58
3148 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3149 // CHECK3:       omp.body.continue:
3150 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3151 // CHECK3:       omp.inner.for.inc:
3152 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
3153 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3154 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
3155 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP59:![0-9]+]]
3156 // CHECK3:       omp.inner.for.end:
3157 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3158 // CHECK3:       omp.loop.exit:
3159 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3160 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3161 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3162 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3163 // CHECK3:       .omp.final.then:
3164 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
3165 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3166 // CHECK3:       .omp.final.done:
3167 // CHECK3-NEXT:    ret void
3168 //
3169 //
3170 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
3171 // CHECK3-SAME: () #[[ATTR5:[0-9]+]] {
3172 // CHECK3-NEXT:  entry:
3173 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
3174 // CHECK3-NEXT:    ret void
3175 //
3176 //
3177 // CHECK5-LABEL: define {{[^@]+}}@_Z9gtid_testv
3178 // CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
3179 // CHECK5-NEXT:  entry:
3180 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3181 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3182 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3183 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3184 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
3185 // CHECK5-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
3186 // CHECK5-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
3187 // CHECK5-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
3188 // CHECK5-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
3189 // CHECK5-NEXT:    [[I6:%.*]] = alloca i32, align 4
3190 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3191 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3192 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3193 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
3194 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3195 // CHECK5:       omp.inner.for.cond:
3196 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
3197 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
3198 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
3199 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3200 // CHECK5:       omp.inner.for.body:
3201 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
3202 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
3203 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3204 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
3205 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3206 // CHECK5:       omp.body.continue:
3207 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3208 // CHECK5:       omp.inner.for.inc:
3209 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
3210 // CHECK5-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
3211 // CHECK5-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
3212 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
3213 // CHECK5:       omp.inner.for.end:
3214 // CHECK5-NEXT:    store i32 100, i32* [[I]], align 4
3215 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
3216 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
3217 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
3218 // CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
3219 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
3220 // CHECK5:       omp.inner.for.cond7:
3221 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
3222 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !6
3223 // CHECK5-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3224 // CHECK5-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
3225 // CHECK5:       omp.inner.for.body9:
3226 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
3227 // CHECK5-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
3228 // CHECK5-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
3229 // CHECK5-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !6
3230 // CHECK5-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !6
3231 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
3232 // CHECK5:       omp.body.continue12:
3233 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
3234 // CHECK5:       omp.inner.for.inc13:
3235 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
3236 // CHECK5-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
3237 // CHECK5-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
3238 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP7:![0-9]+]]
3239 // CHECK5:       omp.inner.for.end15:
3240 // CHECK5-NEXT:    store i32 100, i32* [[I6]], align 4
3241 // CHECK5-NEXT:    ret void
3242 //
3243 //
3244 // CHECK5-LABEL: define {{[^@]+}}@main
3245 // CHECK5-SAME: () #[[ATTR1:[0-9]+]] {
3246 // CHECK5-NEXT:  entry:
3247 // CHECK5-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
3248 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3249 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3250 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3251 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3252 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
3253 // CHECK5-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
3254 // CHECK5-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
3255 // CHECK5-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
3256 // CHECK5-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
3257 // CHECK5-NEXT:    [[I6:%.*]] = alloca i32, align 4
3258 // CHECK5-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
3259 // CHECK5-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
3260 // CHECK5-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
3261 // CHECK5-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
3262 // CHECK5-NEXT:    [[I20:%.*]] = alloca i32, align 4
3263 // CHECK5-NEXT:    store i32 0, i32* [[RETVAL]], align 4
3264 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3265 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3266 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3267 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
3268 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3269 // CHECK5:       omp.inner.for.cond:
3270 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
3271 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
3272 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
3273 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3274 // CHECK5:       omp.inner.for.body:
3275 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
3276 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
3277 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3278 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
3279 // CHECK5-NEXT:    call void @_Z3fn4v(), !llvm.access.group !9
3280 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3281 // CHECK5:       omp.body.continue:
3282 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3283 // CHECK5:       omp.inner.for.inc:
3284 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
3285 // CHECK5-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
3286 // CHECK5-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
3287 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
3288 // CHECK5:       omp.inner.for.end:
3289 // CHECK5-NEXT:    store i32 100, i32* [[I]], align 4
3290 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
3291 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
3292 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
3293 // CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
3294 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
3295 // CHECK5:       omp.inner.for.cond7:
3296 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !12
3297 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !12
3298 // CHECK5-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3299 // CHECK5-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
3300 // CHECK5:       omp.inner.for.body9:
3301 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !12
3302 // CHECK5-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
3303 // CHECK5-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
3304 // CHECK5-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !12
3305 // CHECK5-NEXT:    call void @_Z3fn5v(), !llvm.access.group !12
3306 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
3307 // CHECK5:       omp.body.continue12:
3308 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
3309 // CHECK5:       omp.inner.for.inc13:
3310 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !12
3311 // CHECK5-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
3312 // CHECK5-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !12
3313 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP13:![0-9]+]]
3314 // CHECK5:       omp.inner.for.end15:
3315 // CHECK5-NEXT:    store i32 100, i32* [[I6]], align 4
3316 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
3317 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
3318 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
3319 // CHECK5-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
3320 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
3321 // CHECK5:       omp.inner.for.cond21:
3322 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !15
3323 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !15
3324 // CHECK5-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
3325 // CHECK5-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
3326 // CHECK5:       omp.inner.for.body23:
3327 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !15
3328 // CHECK5-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
3329 // CHECK5-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
3330 // CHECK5-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !15
3331 // CHECK5-NEXT:    call void @_Z3fn6v(), !llvm.access.group !15
3332 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
3333 // CHECK5:       omp.body.continue26:
3334 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
3335 // CHECK5:       omp.inner.for.inc27:
3336 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !15
3337 // CHECK5-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
3338 // CHECK5-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !15
3339 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP16:![0-9]+]]
3340 // CHECK5:       omp.inner.for.end29:
3341 // CHECK5-NEXT:    store i32 100, i32* [[I20]], align 4
3342 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
3343 // CHECK5-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
3344 // CHECK5-NEXT:    ret i32 [[CALL]]
3345 //
3346 //
3347 // CHECK5-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
3348 // CHECK5-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
3349 // CHECK5-NEXT:  entry:
3350 // CHECK5-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
3351 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3352 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3353 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3354 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3355 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
3356 // CHECK5-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
3357 // CHECK5-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
3358 // CHECK5-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
3359 // CHECK5-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
3360 // CHECK5-NEXT:    [[I6:%.*]] = alloca i32, align 4
3361 // CHECK5-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
3362 // CHECK5-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
3363 // CHECK5-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
3364 // CHECK5-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
3365 // CHECK5-NEXT:    [[I20:%.*]] = alloca i32, align 4
3366 // CHECK5-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
3367 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3368 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3369 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3370 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
3371 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3372 // CHECK5:       omp.inner.for.cond:
3373 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
3374 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
3375 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
3376 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3377 // CHECK5:       omp.inner.for.body:
3378 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
3379 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
3380 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3381 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !18
3382 // CHECK5-NEXT:    call void @_Z3fn1v(), !llvm.access.group !18
3383 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3384 // CHECK5:       omp.body.continue:
3385 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3386 // CHECK5:       omp.inner.for.inc:
3387 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
3388 // CHECK5-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
3389 // CHECK5-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
3390 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
3391 // CHECK5:       omp.inner.for.end:
3392 // CHECK5-NEXT:    store i32 100, i32* [[I]], align 4
3393 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
3394 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
3395 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
3396 // CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
3397 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
3398 // CHECK5:       omp.inner.for.cond7:
3399 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
3400 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !21
3401 // CHECK5-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3402 // CHECK5-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
3403 // CHECK5:       omp.inner.for.body9:
3404 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
3405 // CHECK5-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
3406 // CHECK5-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
3407 // CHECK5-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !21
3408 // CHECK5-NEXT:    call void @_Z3fn2v(), !llvm.access.group !21
3409 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
3410 // CHECK5:       omp.body.continue12:
3411 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
3412 // CHECK5:       omp.inner.for.inc13:
3413 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
3414 // CHECK5-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
3415 // CHECK5-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
3416 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP22:![0-9]+]]
3417 // CHECK5:       omp.inner.for.end15:
3418 // CHECK5-NEXT:    store i32 100, i32* [[I6]], align 4
3419 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
3420 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
3421 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
3422 // CHECK5-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
3423 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
3424 // CHECK5:       omp.inner.for.cond21:
3425 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !24
3426 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !24
3427 // CHECK5-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
3428 // CHECK5-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
3429 // CHECK5:       omp.inner.for.body23:
3430 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !24
3431 // CHECK5-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
3432 // CHECK5-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
3433 // CHECK5-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !24
3434 // CHECK5-NEXT:    call void @_Z3fn3v(), !llvm.access.group !24
3435 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
3436 // CHECK5:       omp.body.continue26:
3437 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
3438 // CHECK5:       omp.inner.for.inc27:
3439 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !24
3440 // CHECK5-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
3441 // CHECK5-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !24
3442 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP25:![0-9]+]]
3443 // CHECK5:       omp.inner.for.end29:
3444 // CHECK5-NEXT:    store i32 100, i32* [[I20]], align 4
3445 // CHECK5-NEXT:    ret i32 0
3446 //
3447 //
3448 // CHECK7-LABEL: define {{[^@]+}}@_Z9gtid_testv
3449 // CHECK7-SAME: () #[[ATTR0:[0-9]+]] {
3450 // CHECK7-NEXT:  entry:
3451 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3452 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3453 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3454 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3455 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
3456 // CHECK7-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
3457 // CHECK7-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
3458 // CHECK7-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
3459 // CHECK7-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
3460 // CHECK7-NEXT:    [[I6:%.*]] = alloca i32, align 4
3461 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3462 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3463 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3464 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
3465 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3466 // CHECK7:       omp.inner.for.cond:
3467 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
3468 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
3469 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
3470 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3471 // CHECK7:       omp.inner.for.body:
3472 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
3473 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
3474 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3475 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
3476 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3477 // CHECK7:       omp.body.continue:
3478 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3479 // CHECK7:       omp.inner.for.inc:
3480 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
3481 // CHECK7-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
3482 // CHECK7-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
3483 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
3484 // CHECK7:       omp.inner.for.end:
3485 // CHECK7-NEXT:    store i32 100, i32* [[I]], align 4
3486 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
3487 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
3488 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
3489 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
3490 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
3491 // CHECK7:       omp.inner.for.cond7:
3492 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
3493 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !6
3494 // CHECK7-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3495 // CHECK7-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
3496 // CHECK7:       omp.inner.for.body9:
3497 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
3498 // CHECK7-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
3499 // CHECK7-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
3500 // CHECK7-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !6
3501 // CHECK7-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !6
3502 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
3503 // CHECK7:       omp.body.continue12:
3504 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
3505 // CHECK7:       omp.inner.for.inc13:
3506 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
3507 // CHECK7-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
3508 // CHECK7-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
3509 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP7:![0-9]+]]
3510 // CHECK7:       omp.inner.for.end15:
3511 // CHECK7-NEXT:    store i32 100, i32* [[I6]], align 4
3512 // CHECK7-NEXT:    ret void
3513 //
3514 //
3515 // CHECK7-LABEL: define {{[^@]+}}@main
3516 // CHECK7-SAME: () #[[ATTR1:[0-9]+]] {
3517 // CHECK7-NEXT:  entry:
3518 // CHECK7-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
3519 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3520 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3521 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3522 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3523 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
3524 // CHECK7-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
3525 // CHECK7-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
3526 // CHECK7-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
3527 // CHECK7-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
3528 // CHECK7-NEXT:    [[I6:%.*]] = alloca i32, align 4
3529 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
3530 // CHECK7-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
3531 // CHECK7-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
3532 // CHECK7-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
3533 // CHECK7-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
3534 // CHECK7-NEXT:    [[I20:%.*]] = alloca i32, align 4
3535 // CHECK7-NEXT:    store i32 0, i32* [[RETVAL]], align 4
3536 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3537 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3538 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3539 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
3540 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3541 // CHECK7:       omp.inner.for.cond:
3542 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
3543 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
3544 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
3545 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3546 // CHECK7:       omp.inner.for.body:
3547 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
3548 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
3549 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3550 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
3551 // CHECK7-NEXT:    call void @_Z3fn4v(), !llvm.access.group !9
3552 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3553 // CHECK7:       omp.body.continue:
3554 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3555 // CHECK7:       omp.inner.for.inc:
3556 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
3557 // CHECK7-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
3558 // CHECK7-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
3559 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
3560 // CHECK7:       omp.inner.for.end:
3561 // CHECK7-NEXT:    store i32 100, i32* [[I]], align 4
3562 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
3563 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
3564 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
3565 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
3566 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
3567 // CHECK7:       omp.inner.for.cond7:
3568 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
3569 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4
3570 // CHECK7-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3571 // CHECK7-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
3572 // CHECK7:       omp.inner.for.body9:
3573 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
3574 // CHECK7-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
3575 // CHECK7-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
3576 // CHECK7-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4
3577 // CHECK7-NEXT:    call void @_Z3fn5v()
3578 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
3579 // CHECK7:       omp.body.continue12:
3580 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
3581 // CHECK7:       omp.inner.for.inc13:
3582 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
3583 // CHECK7-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
3584 // CHECK7-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4
3585 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP12:![0-9]+]]
3586 // CHECK7:       omp.inner.for.end15:
3587 // CHECK7-NEXT:    store i32 100, i32* [[I6]], align 4
3588 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* @Arg, align 4
3589 // CHECK7-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP10]], 0
3590 // CHECK7-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
3591 // CHECK7-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
3592 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
3593 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
3594 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
3595 // CHECK7-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV19]], align 4
3596 // CHECK7-NEXT:    [[TMP12:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
3597 // CHECK7-NEXT:    [[TOBOOL21:%.*]] = trunc i8 [[TMP12]] to i1
3598 // CHECK7-NEXT:    br i1 [[TOBOOL21]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3599 // CHECK7:       omp_if.then:
3600 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND22:%.*]]
3601 // CHECK7:       omp.inner.for.cond22:
3602 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !14
3603 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !14
3604 // CHECK7-NEXT:    [[CMP23:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
3605 // CHECK7-NEXT:    br i1 [[CMP23]], label [[OMP_INNER_FOR_BODY24:%.*]], label [[OMP_INNER_FOR_END30:%.*]]
3606 // CHECK7:       omp.inner.for.body24:
3607 // CHECK7-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !14
3608 // CHECK7-NEXT:    [[MUL25:%.*]] = mul nsw i32 [[TMP15]], 1
3609 // CHECK7-NEXT:    [[ADD26:%.*]] = add nsw i32 0, [[MUL25]]
3610 // CHECK7-NEXT:    store i32 [[ADD26]], i32* [[I20]], align 4, !llvm.access.group !14
3611 // CHECK7-NEXT:    call void @_Z3fn6v(), !llvm.access.group !14
3612 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE27:%.*]]
3613 // CHECK7:       omp.body.continue27:
3614 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC28:%.*]]
3615 // CHECK7:       omp.inner.for.inc28:
3616 // CHECK7-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !14
3617 // CHECK7-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP16]], 1
3618 // CHECK7-NEXT:    store i32 [[ADD29]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !14
3619 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND22]], !llvm.loop [[LOOP15:![0-9]+]]
3620 // CHECK7:       omp.inner.for.end30:
3621 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
3622 // CHECK7:       omp_if.else:
3623 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND31:%.*]]
3624 // CHECK7:       omp.inner.for.cond31:
3625 // CHECK7-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
3626 // CHECK7-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4
3627 // CHECK7-NEXT:    [[CMP32:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
3628 // CHECK7-NEXT:    br i1 [[CMP32]], label [[OMP_INNER_FOR_BODY33:%.*]], label [[OMP_INNER_FOR_END39:%.*]]
3629 // CHECK7:       omp.inner.for.body33:
3630 // CHECK7-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
3631 // CHECK7-NEXT:    [[MUL34:%.*]] = mul nsw i32 [[TMP19]], 1
3632 // CHECK7-NEXT:    [[ADD35:%.*]] = add nsw i32 0, [[MUL34]]
3633 // CHECK7-NEXT:    store i32 [[ADD35]], i32* [[I20]], align 4
3634 // CHECK7-NEXT:    call void @_Z3fn6v()
3635 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE36:%.*]]
3636 // CHECK7:       omp.body.continue36:
3637 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC37:%.*]]
3638 // CHECK7:       omp.inner.for.inc37:
3639 // CHECK7-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
3640 // CHECK7-NEXT:    [[ADD38:%.*]] = add nsw i32 [[TMP20]], 1
3641 // CHECK7-NEXT:    store i32 [[ADD38]], i32* [[DOTOMP_IV19]], align 4
3642 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND31]], !llvm.loop [[LOOP17:![0-9]+]]
3643 // CHECK7:       omp.inner.for.end39:
3644 // CHECK7-NEXT:    br label [[OMP_IF_END]]
3645 // CHECK7:       omp_if.end:
3646 // CHECK7-NEXT:    store i32 100, i32* [[I20]], align 4
3647 // CHECK7-NEXT:    [[TMP21:%.*]] = load i32, i32* @Arg, align 4
3648 // CHECK7-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP21]])
3649 // CHECK7-NEXT:    ret i32 [[CALL]]
3650 //
3651 //
3652 // CHECK7-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
3653 // CHECK7-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
3654 // CHECK7-NEXT:  entry:
3655 // CHECK7-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
3656 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3657 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3658 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3659 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3660 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
