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=CHECK2
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=CHECK4
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=CHECK6
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=CHECK8
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=CHECK10
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=CHECK12
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=CHECK14
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=CHECK16
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 // CHECK2-LABEL: define {{[^@]+}}@_Z9gtid_testv
1530 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
1531 // CHECK2-NEXT:  entry:
1532 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1533 // CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1534 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
1535 // CHECK2-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 // CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
1537 // CHECK2-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1538 // CHECK2:       omp_offload.failed:
1539 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43() #[[ATTR2:[0-9]+]]
1540 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1541 // CHECK2:       omp_offload.cont:
1542 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1543 // CHECK2-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 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
1545 // CHECK2-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
1546 // CHECK2:       omp_offload.failed2:
1547 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48() #[[ATTR2]]
1548 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
1549 // CHECK2:       omp_offload.cont3:
1550 // CHECK2-NEXT:    ret void
1551 //
1552 //
1553 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43
1554 // CHECK2-SAME: () #[[ATTR1:[0-9]+]] {
1555 // CHECK2-NEXT:  entry:
1556 // CHECK2-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 // CHECK2-NEXT:    ret void
1558 //
1559 //
1560 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
1561 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1562 // CHECK2-NEXT:  entry:
1563 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1564 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1565 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1566 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1567 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1568 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1569 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1570 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1571 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
1572 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1573 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1574 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1575 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1576 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1577 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1578 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1579 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1580 // CHECK2-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 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1582 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1583 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1584 // CHECK2:       cond.true:
1585 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1586 // CHECK2:       cond.false:
1587 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1588 // CHECK2-NEXT:    br label [[COND_END]]
1589 // CHECK2:       cond.end:
1590 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1591 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1592 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1593 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1594 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1595 // CHECK2:       omp.inner.for.cond:
1596 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1597 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !11
1598 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1599 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1600 // CHECK2:       omp.inner.for.body:
1601 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !11
1602 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1603 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !11
1604 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1605 // CHECK2-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 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1607 // CHECK2:       omp.inner.for.inc:
1608 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1609 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !11
1610 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1611 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
1612 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
1613 // CHECK2:       omp.inner.for.end:
1614 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1615 // CHECK2:       omp.loop.exit:
1616 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1617 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1618 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1619 // CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1620 // CHECK2:       .omp.final.then:
1621 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
1622 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1623 // CHECK2:       .omp.final.done:
1624 // CHECK2-NEXT:    ret void
1625 //
1626 //
1627 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
1628 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1629 // CHECK2-NEXT:  entry:
1630 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1631 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1632 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1633 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1634 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1635 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1636 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1637 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1638 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1639 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1640 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
1641 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1642 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1643 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1644 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1645 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1646 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
1647 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1648 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1649 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1650 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1651 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1652 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1653 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1654 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1655 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1656 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1657 // CHECK2-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 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1659 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1660 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1661 // CHECK2:       cond.true:
1662 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1663 // CHECK2:       cond.false:
1664 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1665 // CHECK2-NEXT:    br label [[COND_END]]
1666 // CHECK2:       cond.end:
1667 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1668 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1669 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1670 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1671 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1672 // CHECK2:       omp.inner.for.cond:
1673 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
1674 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !15
1675 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1676 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1677 // CHECK2:       omp.inner.for.body:
1678 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
1679 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1680 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1681 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !15
1682 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1683 // CHECK2:       omp.body.continue:
1684 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1685 // CHECK2:       omp.inner.for.inc:
1686 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
1687 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1688 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
1689 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
1690 // CHECK2:       omp.inner.for.end:
1691 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1692 // CHECK2:       omp.loop.exit:
1693 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1694 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1695 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1696 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1697 // CHECK2:       .omp.final.then:
1698 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
1699 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1700 // CHECK2:       .omp.final.done:
1701 // CHECK2-NEXT:    ret void
1702 //
1703 //
1704 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48
1705 // CHECK2-SAME: () #[[ATTR1]] {
1706 // CHECK2-NEXT:  entry:
1707 // CHECK2-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 // CHECK2-NEXT:    ret void
1709 //
1710 //
1711 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
1712 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1713 // CHECK2-NEXT:  entry:
1714 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1715 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1716 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1717 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1718 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1719 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1720 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1721 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1722 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
1723 // CHECK2-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1724 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1725 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1726 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1727 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1728 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1729 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1730 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1731 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1732 // CHECK2-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 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1734 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1735 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1736 // CHECK2:       cond.true:
1737 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1738 // CHECK2:       cond.false:
1739 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1740 // CHECK2-NEXT:    br label [[COND_END]]
1741 // CHECK2:       cond.end:
1742 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1743 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1744 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1745 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1746 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1747 // CHECK2:       omp.inner.for.cond:
1748 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
1749 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !20
1750 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1751 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1752 // CHECK2:       omp.inner.for.body:
1753 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !20
1754 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1755 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !20
1756 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1757 // CHECK2-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !20
1758 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !20
1759 // CHECK2-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !20
1760 // CHECK2-NEXT:    call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !20
1761 // CHECK2-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !20
1762 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1763 // CHECK2:       omp.inner.for.inc:
1764 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
1765 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !20
1766 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1767 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
1768 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
1769 // CHECK2:       omp.inner.for.end:
1770 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1771 // CHECK2:       omp.loop.exit:
1772 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1773 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1774 // CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
1775 // CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1776 // CHECK2:       .omp.final.then:
1777 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
1778 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1779 // CHECK2:       .omp.final.done:
1780 // CHECK2-NEXT:    ret void
1781 //
1782 //
1783 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
1784 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1785 // CHECK2-NEXT:  entry:
1786 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1787 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1788 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1789 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1790 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1791 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1792 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1793 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1794 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1795 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1796 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
1797 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1798 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1799 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1800 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1801 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1802 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
1803 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1804 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1805 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1806 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1807 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1808 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1809 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1810 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1811 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1812 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1813 // CHECK2-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 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1815 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1816 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1817 // CHECK2:       cond.true:
1818 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1819 // CHECK2:       cond.false:
1820 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1821 // CHECK2-NEXT:    br label [[COND_END]]
1822 // CHECK2:       cond.end:
1823 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1824 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1825 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1826 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1827 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1828 // CHECK2:       omp.inner.for.cond:
1829 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
1830 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !23
1831 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1832 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1833 // CHECK2:       omp.inner.for.body:
1834 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
1835 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1836 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1837 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !23
1838 // CHECK2-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !23
1839 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1840 // CHECK2:       omp.body.continue:
1841 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1842 // CHECK2:       omp.inner.for.inc:
1843 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
1844 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1845 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
1846 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
1847 // CHECK2:       omp.inner.for.end:
1848 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1849 // CHECK2:       omp.loop.exit:
1850 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1851 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1852 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
1853 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1854 // CHECK2:       .omp.final.then