3661 // CHECK7-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
3662 // CHECK7-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
3663 // CHECK7-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
3664 // CHECK7-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
3665 // CHECK7-NEXT:    [[I6:%.*]] = alloca i32, align 4
3666 // CHECK7-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
3667 // CHECK7-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
3668 // CHECK7-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
3669 // CHECK7-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
3670 // CHECK7-NEXT:    [[I20:%.*]] = alloca i32, align 4
3671 // CHECK7-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
3672 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3673 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3674 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3675 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
3676 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3677 // CHECK7:       omp.inner.for.cond:
3678 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
3679 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
3680 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
3681 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3682 // CHECK7:       omp.inner.for.body:
3683 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
3684 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
3685 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3686 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !18
3687 // CHECK7-NEXT:    call void @_Z3fn1v(), !llvm.access.group !18
3688 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3689 // CHECK7:       omp.body.continue:
3690 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3691 // CHECK7:       omp.inner.for.inc:
3692 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
3693 // CHECK7-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
3694 // CHECK7-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
3695 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
3696 // CHECK7:       omp.inner.for.end:
3697 // CHECK7-NEXT:    store i32 100, i32* [[I]], align 4
3698 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
3699 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
3700 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
3701 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
3702 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
3703 // CHECK7:       omp.inner.for.cond7:
3704 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
3705 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4
3706 // CHECK7-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
3707 // CHECK7-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
3708 // CHECK7:       omp.inner.for.body9:
3709 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
3710 // CHECK7-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
3711 // CHECK7-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
3712 // CHECK7-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4
3713 // CHECK7-NEXT:    call void @_Z3fn2v()
3714 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
3715 // CHECK7:       omp.body.continue12:
3716 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
3717 // CHECK7:       omp.inner.for.inc13:
3718 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
3719 // CHECK7-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
3720 // CHECK7-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4
3721 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP21:![0-9]+]]
3722 // CHECK7:       omp.inner.for.end15:
3723 // CHECK7-NEXT:    store i32 100, i32* [[I6]], align 4
3724 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
3725 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
3726 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
3727 // CHECK7-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
3728 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
3729 // CHECK7:       omp.inner.for.cond21:
3730 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !22
3731 // CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !22
3732 // CHECK7-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
3733 // CHECK7-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
3734 // CHECK7:       omp.inner.for.body23:
3735 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !22
3736 // CHECK7-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
3737 // CHECK7-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
3738 // CHECK7-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !22
3739 // CHECK7-NEXT:    call void @_Z3fn3v(), !llvm.access.group !22
3740 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
3741 // CHECK7:       omp.body.continue26:
3742 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
3743 // CHECK7:       omp.inner.for.inc27:
3744 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !22
3745 // CHECK7-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
3746 // CHECK7-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !22
3747 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP23:![0-9]+]]
3748 // CHECK7:       omp.inner.for.end29:
3749 // CHECK7-NEXT:    store i32 100, i32* [[I20]], align 4
3750 // CHECK7-NEXT:    ret i32 0
3751 //
3752 //
3753 // CHECK9-LABEL: define {{[^@]+}}@_Z9gtid_testv
3754 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
3755 // CHECK9-NEXT:  entry:
3756 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3757 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
3758 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
3759 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
3760 // CHECK9-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
3761 // CHECK9-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3762 // CHECK9:       omp_offload.failed:
3763 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43() #[[ATTR2:[0-9]+]]
3764 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3765 // CHECK9:       omp_offload.cont:
3766 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
3767 // CHECK9-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
3768 // CHECK9-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
3769 // CHECK9-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
3770 // CHECK9:       omp_offload.failed2:
3771 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48() #[[ATTR2]]
3772 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
3773 // CHECK9:       omp_offload.cont3:
3774 // CHECK9-NEXT:    ret void
3775 //
3776 //
3777 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43
3778 // CHECK9-SAME: () #[[ATTR1:[0-9]+]] {
3779 // CHECK9-NEXT:  entry:
3780 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
3781 // CHECK9-NEXT:    ret void
3782 //
3783 //
3784 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
3785 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3786 // CHECK9-NEXT:  entry:
3787 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3788 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3789 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3790 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3791 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3792 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3793 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3794 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3795 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
3796 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3797 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3798 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3799 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3800 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3801 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3802 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3803 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3804 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3805 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3806 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
3807 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3808 // CHECK9:       cond.true:
3809 // CHECK9-NEXT:    br label [[COND_END:%.*]]
3810 // CHECK9:       cond.false:
3811 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3812 // CHECK9-NEXT:    br label [[COND_END]]
3813 // CHECK9:       cond.end:
3814 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
3815 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3816 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3817 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
3818 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3819 // CHECK9:       omp.inner.for.cond:
3820 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
3821 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
3822 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3823 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3824 // CHECK9:       omp.inner.for.body:
3825 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !15
3826 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
3827 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
3828 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
3829 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group !15
3830 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3831 // CHECK9:       omp.inner.for.inc:
3832 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
3833 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !15
3834 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
3835 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
3836 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
3837 // CHECK9:       omp.inner.for.end:
3838 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3839 // CHECK9:       omp.loop.exit:
3840 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
3841 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3842 // CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
3843 // CHECK9-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3844 // CHECK9:       .omp.final.then:
3845 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
3846 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3847 // CHECK9:       .omp.final.done:
3848 // CHECK9-NEXT:    ret void
3849 //
3850 //
3851 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
3852 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
3853 // CHECK9-NEXT:  entry:
3854 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3855 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3856 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3857 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3858 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3859 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3860 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3861 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3862 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3863 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3864 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
3865 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3866 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3867 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3868 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3869 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3870 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3871 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3872 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
3873 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3874 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3875 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3876 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3877 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3878 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3879 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3880 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3881 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3882 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3883 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3884 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3885 // CHECK9:       cond.true:
3886 // CHECK9-NEXT:    br label [[COND_END:%.*]]
3887 // CHECK9:       cond.false:
3888 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3889 // CHECK9-NEXT:    br label [[COND_END]]
3890 // CHECK9:       cond.end:
3891 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3892 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3893 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3894 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3895 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3896 // CHECK9:       omp.inner.for.cond:
3897 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
3898 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
3899 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3900 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3901 // CHECK9:       omp.inner.for.body:
3902 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
3903 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3904 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3905 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !19
3906 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3907 // CHECK9:       omp.body.continue:
3908 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3909 // CHECK9:       omp.inner.for.inc:
3910 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
3911 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3912 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
3913 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
3914 // CHECK9:       omp.inner.for.end:
3915 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3916 // CHECK9:       omp.loop.exit:
3917 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3918 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3919 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3920 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3921 // CHECK9:       .omp.final.then:
3922 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
3923 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3924 // CHECK9:       .omp.final.done:
3925 // CHECK9-NEXT:    ret void
3926 //
3927 //
3928 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48
3929 // CHECK9-SAME: () #[[ATTR1]] {
3930 // CHECK9-NEXT:  entry:
3931 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
3932 // CHECK9-NEXT:    ret void
3933 //
3934 //
3935 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
3936 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3937 // CHECK9-NEXT:  entry:
3938 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3939 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3940 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3941 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3942 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3943 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3944 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3945 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3946 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
3947 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
3948 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3949 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3950 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3951 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3952 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3953 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3954 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3955 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3956 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3957 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3958 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
3959 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3960 // CHECK9:       cond.true:
3961 // CHECK9-NEXT:    br label [[COND_END:%.*]]
3962 // CHECK9:       cond.false:
3963 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3964 // CHECK9-NEXT:    br label [[COND_END]]
3965 // CHECK9:       cond.end:
3966 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
3967 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3968 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3969 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
3970 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3971 // CHECK9:       omp.inner.for.cond:
3972 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
3973 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
3974 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3975 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3976 // CHECK9:       omp.inner.for.body:
3977 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
3978 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
3979 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
3980 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
3981 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !24
3982 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !24
3983 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !24
3984 // CHECK9-NEXT:    call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !24
3985 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !24
3986 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3987 // CHECK9:       omp.inner.for.inc:
3988 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
3989 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
3990 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3991 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
3992 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
3993 // CHECK9:       omp.inner.for.end:
3994 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3995 // CHECK9:       omp.loop.exit:
3996 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
3997 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3998 // CHECK9-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
3999 // CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4000 // CHECK9:       .omp.final.then:
4001 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
4002 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4003 // CHECK9:       .omp.final.done:
4004 // CHECK9-NEXT:    ret void
4005 //
4006 //
4007 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
4008 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4009 // CHECK9-NEXT:  entry:
4010 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4011 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4012 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4013 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4014 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4015 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4016 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4017 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4018 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4019 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4020 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4021 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4022 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4023 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4024 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4025 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4026 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
4027 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4028 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4029 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4030 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4031 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4032 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4033 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4034 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4035 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4036 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4037 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4038 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4039 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4040 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4041 // CHECK9:       cond.true:
4042 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4043 // CHECK9:       cond.false:
4044 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4045 // CHECK9-NEXT:    br label [[COND_END]]
4046 // CHECK9:       cond.end:
4047 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4048 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4049 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4050 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4051 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4052 // CHECK9:       omp.inner.for.cond:
4053 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
4054 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
4055 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4056 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4057 // CHECK9:       omp.inner.for.body:
4058 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
4059 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4060 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4061 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !27
4062 // CHECK9-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !27
4063 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4064 // CHECK9:       omp.body.continue:
4065 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4066 // CHECK9:       omp.inner.for.inc:
4067 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
4068 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4069 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
4070 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
4071 // CHECK9:       omp.inner.for.end:
4072 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4073 // CHECK9:       omp.loop.exit:
4074 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4075 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4076 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4077 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4078 // CHECK9:       .omp.final.then:
4079 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
4080 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4081 // CHECK9:       .omp.final.done:
4082 // CHECK9-NEXT:    ret void
4083 //
4084 //
4085 // CHECK9-LABEL: define {{[^@]+}}@main
4086 // CHECK9-SAME: () #[[ATTR3:[0-9]+]] {
4087 // CHECK9-NEXT:  entry:
4088 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
4089 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4090 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
4091 // CHECK9-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
4092 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
4093 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
4094 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
4095 // CHECK9-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
4096 // CHECK9-NEXT:    store i32 0, i32* [[RETVAL]], align 4
4097 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4098 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
4099 // CHECK9-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
4100 // CHECK9-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4101 // CHECK9:       omp_offload.failed:
4102 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81() #[[ATTR2]]
4103 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4104 // CHECK9:       omp_offload.cont:
4105 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4106 // CHECK9-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
4107 // CHECK9-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
4108 // CHECK9-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
4109 // CHECK9:       omp_offload.failed2:
4110 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90() #[[ATTR2]]
4111 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
4112 // CHECK9:       omp_offload.cont3:
4113 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* @Arg, align 4
4114 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
4115 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
4116 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
4117 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4118 // CHECK9-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
4119 // CHECK9-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
4120 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4121 // CHECK9-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
4122 // CHECK9-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
4123 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4124 // CHECK9-NEXT:    store i8* null, i8** [[TMP10]], align 8
4125 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4126 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4127 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4128 // CHECK9-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 1)
4129 // CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4130 // CHECK9-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
4131 // CHECK9:       omp_offload.failed5:
4132 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99(i64 [[TMP5]]) #[[ATTR2]]
4133 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
4134 // CHECK9:       omp_offload.cont6:
4135 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
4136 // CHECK9-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
4137 // CHECK9-NEXT:    ret i32 [[CALL]]
4138 //
4139 //
4140 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
4141 // CHECK9-SAME: () #[[ATTR1]] {
4142 // CHECK9-NEXT:  entry:
4143 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
4144 // CHECK9-NEXT:    ret void
4145 //
4146 //
4147 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
4148 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4149 // CHECK9-NEXT:  entry:
4150 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4151 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4152 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4153 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4154 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4155 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4156 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4157 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4158 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4159 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4160 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4161 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4162 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4163 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4164 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4165 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4166 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4167 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4168 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4169 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4170 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4171 // CHECK9:       cond.true:
4172 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4173 // CHECK9:       cond.false:
4174 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4175 // CHECK9-NEXT:    br label [[COND_END]]
4176 // CHECK9:       cond.end:
4177 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4178 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4179 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4180 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4181 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4182 // CHECK9:       omp.inner.for.cond:
4183 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
4184 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
4185 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4186 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4187 // CHECK9:       omp.inner.for.body:
4188 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
4189 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4190 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
4191 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4192 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group !30
4193 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4194 // CHECK9:       omp.inner.for.inc:
4195 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
4196 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
4197 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
4198 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
4199 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
4200 // CHECK9:       omp.inner.for.end:
4201 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4202 // CHECK9:       omp.loop.exit:
4203 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4204 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4205 // CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4206 // CHECK9-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4207 // CHECK9:       .omp.final.then:
4208 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
4209 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4210 // CHECK9:       .omp.final.done:
4211 // CHECK9-NEXT:    ret void
4212 //
4213 //
4214 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
4215 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4216 // CHECK9-NEXT:  entry:
4217 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4218 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4219 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4220 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4221 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4222 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4223 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4224 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4225 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4226 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4227 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4228 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4229 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4230 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4231 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4232 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4233 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
4234 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4235 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4236 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4237 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4238 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4239 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4240 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4241 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4242 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4243 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4244 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4245 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4246 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4247 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4248 // CHECK9:       cond.true:
4249 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4250 // CHECK9:       cond.false:
4251 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4252 // CHECK9-NEXT:    br label [[COND_END]]
4253 // CHECK9:       cond.end:
4254 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4255 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4256 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4257 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4258 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4259 // CHECK9:       omp.inner.for.cond:
4260 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
4261 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
4262 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4263 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4264 // CHECK9:       omp.inner.for.body:
4265 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
4266 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4267 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4268 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
4269 // CHECK9-NEXT:    call void @_Z3fn4v(), !llvm.access.group !33
4270 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4271 // CHECK9:       omp.body.continue:
4272 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4273 // CHECK9:       omp.inner.for.inc:
4274 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
4275 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4276 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
4277 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
4278 // CHECK9:       omp.inner.for.end:
4279 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4280 // CHECK9:       omp.loop.exit:
4281 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4282 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4283 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4284 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4285 // CHECK9:       .omp.final.then:
4286 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
4287 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4288 // CHECK9:       .omp.final.done:
4289 // CHECK9-NEXT:    ret void
4290 //
4291 //
4292 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90
4293 // CHECK9-SAME: () #[[ATTR1]] {
4294 // CHECK9-NEXT:  entry:
4295 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..6 to void (i32*, i32*, ...)*))
4296 // CHECK9-NEXT:    ret void
4297 //
4298 //
4299 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..6
4300 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4301 // CHECK9-NEXT:  entry:
4302 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4303 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4304 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4305 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4306 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4307 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4308 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4309 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4310 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4311 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
4312 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4313 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4314 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4315 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4316 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4317 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4318 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4319 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4320 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4321 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4322 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4323 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4324 // CHECK9:       cond.true:
4325 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4326 // CHECK9:       cond.false:
4327 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4328 // CHECK9-NEXT:    br label [[COND_END]]
4329 // CHECK9:       cond.end:
4330 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4331 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4332 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4333 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4334 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4335 // CHECK9:       omp.inner.for.cond:
4336 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
4337 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
4338 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4339 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4340 // CHECK9:       omp.inner.for.body:
4341 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !36
4342 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4343 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
4344 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4345 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !36
4346 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !36
4347 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !36
4348 // CHECK9-NEXT:    call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !36
4349 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !36
4350 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4351 // CHECK9:       omp.inner.for.inc:
4352 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
4353 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !36
4354 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4355 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
4356 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
4357 // CHECK9:       omp.inner.for.end:
4358 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4359 // CHECK9:       omp.loop.exit:
4360 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4361 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4362 // CHECK9-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
4363 // CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4364 // CHECK9:       .omp.final.then:
4365 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
4366 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4367 // CHECK9:       .omp.final.done:
4368 // CHECK9-NEXT:    ret void
4369 //
4370 //
4371 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..7
4372 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4373 // CHECK9-NEXT:  entry:
4374 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4375 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4376 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4377 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4378 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4379 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4380 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4381 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4382 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4383 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4384 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4385 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4386 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4387 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4388 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4389 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4390 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
4391 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4392 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4393 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4394 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4395 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4396 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4397 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4398 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4399 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4400 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4401 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4402 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4403 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4404 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4405 // CHECK9:       cond.true:
4406 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4407 // CHECK9:       cond.false:
4408 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4409 // CHECK9-NEXT:    br label [[COND_END]]
4410 // CHECK9:       cond.end:
4411 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4412 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4413 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4414 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4415 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4416 // CHECK9:       omp.inner.for.cond:
4417 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
4418 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
4419 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4420 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4421 // CHECK9:       omp.inner.for.body:
4422 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
4423 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4424 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4425 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
4426 // CHECK9-NEXT:    call void @_Z3fn5v(), !llvm.access.group !39
4427 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4428 // CHECK9:       omp.body.continue:
4429 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4430 // CHECK9:       omp.inner.for.inc:
4431 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
4432 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4433 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
4434 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
4435 // CHECK9:       omp.inner.for.end:
4436 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4437 // CHECK9:       omp.loop.exit:
4438 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4439 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4440 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4441 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4442 // CHECK9:       .omp.final.then:
4443 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
4444 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4445 // CHECK9:       .omp.final.done:
4446 // CHECK9-NEXT:    ret void
4447 //
4448 //
4449 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99
4450 // CHECK9-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
4451 // CHECK9-NEXT:  entry:
4452 // CHECK9-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
4453 // CHECK9-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
4454 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
4455 // 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*, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]])
4456 // CHECK9-NEXT:    ret void
4457 //
4458 //
4459 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..8
4460 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
4461 // CHECK9-NEXT:  entry:
4462 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4463 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4464 // CHECK9-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
4465 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4466 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4467 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4468 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4469 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4470 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4471 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4472 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
4473 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4474 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4475 // CHECK9-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
4476 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
4477 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4478 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4479 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4480 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4481 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4482 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4483 // 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)
4484 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4485 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
4486 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4487 // CHECK9:       cond.true:
4488 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4489 // CHECK9:       cond.false:
4490 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4491 // CHECK9-NEXT:    br label [[COND_END]]
4492 // CHECK9:       cond.end:
4493 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4494 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4495 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4496 // CHECK9-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4497 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4498 // CHECK9:       omp.inner.for.cond:
4499 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
4500 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !42
4501 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4502 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4503 // CHECK9:       omp.inner.for.body:
4504 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !42
4505 // CHECK9-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
4506 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !42
4507 // CHECK9-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
4508 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !42
4509 // CHECK9-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
4510 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4511 // CHECK9:       omp_if.then:
4512 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]]), !llvm.access.group !42
4513 // CHECK9-NEXT:    br label [[OMP_IF_END:%.*]]
4514 // CHECK9:       omp_if.else:
4515 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !42
4516 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !42
4517 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !42
4518 // CHECK9-NEXT:    call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !42
4519 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !42
4520 // CHECK9-NEXT:    br label [[OMP_IF_END]]
4521 // CHECK9:       omp_if.end:
4522 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4523 // CHECK9:       omp.inner.for.inc:
4524 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
4525 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !42
4526 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
4527 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
4528 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP43:![0-9]+]]
4529 // CHECK9:       omp.inner.for.end:
4530 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4531 // CHECK9:       omp.loop.exit:
4532 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4533 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4534 // CHECK9-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
4535 // CHECK9-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4536 // CHECK9:       .omp.final.then:
4537 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
4538 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4539 // CHECK9:       .omp.final.done:
4540 // CHECK9-NEXT:    ret void
4541 //
4542 //
4543 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..9
4544 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4545 // CHECK9-NEXT:  entry:
4546 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4547 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4548 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4549 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4550 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4551 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4552 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4553 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4554 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4555 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4556 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4557 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4558 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4559 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4560 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4561 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4562 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
4563 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4564 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4565 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4566 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4567 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4568 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4569 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4570 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4571 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4572 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4573 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4574 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4575 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4576 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4577 // CHECK9:       cond.true:
4578 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4579 // CHECK9:       cond.false:
4580 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4581 // CHECK9-NEXT:    br label [[COND_END]]
4582 // CHECK9:       cond.end:
4583 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4584 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4585 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4586 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4587 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4588 // CHECK9:       omp.inner.for.cond:
4589 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
4590 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !45
4591 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4592 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4593 // CHECK9:       omp.inner.for.body:
4594 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
4595 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4596 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4597 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !45
4598 // CHECK9-NEXT:    call void @_Z3fn6v(), !llvm.access.group !45
4599 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4600 // CHECK9:       omp.body.continue:
4601 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4602 // CHECK9:       omp.inner.for.inc:
4603 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
4604 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4605 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
4606 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP46:![0-9]+]]
4607 // CHECK9:       omp.inner.for.end:
4608 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4609 // CHECK9:       omp.loop.exit:
4610 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4611 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4612 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4613 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4614 // CHECK9:       .omp.final.then:
4615 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
4616 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4617 // CHECK9:       .omp.final.done:
4618 // CHECK9-NEXT:    ret void
4619 //
4620 //
4621 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
4622 // CHECK9-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
4623 // CHECK9-NEXT:  entry:
4624 // CHECK9-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
4625 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4626 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
4627 // CHECK9-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
4628 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
4629 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
4630 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
4631 // CHECK9-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
4632 // CHECK9-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
4633 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4634 // CHECK9-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
4635 // CHECK9-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
4636 // CHECK9-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4637 // CHECK9:       omp_offload.failed:
4638 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59() #[[ATTR2]]
4639 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4640 // CHECK9:       omp_offload.cont:
4641 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4642 // CHECK9-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
4643 // CHECK9-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
4644 // CHECK9-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
4645 // CHECK9:       omp_offload.failed2:
4646 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65() #[[ATTR2]]
4647 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
4648 // CHECK9:       omp_offload.cont3:
4649 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
4650 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
4651 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
4652 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
4653 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4654 // CHECK9-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
4655 // CHECK9-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
4656 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4657 // CHECK9-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
4658 // CHECK9-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
4659 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4660 // CHECK9-NEXT:    store i8* null, i8** [[TMP10]], align 8
4661 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4662 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4663 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4664 // CHECK9-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.16, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.17, i32 0, i32 0), i8** null, i8** null, i32 0, i32 1)
4665 // CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4666 // CHECK9-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
4667 // CHECK9:       omp_offload.failed5:
4668 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71(i64 [[TMP5]]) #[[ATTR2]]
4669 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
4670 // CHECK9:       omp_offload.cont6:
4671 // CHECK9-NEXT:    ret i32 0
4672 //
4673 //
4674 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59
4675 // CHECK9-SAME: () #[[ATTR1]] {
4676 // CHECK9-NEXT:  entry:
4677 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..10 to void (i32*, i32*, ...)*))
4678 // CHECK9-NEXT:    ret void
4679 //
4680 //
4681 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..10
4682 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4683 // CHECK9-NEXT:  entry:
4684 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4685 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4686 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4687 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4688 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4689 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4690 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4691 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4692 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4693 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4694 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4695 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4696 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4697 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4698 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4699 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4700 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4701 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4702 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4703 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4704 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4705 // CHECK9:       cond.true:
4706 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4707 // CHECK9:       cond.false:
4708 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4709 // CHECK9-NEXT:    br label [[COND_END]]
4710 // CHECK9:       cond.end:
4711 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4712 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4713 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4714 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4715 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4716 // CHECK9:       omp.inner.for.cond:
4717 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !48
4718 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !48
4719 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4720 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4721 // CHECK9:       omp.inner.for.body:
4722 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !48
4723 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4724 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !48
4725 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4726 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..11 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group !48
4727 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4728 // CHECK9:       omp.inner.for.inc:
4729 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !48
4730 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !48
4731 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
4732 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !48
4733 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP49:![0-9]+]]
4734 // CHECK9:       omp.inner.for.end:
4735 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4736 // CHECK9:       omp.loop.exit:
4737 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4738 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4739 // CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4740 // CHECK9-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4741 // CHECK9:       .omp.final.then:
4742 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
4743 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4744 // CHECK9:       .omp.final.done:
4745 // CHECK9-NEXT:    ret void
4746 //
4747 //
4748 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..11
4749 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4750 // CHECK9-NEXT:  entry:
4751 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4752 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4753 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4754 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4755 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4756 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4757 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4758 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4759 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4760 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4761 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4762 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4763 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4764 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4765 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4766 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4767 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
4768 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4769 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4770 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4771 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4772 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4773 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4774 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4775 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4776 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4777 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4778 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4779 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4780 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4781 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4782 // CHECK9:       cond.true:
4783 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4784 // CHECK9:       cond.false:
4785 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4786 // CHECK9-NEXT:    br label [[COND_END]]
4787 // CHECK9:       cond.end:
4788 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4789 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4790 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4791 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4792 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4793 // CHECK9:       omp.inner.for.cond:
4794 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
4795 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !51
4796 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4797 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4798 // CHECK9:       omp.inner.for.body:
4799 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
4800 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4801 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4802 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !51
4803 // CHECK9-NEXT:    call void @_Z3fn1v(), !llvm.access.group !51
4804 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4805 // CHECK9:       omp.body.continue:
4806 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4807 // CHECK9:       omp.inner.for.inc:
4808 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
4809 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4810 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
4811 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP52:![0-9]+]]
4812 // CHECK9:       omp.inner.for.end:
4813 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4814 // CHECK9:       omp.loop.exit:
4815 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4816 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4817 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4818 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4819 // CHECK9:       .omp.final.then:
4820 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
4821 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4822 // CHECK9:       .omp.final.done:
4823 // CHECK9-NEXT:    ret void
4824 //
4825 //
4826 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65
4827 // CHECK9-SAME: () #[[ATTR1]] {
4828 // CHECK9-NEXT:  entry:
4829 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..12 to void (i32*, i32*, ...)*))
4830 // CHECK9-NEXT:    ret void
4831 //
4832 //
4833 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..12
4834 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4835 // CHECK9-NEXT:  entry:
4836 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4837 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4838 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4839 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4840 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4841 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4842 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4843 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4844 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4845 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
4846 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4847 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4848 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4849 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4850 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4851 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4852 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4853 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4854 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4855 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4856 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4857 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4858 // CHECK9:       cond.true:
4859 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4860 // CHECK9:       cond.false:
4861 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4862 // CHECK9-NEXT:    br label [[COND_END]]
4863 // CHECK9:       cond.end:
4864 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4865 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4866 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4867 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4868 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4869 // CHECK9:       omp.inner.for.cond:
4870 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
4871 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !54
4872 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4873 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4874 // CHECK9:       omp.inner.for.body:
4875 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !54
4876 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4877 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !54
4878 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4879 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !54
4880 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !54
4881 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !54
4882 // CHECK9-NEXT:    call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !54
4883 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !54
4884 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4885 // CHECK9:       omp.inner.for.inc:
4886 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
4887 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !54
4888 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4889 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
4890 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP55:![0-9]+]]
4891 // CHECK9:       omp.inner.for.end:
4892 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4893 // CHECK9:       omp.loop.exit:
4894 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4895 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4896 // CHECK9-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
4897 // CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4898 // CHECK9:       .omp.final.then:
4899 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
4900 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4901 // CHECK9:       .omp.final.done:
4902 // CHECK9-NEXT:    ret void
4903 //
4904 //
4905 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..13
4906 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4907 // CHECK9-NEXT:  entry:
4908 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4909 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4910 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4911 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4912 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4913 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4914 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4915 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4916 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4917 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4918 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
4919 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4920 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4921 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4922 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4923 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4924 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
4925 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4926 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4927 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4928 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4929 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4930 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4931 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4932 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4933 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4934 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4935 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4936 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4937 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4938 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4939 // CHECK9:       cond.true:
4940 // CHECK9-NEXT:    br label [[COND_END:%.*]]
4941 // CHECK9:       cond.false:
4942 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4943 // CHECK9-NEXT:    br label [[COND_END]]
4944 // CHECK9:       cond.end:
4945 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4946 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4947 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4948 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4949 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4950 // CHECK9:       omp.inner.for.cond:
4951 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !57
4952 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !57
4953 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4954 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4955 // CHECK9:       omp.inner.for.body:
4956 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !57
4957 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4958 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4959 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !57
4960 // CHECK9-NEXT:    call void @_Z3fn2v(), !llvm.access.group !57
4961 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4962 // CHECK9:       omp.body.continue:
4963 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4964 // CHECK9:       omp.inner.for.inc:
4965 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !57
4966 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4967 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !57
4968 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP58:![0-9]+]]
4969 // CHECK9:       omp.inner.for.end:
4970 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4971 // CHECK9:       omp.loop.exit:
4972 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4973 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4974 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4975 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4976 // CHECK9:       .omp.final.then:
4977 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
4978 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4979 // CHECK9:       .omp.final.done:
4980 // CHECK9-NEXT:    ret void
4981 //
4982 //
4983 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71
4984 // CHECK9-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
4985 // CHECK9-NEXT:  entry:
4986 // CHECK9-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
4987 // CHECK9-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
4988 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
4989 // 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*, i32*)* @.omp_outlined..14 to void (i32*, i32*, ...)*), i32* [[CONV]])
4990 // CHECK9-NEXT:    ret void
4991 //
4992 //
4993 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..14
4994 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
4995 // CHECK9-NEXT:  entry:
4996 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4997 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4998 // CHECK9-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
4999 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5000 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5001 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5002 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5003 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5004 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5005 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
5006 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
5007 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5008 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5009 // CHECK9-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
5010 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
5011 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5012 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5013 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5014 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5015 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5016 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
5017 // 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)
5018 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5019 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
5020 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5021 // CHECK9:       cond.true:
5022 // CHECK9-NEXT:    br label [[COND_END:%.*]]
5023 // CHECK9:       cond.false:
5024 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5025 // CHECK9-NEXT:    br label [[COND_END]]
5026 // CHECK9:       cond.end:
5027 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
5028 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5029 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5030 // CHECK9-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
5031 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5032 // CHECK9:       omp.inner.for.cond:
5033 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !60
5034 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !60
5035 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
5036 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5037 // CHECK9:       omp.inner.for.body:
5038 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !60
5039 // CHECK9-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
5040 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !60
5041 // CHECK9-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
5042 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !60
5043 // CHECK9-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
5044 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
5045 // CHECK9:       omp_if.then:
5046 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]]), !llvm.access.group !60
5047 // CHECK9-NEXT:    br label [[OMP_IF_END:%.*]]
5048 // CHECK9:       omp_if.else:
5049 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !60
5050 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !60
5051 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !60
5052 // CHECK9-NEXT:    call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !60
5053 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !60
5054 // CHECK9-NEXT:    br label [[OMP_IF_END]]
5055 // CHECK9:       omp_if.end:
5056 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5057 // CHECK9:       omp.inner.for.inc:
5058 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !60
5059 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !60
5060 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
5061 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !60
5062 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP61:![0-9]+]]
5063 // CHECK9:       omp.inner.for.end:
5064 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5065 // CHECK9:       omp.loop.exit:
5066 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
5067 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5068 // CHECK9-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
5069 // CHECK9-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5070 // CHECK9:       .omp.final.then:
5071 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
5072 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5073 // CHECK9:       .omp.final.done:
5074 // CHECK9-NEXT:    ret void
5075 //
5076 //
5077 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..15
5078 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5079 // CHECK9-NEXT:  entry:
5080 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5081 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5082 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5083 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5084 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5085 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5086 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5087 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5088 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5089 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5090 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
5091 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5092 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5093 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5094 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5095 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5096 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
5097 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5098 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5099 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5100 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5101 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5102 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5103 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5104 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5105 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5106 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5107 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5108 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5109 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5110 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5111 // CHECK9:       cond.true:
5112 // CHECK9-NEXT:    br label [[COND_END:%.*]]
5113 // CHECK9:       cond.false:
5114 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5115 // CHECK9-NEXT:    br label [[COND_END]]
5116 // CHECK9:       cond.end:
5117 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5118 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5119 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5120 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5121 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5122 // CHECK9:       omp.inner.for.cond:
5123 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !63
5124 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !63
5125 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5126 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5127 // CHECK9:       omp.inner.for.body:
5128 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !63
5129 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5130 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5131 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !63
5132 // CHECK9-NEXT:    call void @_Z3fn3v(), !llvm.access.group !63
5133 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5134 // CHECK9:       omp.body.continue:
5135 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5136 // CHECK9:       omp.inner.for.inc:
5137 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !63
5138 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5139 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !63
5140 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP64:![0-9]+]]
5141 // CHECK9:       omp.inner.for.end:
5142 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5143 // CHECK9:       omp.loop.exit:
5144 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5145 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5146 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
5147 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5148 // CHECK9:       .omp.final.then:
5149 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
5150 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5151 // CHECK9:       .omp.final.done:
5152 // CHECK9-NEXT:    ret void
5153 //
5154 //
5155 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
5156 // CHECK9-SAME: () #[[ATTR5:[0-9]+]] {
5157 // CHECK9-NEXT:  entry:
5158 // CHECK9-NEXT:    call void @__tgt_register_requires(i64 1)
5159 // CHECK9-NEXT:    ret void
5160 //
5161 //
5162 // CHECK11-LABEL: define {{[^@]+}}@_Z9gtid_testv
5163 // CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
5164 // CHECK11-NEXT:  entry:
5165 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5166 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
5167 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
5168 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
5169 // CHECK11-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
5170 // CHECK11-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5171 // CHECK11:       omp_offload.failed:
5172 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43() #[[ATTR2:[0-9]+]]
5173 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
5174 // CHECK11:       omp_offload.cont:
5175 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
5176 // CHECK11-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
5177 // CHECK11-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
5178 // CHECK11-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
5179 // CHECK11:       omp_offload.failed2:
5180 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48() #[[ATTR2]]
5181 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
5182 // CHECK11:       omp_offload.cont3:
5183 // CHECK11-NEXT:    ret void
5184 //
5185 //
5186 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43
5187 // CHECK11-SAME: () #[[ATTR1:[0-9]+]] {
5188 // CHECK11-NEXT:  entry:
5189 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
5190 // CHECK11-NEXT:    ret void
5191 //
5192 //
5193 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
5194 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
5195 // CHECK11-NEXT:  entry:
5196 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5197 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5198 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5199 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5200 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5201 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5202 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5203 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5204 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
5205 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5206 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5207 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5208 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5209 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5210 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5211 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5212 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5213 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5214 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5215 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5216 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5217 // CHECK11:       cond.true:
5218 // CHECK11-NEXT:    br label [[COND_END:%.*]]
5219 // CHECK11:       cond.false:
5220 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5221 // CHECK11-NEXT:    br label [[COND_END]]
5222 // CHECK11:       cond.end:
5223 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5224 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5225 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5226 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5227 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5228 // CHECK11:       omp.inner.for.cond:
5229 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
5230 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
5231 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5232 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5233 // CHECK11:       omp.inner.for.body:
5234 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !15
5235 // CHECK11-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5236 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
5237 // CHECK11-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5238 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group !15
5239 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5240 // CHECK11:       omp.inner.for.inc:
5241 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
5242 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !15
5243 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
5244 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
5245 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
5246 // CHECK11:       omp.inner.for.end:
5247 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5248 // CHECK11:       omp.loop.exit:
5249 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5250 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5251 // CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
5252 // CHECK11-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5253 // CHECK11:       .omp.final.then:
5254 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
5255 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5256 // CHECK11:       .omp.final.done:
5257 // CHECK11-NEXT:    ret void
5258 //
5259 //
5260 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
5261 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5262 // CHECK11-NEXT:  entry:
5263 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5264 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5265 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5266 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5267 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5268 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5269 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5270 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5271 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5272 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5273 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
5274 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5275 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5276 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5277 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5278 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5279 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
5280 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5281 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5282 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5283 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5284 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5285 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5286 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5287 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5288 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5289 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5290 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5291 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5292 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5293 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5294 // CHECK11:       cond.true:
5295 // CHECK11-NEXT:    br label [[COND_END:%.*]]
5296 // CHECK11:       cond.false:
5297 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5298 // CHECK11-NEXT:    br label [[COND_END]]
5299 // CHECK11:       cond.end:
5300 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5301 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5302 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5303 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5304 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5305 // CHECK11:       omp.inner.for.cond:
5306 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
5307 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
5308 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5309 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5310 // CHECK11:       omp.inner.for.body:
5311 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
5312 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5313 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5314 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !19
5315 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5316 // CHECK11:       omp.body.continue:
5317 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5318 // CHECK11:       omp.inner.for.inc:
5319 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
5320 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5321 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
5322 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
5323 // CHECK11:       omp.inner.for.end:
5324 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5325 // CHECK11:       omp.loop.exit:
5326 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5327 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5328 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
5329 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5330 // CHECK11:       .omp.final.then:
5331 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
5332 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5333 // CHECK11:       .omp.final.done:
5334 // CHECK11-NEXT:    ret void
5335 //
5336 //
5337 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48
5338 // CHECK11-SAME: () #[[ATTR1]] {
5339 // CHECK11-NEXT:  entry:
5340 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
5341 // CHECK11-NEXT:    ret void
5342 //
5343 //
5344 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
5345 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
5346 // CHECK11-NEXT:  entry:
5347 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5348 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5349 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5350 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5351 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5352 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5353 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5354 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5355 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
5356 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
5357 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5358 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5359 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5360 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5361 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5362 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5363 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5364 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5365 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5366 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5367 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5368 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5369 // CHECK11:       cond.true:
5370 // CHECK11-NEXT:    br label [[COND_END:%.*]]
5371 // CHECK11:       cond.false:
5372 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5373 // CHECK11-NEXT:    br label [[COND_END]]
5374 // CHECK11:       cond.end:
5375 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5376 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5377 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5378 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5379 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5380 // CHECK11:       omp.inner.for.cond:
5381 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
5382 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
5383 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5384 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5385 // CHECK11:       omp.inner.for.body:
5386 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
5387 // CHECK11-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5388 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
5389 // CHECK11-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5390 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !24
5391 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !24
5392 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !24
5393 // CHECK11-NEXT:    call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !24
5394 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !24
5395 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5396 // CHECK11:       omp.inner.for.inc:
5397 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
5398 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
5399 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5400 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
5401 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
5402 // CHECK11:       omp.inner.for.end:
5403 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5404 // CHECK11:       omp.loop.exit:
5405 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5406 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5407 // CHECK11-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
5408 // CHECK11-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5409 // CHECK11:       .omp.final.then:
5410 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
5411 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5412 // CHECK11:       .omp.final.done:
5413 // CHECK11-NEXT:    ret void
5414 //
5415 //
5416 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
5417 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5418 // CHECK11-NEXT:  entry:
5419 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5420 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5421 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5422 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5423 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5424 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5425 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5426 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5427 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5428 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5429 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
5430 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5431 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5432 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5433 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5434 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5435 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
5436 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5437 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5438 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5439 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5440 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5441 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5442 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5443 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5444 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5445 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5446 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5447 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5448 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5449 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5450 // CHECK11:       cond.true:
5451 // CHECK11-NEXT:    br label [[COND_END:%.*]]
5452 // CHECK11:       cond.false:
5453 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5454 // CHECK11-NEXT:    br label [[COND_END]]
5455 // CHECK11:       cond.end:
5456 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5457 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5458 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5459 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5460 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5461 // CHECK11:       omp.inner.for.cond:
5462 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
5463 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
5464 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5465 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5466 // CHECK11:       omp.inner.for.body:
5467 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
5468 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5469 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5470 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !27
5471 // CHECK11-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !27
5472 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5473 // CHECK11:       omp.body.continue:
5474 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5475 // CHECK11:       omp.inner.for.inc:
5476 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
5477 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5478 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
5479 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
5480 // CHECK11:       omp.inner.for.end:
5481 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5482 // CHECK11:       omp.loop.exit:
5483 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5484 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5485 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
5486 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5487 // CHECK11:       .omp.final.then:
5488 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
5489 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5490 // CHECK11:       .omp.final.done:
5491 // CHECK11-NEXT:    ret void
5492 //
5493 //
5494 // CHECK11-LABEL: define {{[^@]+}}@main
5495 // CHECK11-SAME: () #[[ATTR3:[0-9]+]] {
5496 // CHECK11-NEXT:  entry:
5497 // CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
5498 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5499 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
5500 // CHECK11-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
5501 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
5502 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
5503 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
5504 // CHECK11-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
5505 // CHECK11-NEXT:    store i32 0, i32* [[RETVAL]], align 4
5506 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
5507 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
5508 // CHECK11-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
5509 // CHECK11-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5510 // CHECK11:       omp_offload.failed:
5511 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81() #[[ATTR2]]
5512 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
5513 // CHECK11:       omp_offload.cont:
5514 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
5515 // CHECK11-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
5516 // CHECK11-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
5517 // CHECK11-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
5518 // CHECK11:       omp_offload.failed2:
5519 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90() #[[ATTR2]]
5520 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
5521 // CHECK11:       omp_offload.cont3:
5522 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* @Arg, align 4
5523 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
5524 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
5525 // CHECK11-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
5526 // CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5527 // CHECK11-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
5528 // CHECK11-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
5529 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5530 // CHECK11-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
5531 // CHECK11-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
5532 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
5533 // CHECK11-NEXT:    store i8* null, i8** [[TMP10]], align 8
5534 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5535 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5536 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
5537 // CHECK11-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 1)
5538 // CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
5539 // CHECK11-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
5540 // CHECK11:       omp_offload.failed5:
5541 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99(i64 [[TMP5]]) #[[ATTR2]]
5542 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
5543 // CHECK11:       omp_offload.cont6:
5544 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
5545 // CHECK11-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
5546 // CHECK11-NEXT:    ret i32 [[CALL]]
5547 //
5548 //
5549 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
5550 // CHECK11-SAME: () #[[ATTR1]] {
5551 // CHECK11-NEXT:  entry:
5552 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
5553 // CHECK11-NEXT:    ret void
5554 //
5555 //
5556 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
5557 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
5558 // CHECK11-NEXT:  entry:
5559 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5560 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5561 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5562 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5563 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5564 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5565 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5566 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5567 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
5568 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5569 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5570 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5571 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5572 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5573 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5574 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5575 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5576 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5577 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5578 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5579 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5580 // CHECK11:       cond.true:
5581 // CHECK11-NEXT:    br label [[COND_END:%.*]]
5582 // CHECK11:       cond.false:
5583 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5584 // CHECK11-NEXT:    br label [[COND_END]]
5585 // CHECK11:       cond.end:
5586 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5587 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5588 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5589 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5590 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5591 // CHECK11:       omp.inner.for.cond:
5592 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
5593 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
5594 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5595 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5596 // CHECK11:       omp.inner.for.body:
5597 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
5598 // CHECK11-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5599 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
5600 // CHECK11-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5601 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group !30
5602 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5603 // CHECK11:       omp.inner.for.inc:
5604 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
5605 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
5606 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
5607 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
5608 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
5609 // CHECK11:       omp.inner.for.end:
5610 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5611 // CHECK11:       omp.loop.exit:
5612 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5613 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5614 // CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
5615 // CHECK11-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5616 // CHECK11:       .omp.final.then:
5617 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
5618 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5619 // CHECK11:       .omp.final.done:
5620 // CHECK11-NEXT:    ret void
5621 //
5622 //
5623 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
5624 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5625 // CHECK11-NEXT:  entry:
5626 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5627 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5628 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5629 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5630 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5631 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5632 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5633 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5634 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5635 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5636 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
5637 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5638 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5639 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5640 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5641 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5642 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
5643 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5644 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5645 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5646 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5647 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5648 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5649 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5650 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5651 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5652 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5653 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5654 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5655 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5656 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5657 // CHECK11:       cond.true:
5658 // CHECK11-NEXT:    br label [[COND_END:%.*]]
5659 // CHECK11:       cond.false:
5660 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5661 // CHECK11-NEXT:    br label [[COND_END]]
5662 // CHECK11:       cond.end:
5663 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5664 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5665 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5666 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5667 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5668 // CHECK11:       omp.inner.for.cond:
5669 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
5670 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
5671 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5672 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5673 // CHECK11:       omp.inner.for.body:
5674 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
5675 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5676 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5677 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
5678 // CHECK11-NEXT:    call void @_Z3fn4v(), !llvm.access.group !33
5679 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5680 // CHECK11:       omp.body.continue:
5681 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5682 // CHECK11:       omp.inner.for.inc:
5683 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
5684 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5685 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
5686 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
5687 // CHECK11:       omp.inner.for.end:
5688 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5689 // CHECK11:       omp.loop.exit:
5690 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5691 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5692 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
5693 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5694 // CHECK11:       .omp.final.then:
5695 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
5696 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5697 // CHECK11:       .omp.final.done:
5698 // CHECK11-NEXT:    ret void
5699 //
5700 //
5701 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90
5702 // CHECK11-SAME: () #[[ATTR1]] {
5703 // CHECK11-NEXT:  entry:
5704 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..6 to void (i32*, i32*, ...)*))
5705 // CHECK11-NEXT:    ret void
5706 //
5707 //
5708 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..6
5709 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
5710 // CHECK11-NEXT:  entry:
5711 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5712 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5713 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5714 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5715 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5716 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5717 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5718 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5719 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
5720 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
5721 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5722 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5723 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5724 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5725 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5726 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5727 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5728 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5729 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5730 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5731 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5732 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5733 // CHECK11:       cond.true:
5734 // CHECK11-NEXT:    br label [[COND_END:%.*]]
5735 // CHECK11:       cond.false:
5736 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5737 // CHECK11-NEXT:    br label [[COND_END]]
5738 // CHECK11:       cond.end:
5739 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5740 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5741 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5742 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5743 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5744 // CHECK11:       omp.inner.for.cond:
5745 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5746 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5747 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5748 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5749 // CHECK11:       omp.inner.for.body:
5750 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5751 // CHECK11-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5752 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5753 // CHECK11-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5754 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
5755 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5756 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
5757 // CHECK11-NEXT:    call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
5758 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
5759 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5760 // CHECK11:       omp.inner.for.inc:
5761 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5762 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5763 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5764 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5765 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
5766 // CHECK11:       omp.inner.for.end:
5767 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5768 // CHECK11:       omp.loop.exit:
5769 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5770 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5771 // CHECK11-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
5772 // CHECK11-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5773 // CHECK11:       .omp.final.then:
5774 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
5775 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5776 // CHECK11:       .omp.final.done:
5777 // CHECK11-NEXT:    ret void
5778 //
5779 //
5780 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..7
5781 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5782 // CHECK11-NEXT:  entry:
5783 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5784 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5785 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5786 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5787 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5788 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5789 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5790 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5791 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5792 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5793 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
5794 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5795 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5796 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5797 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5798 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5799 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
5800 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5801 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5802 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5803 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5804 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5805 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5806 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5807 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5808 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5809 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5810 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5811 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5812 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5813 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5814 // CHECK11:       cond.true:
5815 // CHECK11-NEXT:    br label [[COND_END:%.*]]
5816 // CHECK11:       cond.false:
5817 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5818 // CHECK11-NEXT:    br label [[COND_END]]
5819 // CHECK11:       cond.end:
5820 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5821 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5822 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5823 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5824 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5825 // CHECK11:       omp.inner.for.cond:
5826 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5827 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5828 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5829 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5830 // CHECK11:       omp.inner.for.body:
5831 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5832 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5833 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5834 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
5835 // CHECK11-NEXT:    call void @_Z3fn5v()
5836 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5837 // CHECK11:       omp.body.continue:
5838 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5839 // CHECK11:       omp.inner.for.inc:
5840 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5841 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5842 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
5843 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP38:![0-9]+]]
5844 // CHECK11:       omp.inner.for.end:
5845 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5846 // CHECK11:       omp.loop.exit:
5847 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5848 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5849 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
5850 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5851 // CHECK11:       .omp.final.then:
5852 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
5853 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5854 // CHECK11:       .omp.final.done:
5855 // CHECK11-NEXT:    ret void
5856 //
5857 //
5858 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99
5859 // CHECK11-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
5860 // CHECK11-NEXT:  entry:
5861 // CHECK11-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
5862 // CHECK11-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
5863 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
5864 // 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*, i32*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), i32* [[CONV]])
5865 // CHECK11-NEXT:    ret void
5866 //
5867 //
5868 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..8
5869 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
5870 // CHECK11-NEXT:  entry:
5871 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5872 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5873 // CHECK11-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
5874 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
5875 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5876 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5877 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5878 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5879 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5880 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5881 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
5882 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
5883 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
5884 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__CASTED12:%.*]] = alloca i64, align 8
5885 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR18:%.*]] = alloca i32, align 4
5886 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5887 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5888 // CHECK11-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
5889 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
5890 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5891 // CHECK11-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
5892 // CHECK11-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
5893 // CHECK11-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
5894 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5895 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5896 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5897 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5898 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5899 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5900 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
5901 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5902 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5903 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5904 // CHECK11:       cond.true:
5905 // CHECK11-NEXT:    br label [[COND_END:%.*]]
5906 // CHECK11:       cond.false:
5907 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5908 // CHECK11-NEXT:    br label [[COND_END]]
5909 // CHECK11:       cond.end:
5910 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5911 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5912 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5913 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5914 // CHECK11-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
5915 // CHECK11-NEXT:    [[TOBOOL1:%.*]] = trunc i8 [[TMP7]] to i1
5916 // CHECK11-NEXT:    br i1 [[TOBOOL1]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE7:%.*]]
5917 // CHECK11:       omp_if.then:
5918 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5919 // CHECK11:       omp.inner.for.cond:
5920 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
5921 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !39
5922 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
5923 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5924 // CHECK11:       omp.inner.for.body:
5925 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !39
5926 // CHECK11-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
5927 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !39
5928 // CHECK11-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
5929 // CHECK11-NEXT:    [[TMP14:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !39
5930 // CHECK11-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP14]] to i1
5931 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
5932 // CHECK11-NEXT:    [[FROMBOOL4:%.*]] = zext i1 [[TOBOOL3]] to i8
5933 // CHECK11-NEXT:    store i8 [[FROMBOOL4]], i8* [[CONV]], align 1, !llvm.access.group !39
5934 // CHECK11-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !39
5935 // CHECK11-NEXT:    [[TMP16:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !39
5936 // CHECK11-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP16]] to i1
5937 // CHECK11-NEXT:    br i1 [[TOBOOL5]], label [[OMP_IF_THEN6:%.*]], label [[OMP_IF_ELSE:%.*]]
5938 // CHECK11:       omp_if.then6:
5939 // 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*, i64, i64, i64)* @.omp_outlined..9 to void (i32*, i32*, ...)*), i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]]), !llvm.access.group !39
5940 // CHECK11-NEXT:    br label [[OMP_IF_END:%.*]]
5941 // CHECK11:       omp_if.else:
5942 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]]), !llvm.access.group !39
5943 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !39
5944 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !39
5945 // CHECK11-NEXT:    call void @.omp_outlined..9(i32* [[TMP17]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]]) #[[ATTR2]], !llvm.access.group !39
5946 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]]), !llvm.access.group !39
5947 // CHECK11-NEXT:    br label [[OMP_IF_END]]
5948 // CHECK11:       omp_if.end:
5949 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5950 // CHECK11:       omp.inner.for.inc:
5951 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
5952 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !39
5953 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
5954 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
5955 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
5956 // CHECK11:       omp.inner.for.end:
5957 // CHECK11-NEXT:    br label [[OMP_IF_END23:%.*]]
5958 // CHECK11:       omp_if.else7:
5959 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
5960 // CHECK11:       omp.inner.for.cond8:
5961 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5962 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5963 // CHECK11-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
5964 // CHECK11-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END22:%.*]]
5965 // CHECK11:       omp.inner.for.body10:
5966 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5967 // CHECK11-NEXT:    [[TMP23:%.*]] = zext i32 [[TMP22]] to i64
5968 // CHECK11-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5969 // CHECK11-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP24]] to i64
5970 // CHECK11-NEXT:    [[TMP26:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
5971 // CHECK11-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP26]] to i1
5972 // CHECK11-NEXT:    [[CONV13:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED12]] to i8*
5973 // CHECK11-NEXT:    [[FROMBOOL14:%.*]] = zext i1 [[TOBOOL11]] to i8
5974 // CHECK11-NEXT:    store i8 [[FROMBOOL14]], i8* [[CONV13]], align 1
5975 // CHECK11-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED12]], align 8
5976 // CHECK11-NEXT:    [[TMP28:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
5977 // CHECK11-NEXT:    [[TOBOOL15:%.*]] = trunc i8 [[TMP28]] to i1
5978 // CHECK11-NEXT:    br i1 [[TOBOOL15]], label [[OMP_IF_THEN16:%.*]], label [[OMP_IF_ELSE17:%.*]]
5979 // CHECK11:       omp_if.then16:
5980 // 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*, i64, i64, i64)* @.omp_outlined..10 to void (i32*, i32*, ...)*), i64 [[TMP23]], i64 [[TMP25]], i64 [[TMP27]])
5981 // CHECK11-NEXT:    br label [[OMP_IF_END19:%.*]]
5982 // CHECK11:       omp_if.else17:
5983 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
5984 // CHECK11-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5985 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR18]], align 4
5986 // CHECK11-NEXT:    call void @.omp_outlined..10(i32* [[TMP29]], i32* [[DOTBOUND_ZERO_ADDR18]], i64 [[TMP23]], i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR2]]
5987 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
5988 // CHECK11-NEXT:    br label [[OMP_IF_END19]]
5989 // CHECK11:       omp_if.end19:
5990 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC20:%.*]]
5991 // CHECK11:       omp.inner.for.inc20:
5992 // CHECK11-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5993 // CHECK11-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5994 // CHECK11-NEXT:    [[ADD21:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
5995 // CHECK11-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
5996 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP42:![0-9]+]]
5997 // CHECK11:       omp.inner.for.end22:
5998 // CHECK11-NEXT:    br label [[OMP_IF_END23]]
5999 // CHECK11:       omp_if.end23:
6000 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6001 // CHECK11:       omp.loop.exit:
6002 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6003 // CHECK11-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6004 // CHECK11-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
6005 // CHECK11-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6006 // CHECK11:       .omp.final.then:
6007 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
6008 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6009 // CHECK11:       .omp.final.done:
6010 // CHECK11-NEXT:    ret void
6011 //
6012 //
6013 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..9
6014 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
6015 // CHECK11-NEXT:  entry:
6016 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6017 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6018 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6019 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6020 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
6021 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6022 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6023 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6024 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6025 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6026 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6027 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
6028 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6029 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6030 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6031 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6032 // CHECK11-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
6033 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
6034 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6035 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6036 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6037 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP0]] to i32
6038 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6039 // CHECK11-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
6040 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
6041 // CHECK11-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
6042 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6043 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6044 // CHECK11-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
6045 // CHECK11-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
6046 // CHECK11-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6047 // CHECK11:       omp_if.then:
6048 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6049 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
6050 // 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)
6051 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6052 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
6053 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6054 // CHECK11:       cond.true:
6055 // CHECK11-NEXT:    br label [[COND_END:%.*]]
6056 // CHECK11:       cond.false:
6057 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6058 // CHECK11-NEXT:    br label [[COND_END]]
6059 // CHECK11:       cond.end:
6060 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
6061 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6062 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6063 // CHECK11-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
6064 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6065 // CHECK11:       omp.inner.for.cond:
6066 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
6067 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !43
6068 // CHECK11-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
6069 // CHECK11-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6070 // CHECK11:       omp.inner.for.body:
6071 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
6072 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
6073 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6074 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !43
6075 // CHECK11-NEXT:    call void @_Z3fn6v(), !llvm.access.group !43
6076 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6077 // CHECK11:       omp.body.continue:
6078 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6079 // CHECK11:       omp.inner.for.inc:
6080 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
6081 // CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
6082 // CHECK11-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
6083 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
6084 // CHECK11:       omp.inner.for.end:
6085 // CHECK11-NEXT:    br label [[OMP_IF_END:%.*]]
6086 // CHECK11:       omp_if.else:
6087 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6088 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
6089 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6090 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6091 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP14]], 99
6092 // CHECK11-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
6093 // CHECK11:       cond.true6:
6094 // CHECK11-NEXT:    br label [[COND_END8:%.*]]
6095 // CHECK11:       cond.false7:
6096 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6097 // CHECK11-NEXT:    br label [[COND_END8]]
6098 // CHECK11:       cond.end8:
6099 // CHECK11-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP15]], [[COND_FALSE7]] ]
6100 // CHECK11-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_UB]], align 4
6101 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6102 // CHECK11-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
6103 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND10:%.*]]
6104 // CHECK11:       omp.inner.for.cond10:
6105 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6106 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6107 // CHECK11-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
6108 // CHECK11-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY12:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
6109 // CHECK11:       omp.inner.for.body12:
6110 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6111 // CHECK11-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP19]], 1
6112 // CHECK11-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
6113 // CHECK11-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
6114 // CHECK11-NEXT:    call void @_Z3fn6v()
6115 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
6116 // CHECK11:       omp.body.continue15:
6117 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
6118 // CHECK11:       omp.inner.for.inc16:
6119 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6120 // CHECK11-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
6121 // CHECK11-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
6122 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND10]], !llvm.loop [[LOOP46:![0-9]+]]
6123 // CHECK11:       omp.inner.for.end18:
6124 // CHECK11-NEXT:    br label [[OMP_IF_END]]
6125 // CHECK11:       omp_if.end:
6126 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6127 // CHECK11:       omp.loop.exit:
6128 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6129 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
6130 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
6131 // CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6132 // CHECK11-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
6133 // CHECK11-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6134 // CHECK11:       .omp.final.then:
6135 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
6136 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6137 // CHECK11:       .omp.final.done:
6138 // CHECK11-NEXT:    ret void
6139 //
6140 //
6141 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..10
6142 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
6143 // CHECK11-NEXT:  entry:
6144 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6145 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6146 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6147 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6148 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
6149 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6150 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6151 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6152 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6153 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6154 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6155 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
6156 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6157 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6158 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6159 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6160 // CHECK11-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
6161 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
6162 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6163 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6164 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6165 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP0]] to i32
6166 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6167 // CHECK11-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
6168 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
6169 // CHECK11-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
6170 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6171 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6172 // CHECK11-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
6173 // CHECK11-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
6174 // CHECK11-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6175 // CHECK11:       omp_if.then:
6176 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6177 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
6178 // 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)
6179 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6180 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
6181 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6182 // CHECK11:       cond.true:
6183 // CHECK11-NEXT:    br label [[COND_END:%.*]]
6184 // CHECK11:       cond.false:
6185 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6186 // CHECK11-NEXT:    br label [[COND_END]]
6187 // CHECK11:       cond.end:
6188 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
6189 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6190 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6191 // CHECK11-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
6192 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6193 // CHECK11:       omp.inner.for.cond:
6194 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
6195 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !47
6196 // CHECK11-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
6197 // CHECK11-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6198 // CHECK11:       omp.inner.for.body:
6199 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
6200 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
6201 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6202 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !47
6203 // CHECK11-NEXT:    call void @_Z3fn6v(), !llvm.access.group !47
6204 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6205 // CHECK11:       omp.body.continue:
6206 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6207 // CHECK11:       omp.inner.for.inc:
6208 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
6209 // CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
6210 // CHECK11-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
6211 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
6212 // CHECK11:       omp.inner.for.end:
6213 // CHECK11-NEXT:    br label [[OMP_IF_END:%.*]]
6214 // CHECK11:       omp_if.else:
6215 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6216 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
6217 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6218 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6219 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP14]], 99
6220 // CHECK11-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
6221 // CHECK11:       cond.true6:
6222 // CHECK11-NEXT:    br label [[COND_END8:%.*]]
6223 // CHECK11:       cond.false7:
6224 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6225 // CHECK11-NEXT:    br label [[COND_END8]]
6226 // CHECK11:       cond.end8:
6227 // CHECK11-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP15]], [[COND_FALSE7]] ]
6228 // CHECK11-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_UB]], align 4
6229 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6230 // CHECK11-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
6231 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND10:%.*]]
6232 // CHECK11:       omp.inner.for.cond10:
6233 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6234 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6235 // CHECK11-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
6236 // CHECK11-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY12:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
6237 // CHECK11:       omp.inner.for.body12:
6238 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6239 // CHECK11-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP19]], 1
6240 // CHECK11-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
6241 // CHECK11-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
6242 // CHECK11-NEXT:    call void @_Z3fn6v()
6243 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
6244 // CHECK11:       omp.body.continue15:
6245 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
6246 // CHECK11:       omp.inner.for.inc16:
6247 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6248 // CHECK11-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
6249 // CHECK11-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
6250 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND10]], !llvm.loop [[LOOP50:![0-9]+]]
6251 // CHECK11:       omp.inner.for.end18:
6252 // CHECK11-NEXT:    br label [[OMP_IF_END]]
6253 // CHECK11:       omp_if.end:
6254 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6255 // CHECK11:       omp.loop.exit:
6256 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6257 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
6258 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
6259 // CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6260 // CHECK11-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
6261 // CHECK11-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6262 // CHECK11:       .omp.final.then:
6263 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
6264 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6265 // CHECK11:       .omp.final.done:
6266 // CHECK11-NEXT:    ret void
6267 //
6268 //
6269 // CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
6270 // CHECK11-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
6271 // CHECK11-NEXT:  entry:
6272 // CHECK11-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
6273 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6274 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
6275 // CHECK11-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
6276 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
6277 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
6278 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
6279 // CHECK11-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
6280 // CHECK11-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
6281 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
6282 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
6283 // CHECK11-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
6284 // CHECK11-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
6285 // CHECK11:       omp_offload.failed:
6286 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59() #[[ATTR2]]
6287 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
6288 // CHECK11:       omp_offload.cont:
6289 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
6290 // CHECK11-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 1)
6291 // CHECK11-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
6292 // CHECK11-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
6293 // CHECK11:       omp_offload.failed2:
6294 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65() #[[ATTR2]]
6295 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
6296 // CHECK11:       omp_offload.cont3:
6297 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
6298 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
6299 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
6300 // CHECK11-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
6301 // CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6302 // CHECK11-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
6303 // CHECK11-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
6304 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6305 // CHECK11-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
6306 // CHECK11-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
6307 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
6308 // CHECK11-NEXT:    store i8* null, i8** [[TMP10]], align 8
6309 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
6310 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
6311 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
6312 // CHECK11-NEXT:    [[TMP13:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71.region_id, i32 1, i8** [[TMP11]], i8** [[TMP12]], i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_sizes.17, i32 0, i32 0), i64* getelementptr inbounds ([1 x i64], [1 x i64]* @.offload_maptypes.18, i32 0, i32 0), i8** null, i8** null, i32 0, i32 1)
6313 // CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
6314 // CHECK11-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
6315 // CHECK11:       omp_offload.failed5:
6316 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71(i64 [[TMP5]]) #[[ATTR2]]
6317 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
6318 // CHECK11:       omp_offload.cont6:
6319 // CHECK11-NEXT:    ret i32 0
6320 //
6321 //
6322 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59
6323 // CHECK11-SAME: () #[[ATTR1]] {
6324 // CHECK11-NEXT:  entry:
6325 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..11 to void (i32*, i32*, ...)*))
6326 // CHECK11-NEXT:    ret void
6327 //
6328 //
6329 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..11
6330 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
6331 // CHECK11-NEXT:  entry:
6332 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6333 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6334 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6335 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6336 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6337 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6338 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6339 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6340 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
6341 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6342 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6343 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6344 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
6345 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6346 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6347 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6348 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
6349 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6350 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6351 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
6352 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6353 // CHECK11:       cond.true:
6354 // CHECK11-NEXT:    br label [[COND_END:%.*]]
6355 // CHECK11:       cond.false:
6356 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6357 // CHECK11-NEXT:    br label [[COND_END]]
6358 // CHECK11:       cond.end:
6359 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
6360 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6361 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6362 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
6363 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6364 // CHECK11:       omp.inner.for.cond:
6365 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
6366 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !51
6367 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
6368 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6369 // CHECK11:       omp.inner.for.body:
6370 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !51
6371 // CHECK11-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
6372 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !51
6373 // CHECK11-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
6374 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..12 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]]), !llvm.access.group !51
6375 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6376 // CHECK11:       omp.inner.for.inc:
6377 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
6378 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !51
6379 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
6380 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
6381 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP52:![0-9]+]]
6382 // CHECK11:       omp.inner.for.end:
6383 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6384 // CHECK11:       omp.loop.exit:
6385 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
6386 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6387 // CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
6388 // CHECK11-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6389 // CHECK11:       .omp.final.then:
6390 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
6391 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6392 // CHECK11:       .omp.final.done:
6393 // CHECK11-NEXT:    ret void
6394 //
6395 //
6396 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..12
6397 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
6398 // CHECK11-NEXT:  entry:
6399 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6400 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6401 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6402 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6403 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6404 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6405 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6406 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6407 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6408 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6409 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
6410 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6411 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6412 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6413 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6414 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6415 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6416 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6417 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
6418 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6419 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
6420 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
6421 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
6422 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6423 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6424 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6425 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
6426 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6427 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6428 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
6429 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6430 // CHECK11:       cond.true:
6431 // CHECK11-NEXT:    br label [[COND_END:%.*]]
6432 // CHECK11:       cond.false:
6433 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6434 // CHECK11-NEXT:    br label [[COND_END]]
6435 // CHECK11:       cond.end:
6436 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6437 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6438 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6439 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6440 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6441 // CHECK11:       omp.inner.for.cond:
6442 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
6443 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !54
6444 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6445 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6446 // CHECK11:       omp.inner.for.body:
6447 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
6448 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6449 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6450 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !54
6451 // CHECK11-NEXT:    call void @_Z3fn1v(), !llvm.access.group !54
6452 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6453 // CHECK11:       omp.body.continue:
6454 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6455 // CHECK11:       omp.inner.for.inc:
6456 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
6457 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
6458 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
6459 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP55:![0-9]+]]
6460 // CHECK11:       omp.inner.for.end:
6461 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6462 // CHECK11:       omp.loop.exit:
6463 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6464 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6465 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
6466 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6467 // CHECK11:       .omp.final.then:
6468 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
6469 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6470 // CHECK11:       .omp.final.done:
6471 // CHECK11-NEXT:    ret void
6472 //
6473 //
6474 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65
6475 // CHECK11-SAME: () #[[ATTR1]] {
6476 // CHECK11-NEXT:  entry:
6477 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..13 to void (i32*, i32*, ...)*))
6478 // CHECK11-NEXT:    ret void
6479 //
6480 //
6481 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..13
6482 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
6483 // CHECK11-NEXT:  entry:
6484 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6485 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6486 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6487 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6488 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6489 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6490 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6491 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6492 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
6493 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
6494 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6495 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6496 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6497 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
6498 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6499 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6500 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6501 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
6502 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6503 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6504 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
6505 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6506 // CHECK11:       cond.true:
6507 // CHECK11-NEXT:    br label [[COND_END:%.*]]
6508 // CHECK11:       cond.false:
6509 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6510 // CHECK11-NEXT:    br label [[COND_END]]
6511 // CHECK11:       cond.end:
6512 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
6513 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6514 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6515 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
6516 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6517 // CHECK11:       omp.inner.for.cond:
6518 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6519 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6520 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
6521 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6522 // CHECK11:       omp.inner.for.body:
6523 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6524 // CHECK11-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
6525 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6526 // CHECK11-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
6527 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
6528 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6529 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
6530 // CHECK11-NEXT:    call void @.omp_outlined..14(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
6531 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
6532 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6533 // CHECK11:       omp.inner.for.inc:
6534 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6535 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6536 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
6537 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6538 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP57:![0-9]+]]
6539 // CHECK11:       omp.inner.for.end:
6540 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6541 // CHECK11:       omp.loop.exit:
6542 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
6543 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6544 // CHECK11-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
6545 // CHECK11-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6546 // CHECK11:       .omp.final.then:
6547 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
6548 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6549 // CHECK11:       .omp.final.done:
6550 // CHECK11-NEXT:    ret void
6551 //
6552 //
6553 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..14
6554 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
6555 // CHECK11-NEXT:  entry:
6556 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6557 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6558 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6559 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6560 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6561 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6562 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6563 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6564 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6565 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6566 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
6567 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6568 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6569 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6570 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6571 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6572 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6573 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6574 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
6575 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6576 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
6577 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
6578 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
6579 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6580 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6581 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6582 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
6583 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6584 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6585 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
6586 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6587 // CHECK11:       cond.true:
6588 // CHECK11-NEXT:    br label [[COND_END:%.*]]
6589 // CHECK11:       cond.false:
6590 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6591 // CHECK11-NEXT:    br label [[COND_END]]
6592 // CHECK11:       cond.end:
6593 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6594 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6595 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6596 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6597 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6598 // CHECK11:       omp.inner.for.cond:
6599 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6600 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6601 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6602 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6603 // CHECK11:       omp.inner.for.body:
6604 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6605 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6606 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6607 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
6608 // CHECK11-NEXT:    call void @_Z3fn2v()
6609 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6610 // CHECK11:       omp.body.continue:
6611 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6612 // CHECK11:       omp.inner.for.inc:
6613 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6614 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
6615 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
6616 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP58:![0-9]+]]
6617 // CHECK11:       omp.inner.for.end:
6618 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6619 // CHECK11:       omp.loop.exit:
6620 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6621 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6622 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
6623 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6624 // CHECK11:       .omp.final.then:
6625 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
6626 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6627 // CHECK11:       .omp.final.done:
6628 // CHECK11-NEXT:    ret void
6629 //
6630 //
6631 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71
6632 // CHECK11-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
6633 // CHECK11-NEXT:  entry:
6634 // CHECK11-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
6635 // CHECK11-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
6636 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
6637 // 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*, i32*)* @.omp_outlined..15 to void (i32*, i32*, ...)*), i32* [[CONV]])
6638 // CHECK11-NEXT:    ret void
6639 //
6640 //
6641 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..15
6642 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
6643 // CHECK11-NEXT:  entry:
6644 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6645 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6646 // CHECK11-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
6647 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6648 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6649 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6650 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6651 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6652 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6653 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
6654 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
6655 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6656 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6657 // CHECK11-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
6658 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
6659 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6660 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
6661 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6662 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6663 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6664 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6665 // 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)
6666 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6667 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
6668 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6669 // CHECK11:       cond.true:
6670 // CHECK11-NEXT:    br label [[COND_END:%.*]]
6671 // CHECK11:       cond.false:
6672 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6673 // CHECK11-NEXT:    br label [[COND_END]]
6674 // CHECK11:       cond.end:
6675 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
6676 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6677 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6678 // CHECK11-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
6679 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6680 // CHECK11:       omp.inner.for.cond:
6681 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
6682 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
6683 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6684 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6685 // CHECK11:       omp.inner.for.body:
6686 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !59
6687 // CHECK11-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
6688 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
6689 // CHECK11-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
6690 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !59
6691 // CHECK11-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
6692 // CHECK11-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6693 // CHECK11:       omp_if.then:
6694 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..16 to void (i32*, i32*, ...)*), i64 [[TMP9]], i64 [[TMP11]]), !llvm.access.group !59
6695 // CHECK11-NEXT:    br label [[OMP_IF_END:%.*]]
6696 // CHECK11:       omp_if.else:
6697 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !59
6698 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !59
6699 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !59
6700 // CHECK11-NEXT:    call void @.omp_outlined..16(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !59
6701 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !59
6702 // CHECK11-NEXT:    br label [[OMP_IF_END]]
6703 // CHECK11:       omp_if.end:
6704 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6705 // CHECK11:       omp.inner.for.inc:
6706 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
6707 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !59
6708 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
6709 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
6710 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP60:![0-9]+]]
6711 // CHECK11:       omp.inner.for.end:
6712 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6713 // CHECK11:       omp.loop.exit:
6714 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
6715 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6716 // CHECK11-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
6717 // CHECK11-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6718 // CHECK11:       .omp.final.then:
6719 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
6720 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6721 // CHECK11:       .omp.final.done:
6722 // CHECK11-NEXT:    ret void
6723 //
6724 //
6725 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..16
6726 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
6727 // CHECK11-NEXT:  entry:
6728 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6729 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6730 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6731 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6732 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6733 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6734 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6735 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6736 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6737 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6738 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
6739 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6740 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6741 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6742 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6743 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6744 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6745 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6746 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
6747 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6748 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
6749 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
6750 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
6751 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6752 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6753 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6754 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
6755 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6756 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6757 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
6758 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6759 // CHECK11:       cond.true:
6760 // CHECK11-NEXT:    br label [[COND_END:%.*]]
6761 // CHECK11:       cond.false:
6762 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6763 // CHECK11-NEXT:    br label [[COND_END]]
6764 // CHECK11:       cond.end:
6765 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6766 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6767 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6768 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6769 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6770 // CHECK11:       omp.inner.for.cond:
6771 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
6772 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !62
6773 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6774 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6775 // CHECK11:       omp.inner.for.body:
6776 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
6777 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6778 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6779 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !62
6780 // CHECK11-NEXT:    call void @_Z3fn3v(), !llvm.access.group !62
6781 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6782 // CHECK11:       omp.body.continue:
6783 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6784 // CHECK11:       omp.inner.for.inc:
6785 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
6786 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
6787 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
6788 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP63:![0-9]+]]
6789 // CHECK11:       omp.inner.for.end:
6790 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6791 // CHECK11:       omp.loop.exit:
6792 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6793 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6794 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
6795 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6796 // CHECK11:       .omp.final.then:
6797 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
6798 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6799 // CHECK11:       .omp.final.done:
6800 // CHECK11-NEXT:    ret void
6801 //
6802 //
6803 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
6804 // CHECK11-SAME: () #[[ATTR5:[0-9]+]] {
6805 // CHECK11-NEXT:  entry:
6806 // CHECK11-NEXT:    call void @__tgt_register_requires(i64 1)
6807 // CHECK11-NEXT:    ret void
6808 //
6809 //
6810 // CHECK13-LABEL: define {{[^@]+}}@_Z9gtid_testv
6811 // CHECK13-SAME: () #[[ATTR0:[0-9]+]] {
6812 // CHECK13-NEXT:  entry:
6813 // CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6814 // CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6815 // CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6816 // CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6817 // CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
6818 // CHECK13-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
6819 // CHECK13-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
6820 // CHECK13-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
6821 // CHECK13-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
6822 // CHECK13-NEXT:    [[I6:%.*]] = alloca i32, align 4
6823 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6824 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6825 // CHECK13-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6826 // CHECK13-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
6827 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6828 // CHECK13:       omp.inner.for.cond:
6829 // CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
6830 // CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6
6831 // CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
6832 // CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6833 // CHECK13:       omp.inner.for.body:
6834 // CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
6835 // CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
6836 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6837 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !6
6838 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6839 // CHECK13:       omp.body.continue:
6840 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6841 // CHECK13:       omp.inner.for.inc:
6842 // CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
6843 // CHECK13-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
6844 // CHECK13-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
6845 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
6846 // CHECK13:       omp.inner.for.end:
6847 // CHECK13-NEXT:    store i32 100, i32* [[I]], align 4
6848 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
6849 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
6850 // CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
6851 // CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
6852 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
6853 // CHECK13:       omp.inner.for.cond7:
6854 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
6855 // CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !10
6856 // CHECK13-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6857 // CHECK13-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
6858 // CHECK13:       omp.inner.for.body9:
6859 // CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
6860 // CHECK13-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
6861 // CHECK13-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
6862 // CHECK13-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !10
6863 // CHECK13-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !10
6864 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
6865 // CHECK13:       omp.body.continue12:
6866 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
6867 // CHECK13:       omp.inner.for.inc13:
6868 // CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
6869 // CHECK13-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
6870 // CHECK13-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
6871 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP11:![0-9]+]]
6872 // CHECK13:       omp.inner.for.end15:
6873 // CHECK13-NEXT:    store i32 100, i32* [[I6]], align 4
6874 // CHECK13-NEXT:    ret void
6875 //
6876 //
6877 // CHECK13-LABEL: define {{[^@]+}}@main
6878 // CHECK13-SAME: () #[[ATTR1:[0-9]+]] {
6879 // CHECK13-NEXT:  entry:
6880 // CHECK13-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
6881 // CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6882 // CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6883 // CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6884 // CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6885 // CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
6886 // CHECK13-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
6887 // CHECK13-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
6888 // CHECK13-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
6889 // CHECK13-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
6890 // CHECK13-NEXT:    [[I6:%.*]] = alloca i32, align 4
6891 // CHECK13-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
6892 // CHECK13-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
6893 // CHECK13-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
6894 // CHECK13-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
6895 // CHECK13-NEXT:    [[I20:%.*]] = alloca i32, align 4
6896 // CHECK13-NEXT:    store i32 0, i32* [[RETVAL]], align 4
6897 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6898 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6899 // CHECK13-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6900 // CHECK13-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
6901 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6902 // CHECK13:       omp.inner.for.cond:
6903 // CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
6904 // CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
6905 // CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
6906 // CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6907 // CHECK13:       omp.inner.for.body:
6908 // CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
6909 // CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
6910 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6911 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
6912 // CHECK13-NEXT:    call void @_Z3fn4v(), !llvm.access.group !13
6913 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6914 // CHECK13:       omp.body.continue:
6915 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6916 // CHECK13:       omp.inner.for.inc:
6917 // CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
6918 // CHECK13-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
6919 // CHECK13-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
6920 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
6921 // CHECK13:       omp.inner.for.end:
6922 // CHECK13-NEXT:    store i32 100, i32* [[I]], align 4
6923 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
6924 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
6925 // CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
6926 // CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
6927 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
6928 // CHECK13:       omp.inner.for.cond7:
6929 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
6930 // CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !16
6931 // CHECK13-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6932 // CHECK13-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
6933 // CHECK13:       omp.inner.for.body9:
6934 // CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
6935 // CHECK13-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
6936 // CHECK13-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
6937 // CHECK13-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !16
6938 // CHECK13-NEXT:    call void @_Z3fn5v(), !llvm.access.group !16
6939 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
6940 // CHECK13:       omp.body.continue12:
6941 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
6942 // CHECK13:       omp.inner.for.inc13:
6943 // CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
6944 // CHECK13-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
6945 // CHECK13-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
6946 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP17:![0-9]+]]
6947 // CHECK13:       omp.inner.for.end15:
6948 // CHECK13-NEXT:    store i32 100, i32* [[I6]], align 4
6949 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
6950 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
6951 // CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
6952 // CHECK13-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
6953 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
6954 // CHECK13:       omp.inner.for.cond21:
6955 // CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !19
6956 // CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !19
6957 // CHECK13-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
6958 // CHECK13-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
6959 // CHECK13:       omp.inner.for.body23:
6960 // CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !19
6961 // CHECK13-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
6962 // CHECK13-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
6963 // CHECK13-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !19
6964 // CHECK13-NEXT:    call void @_Z3fn6v(), !llvm.access.group !19
6965 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
6966 // CHECK13:       omp.body.continue26:
6967 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
6968 // CHECK13:       omp.inner.for.inc27:
6969 // CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !19
6970 // CHECK13-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
6971 // CHECK13-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !19
6972 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP20:![0-9]+]]
6973 // CHECK13:       omp.inner.for.end29:
6974 // CHECK13-NEXT:    store i32 100, i32* [[I20]], align 4
6975 // CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
6976 // CHECK13-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
6977 // CHECK13-NEXT:    ret i32 [[CALL]]
6978 //
6979 //
6980 // CHECK13-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
6981 // CHECK13-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
6982 // CHECK13-NEXT:  entry:
6983 // CHECK13-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
6984 // CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6985 // CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6986 // CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6987 // CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6988 // CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
6989 // CHECK13-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
6990 // CHECK13-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
6991 // CHECK13-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
6992 // CHECK13-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
6993 // CHECK13-NEXT:    [[I6:%.*]] = alloca i32, align 4
6994 // CHECK13-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
6995 // CHECK13-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
6996 // CHECK13-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
6997 // CHECK13-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
6998 // CHECK13-NEXT:    [[I20:%.*]] = alloca i32, align 4
6999 // CHECK13-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
7000 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7001 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
7002 // CHECK13-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7003 // CHECK13-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
7004 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7005 // CHECK13:       omp.inner.for.cond:
7006 // CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
7007 // CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
7008 // CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
7009 // CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7010 // CHECK13:       omp.inner.for.body:
7011 // CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
7012 // CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
7013 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7014 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
7015 // CHECK13-NEXT:    call void @_Z3fn1v(), !llvm.access.group !22
7016 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7017 // CHECK13:       omp.body.continue:
7018 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7019 // CHECK13:       omp.inner.for.inc:
7020 // CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
7021 // CHECK13-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
7022 // CHECK13-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
7023 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
7024 // CHECK13:       omp.inner.for.end:
7025 // CHECK13-NEXT:    store i32 100, i32* [[I]], align 4
7026 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
7027 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
7028 // CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
7029 // CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
7030 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
7031 // CHECK13:       omp.inner.for.cond7:
7032 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !25
7033 // CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !25
7034 // CHECK13-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7035 // CHECK13-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
7036 // CHECK13:       omp.inner.for.body9:
7037 // CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !25
7038 // CHECK13-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
7039 // CHECK13-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
7040 // CHECK13-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !25
7041 // CHECK13-NEXT:    call void @_Z3fn2v(), !llvm.access.group !25
7042 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
7043 // CHECK13:       omp.body.continue12:
7044 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
7045 // CHECK13:       omp.inner.for.inc13:
7046 // CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !25
7047 // CHECK13-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
7048 // CHECK13-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !25
7049 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP26:![0-9]+]]
7050 // CHECK13:       omp.inner.for.end15:
7051 // CHECK13-NEXT:    store i32 100, i32* [[I6]], align 4
7052 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
7053 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
7054 // CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
7055 // CHECK13-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
7056 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
7057 // CHECK13:       omp.inner.for.cond21:
7058 // CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !28
7059 // CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !28
7060 // CHECK13-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
7061 // CHECK13-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
7062 // CHECK13:       omp.inner.for.body23:
7063 // CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !28
7064 // CHECK13-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
7065 // CHECK13-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
7066 // CHECK13-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !28
7067 // CHECK13-NEXT:    call void @_Z3fn3v(), !llvm.access.group !28
7068 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
7069 // CHECK13:       omp.body.continue26:
7070 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
7071 // CHECK13:       omp.inner.for.inc27:
7072 // CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !28
7073 // CHECK13-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
7074 // CHECK13-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !28
7075 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP29:![0-9]+]]
7076 // CHECK13:       omp.inner.for.end29:
7077 // CHECK13-NEXT:    store i32 100, i32* [[I20]], align 4
7078 // CHECK13-NEXT:    ret i32 0
7079 //
7080 //
7081 // CHECK15-LABEL: define {{[^@]+}}@_Z9gtid_testv
7082 // CHECK15-SAME: () #[[ATTR0:[0-9]+]] {
7083 // CHECK15-NEXT:  entry:
7084 // CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7085 // CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7086 // CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7087 // CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7088 // CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
7089 // CHECK15-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
7090 // CHECK15-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
7091 // CHECK15-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
7092 // CHECK15-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
7093 // CHECK15-NEXT:    [[I6:%.*]] = alloca i32, align 4
7094 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7095 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
7096 // CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7097 // CHECK15-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
7098 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7099 // CHECK15:       omp.inner.for.cond:
7100 // CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
7101 // CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6
7102 // CHECK15-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
7103 // CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7104 // CHECK15:       omp.inner.for.body:
7105 // CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
7106 // CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
7107 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7108 // CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !6
7109 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7110 // CHECK15:       omp.body.continue:
7111 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7112 // CHECK15:       omp.inner.for.inc:
7113 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
7114 // CHECK15-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
7115 // CHECK15-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
7116 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
7117 // CHECK15:       omp.inner.for.end:
7118 // CHECK15-NEXT:    store i32 100, i32* [[I]], align 4
7119 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
7120 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
7121 // CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
7122 // CHECK15-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
7123 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
7124 // CHECK15:       omp.inner.for.cond7:
7125 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
7126 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !10
7127 // CHECK15-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7128 // CHECK15-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
7129 // CHECK15:       omp.inner.for.body9:
7130 // CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
7131 // CHECK15-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
7132 // CHECK15-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
7133 // CHECK15-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !10
7134 // CHECK15-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !10
7135 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
7136 // CHECK15:       omp.body.continue12:
7137 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
7138 // CHECK15:       omp.inner.for.inc13:
7139 // CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
7140 // CHECK15-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
7141 // CHECK15-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
7142 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP11:![0-9]+]]
7143 // CHECK15:       omp.inner.for.end15:
7144 // CHECK15-NEXT:    store i32 100, i32* [[I6]], align 4
7145 // CHECK15-NEXT:    ret void
7146 //
7147 //
7148 // CHECK15-LABEL: define {{[^@]+}}@main
7149 // CHECK15-SAME: () #[[ATTR1:[0-9]+]] {
7150 // CHECK15-NEXT:  entry:
7151 // CHECK15-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
7152 // CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7153 // CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7154 // CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7155 // CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7156 // CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
7157 // CHECK15-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
7158 // CHECK15-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
7159 // CHECK15-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
7160 // CHECK15-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
7161 // CHECK15-NEXT:    [[I6:%.*]] = alloca i32, align 4
7162 // CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
7163 // CHECK15-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
7164 // CHECK15-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
7165 // CHECK15-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
7166 // CHECK15-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
7167 // CHECK15-NEXT:    [[I20:%.*]] = alloca i32, align 4
7168 // CHECK15-NEXT:    store i32 0, i32* [[RETVAL]], align 4
7169 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7170 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
7171 // CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7172 // CHECK15-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
7173 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7174 // CHECK15:       omp.inner.for.cond:
7175 // CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
7176 // CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
7177 // CHECK15-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
7178 // CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7179 // CHECK15:       omp.inner.for.body:
7180 // CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
7181 // CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
7182 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7183 // CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
7184 // CHECK15-NEXT:    call void @_Z3fn4v(), !llvm.access.group !13
7185 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7186 // CHECK15:       omp.body.continue:
7187 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7188 // CHECK15:       omp.inner.for.inc:
7189 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
7190 // CHECK15-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
7191 // CHECK15-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
7192 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
7193 // CHECK15:       omp.inner.for.end:
7194 // CHECK15-NEXT:    store i32 100, i32* [[I]], align 4
7195 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
7196 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
7197 // CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
7198 // CHECK15-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
7199 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
7200 // CHECK15:       omp.inner.for.cond7:
7201 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7202 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4
7203 // CHECK15-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7204 // CHECK15-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
7205 // CHECK15:       omp.inner.for.body9:
7206 // CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7207 // CHECK15-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
7208 // CHECK15-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
7209 // CHECK15-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4
7210 // CHECK15-NEXT:    call void @_Z3fn5v()
7211 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
7212 // CHECK15:       omp.body.continue12:
7213 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
7214 // CHECK15:       omp.inner.for.inc13:
7215 // CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7216 // CHECK15-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
7217 // CHECK15-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4
7218 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP16:![0-9]+]]
7219 // CHECK15:       omp.inner.for.end15:
7220 // CHECK15-NEXT:    store i32 100, i32* [[I6]], align 4
7221 // CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* @Arg, align 4
7222 // CHECK15-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP10]], 0
7223 // CHECK15-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
7224 // CHECK15-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
7225 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
7226 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
7227 // CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
7228 // CHECK15-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV19]], align 4
7229 // CHECK15-NEXT:    [[TMP12:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
7230 // CHECK15-NEXT:    [[TOBOOL21:%.*]] = trunc i8 [[TMP12]] to i1
7231 // CHECK15-NEXT:    br i1 [[TOBOOL21]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
7232 // CHECK15:       omp_if.then:
7233 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND22:%.*]]
7234 // CHECK15:       omp.inner.for.cond22:
7235 // CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !18
7236 // CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !18
7237 // CHECK15-NEXT:    [[CMP23:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
7238 // CHECK15-NEXT:    br i1 [[CMP23]], label [[OMP_INNER_FOR_BODY24:%.*]], label [[OMP_INNER_FOR_END30:%.*]]
7239 // CHECK15:       omp.inner.for.body24:
7240 // CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !18
7241 // CHECK15-NEXT:    [[MUL25:%.*]] = mul nsw i32 [[TMP15]], 1
7242 // CHECK15-NEXT:    [[ADD26:%.*]] = add nsw i32 0, [[MUL25]]
7243 // CHECK15-NEXT:    store i32 [[ADD26]], i32* [[I20]], align 4, !llvm.access.group !18
7244 // CHECK15-NEXT:    call void @_Z3fn6v(), !llvm.access.group !18
7245 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE27:%.*]]
7246 // CHECK15:       omp.body.continue27:
7247 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC28:%.*]]
7248 // CHECK15:       omp.inner.for.inc28:
7249 // CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !18
7250 // CHECK15-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP16]], 1
7251 // CHECK15-NEXT:    store i32 [[ADD29]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !18
7252 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND22]], !llvm.loop [[LOOP19:![0-9]+]]
7253 // CHECK15:       omp.inner.for.end30:
7254 // CHECK15-NEXT:    br label [[OMP_IF_END:%.*]]
7255 // CHECK15:       omp_if.else:
7256 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND31:%.*]]
7257 // CHECK15:       omp.inner.for.cond31:
7258 // CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
7259 // CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4
7260 // CHECK15-NEXT:    [[CMP32:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
7261 // CHECK15-NEXT:    br i1 [[CMP32]], label [[OMP_INNER_FOR_BODY33:%.*]], label [[OMP_INNER_FOR_END39:%.*]]
7262 // CHECK15:       omp.inner.for.body33:
7263 // CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
7264 // CHECK15-NEXT:    [[MUL34:%.*]] = mul nsw i32 [[TMP19]], 1
7265 // CHECK15-NEXT:    [[ADD35:%.*]] = add nsw i32 0, [[MUL34]]
7266 // CHECK15-NEXT:    store i32 [[ADD35]], i32* [[I20]], align 4
7267 // CHECK15-NEXT:    call void @_Z3fn6v()
7268 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE36:%.*]]
7269 // CHECK15:       omp.body.continue36:
7270 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC37:%.*]]
7271 // CHECK15:       omp.inner.for.inc37:
7272 // CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
7273 // CHECK15-NEXT:    [[ADD38:%.*]] = add nsw i32 [[TMP20]], 1
7274 // CHECK15-NEXT:    store i32 [[ADD38]], i32* [[DOTOMP_IV19]], align 4
7275 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND31]], !llvm.loop [[LOOP21:![0-9]+]]
7276 // CHECK15:       omp.inner.for.end39:
7277 // CHECK15-NEXT:    br label [[OMP_IF_END]]
7278 // CHECK15:       omp_if.end:
7279 // CHECK15-NEXT:    store i32 100, i32* [[I20]], align 4
7280 // CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* @Arg, align 4
7281 // CHECK15-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP21]])
7282 // CHECK15-NEXT:    ret i32 [[CALL]]
7283 //
7284 //
7285 // CHECK15-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
7286 // CHECK15-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
7287 // CHECK15-NEXT:  entry:
7288 // CHECK15-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
7289 // CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7290 // CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7291 // CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7292 // CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7293 // CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
7294 // CHECK15-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
7295 // CHECK15-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
7296 // CHECK15-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
7297 // CHECK15-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
7298 // CHECK15-NEXT:    [[I6:%.*]] = alloca i32, align 4
7299 // CHECK15-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
7300 // CHECK15-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
7301 // CHECK15-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
7302 // CHECK15-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
7303 // CHECK15-NEXT:    [[I20:%.*]] = alloca i32, align 4
7304 // CHECK15-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
7305 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7306 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
7307 // CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7308 // CHECK15-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
7309 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7310 // CHECK15:       omp.inner.for.cond:
7311 // CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
7312 // CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
7313 // CHECK15-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
7314 // CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7315 // CHECK15:       omp.inner.for.body:
7316 // CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
7317 // CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
7318 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7319 // CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
7320 // CHECK15-NEXT:    call void @_Z3fn1v(), !llvm.access.group !22
7321 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7322 // CHECK15:       omp.body.continue:
7323 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7324 // CHECK15:       omp.inner.for.inc:
7325 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
7326 // CHECK15-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
7327 // CHECK15-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
7328 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
7329 // CHECK15:       omp.inner.for.end:
7330 // CHECK15-NEXT:    store i32 100, i32* [[I]], align 4
7331 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
7332 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
7333 // CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
7334 // CHECK15-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
7335 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
7336 // CHECK15:       omp.inner.for.cond7:
7337 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7338 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4
7339 // CHECK15-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7340 // CHECK15-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
7341 // CHECK15:       omp.inner.for.body9:
7342 // CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7343 // CHECK15-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
7344 // CHECK15-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
7345 // CHECK15-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4
7346 // CHECK15-NEXT:    call void @_Z3fn2v()
7347 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
7348 // CHECK15:       omp.body.continue12:
7349 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
7350 // CHECK15:       omp.inner.for.inc13:
7351 // CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7352 // CHECK15-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
7353 // CHECK15-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4
7354 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP25:![0-9]+]]
7355 // CHECK15:       omp.inner.for.end15:
7356 // CHECK15-NEXT:    store i32 100, i32* [[I6]], align 4
7357 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
7358 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
7359 // CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
7360 // CHECK15-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
7361 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
7362 // CHECK15:       omp.inner.for.cond21:
7363 // CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !26
7364 // CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !26
7365 // CHECK15-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
7366 // CHECK15-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
7367 // CHECK15:       omp.inner.for.body23:
7368 // CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !26
7369 // CHECK15-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
7370 // CHECK15-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
7371 // CHECK15-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !26
7372 // CHECK15-NEXT:    call void @_Z3fn3v(), !llvm.access.group !26
7373 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
7374 // CHECK15:       omp.body.continue26:
7375 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
7376 // CHECK15:       omp.inner.for.inc27:
7377 // CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !26
7378 // CHECK15-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
7379 // CHECK15-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !26
7380 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP27:![0-9]+]]
7381 // CHECK15:       omp.inner.for.end29:
7382 // CHECK15-NEXT:    store i32 100, i32* [[I20]], align 4
7383 // CHECK15-NEXT:    ret i32 0
7384 //
7385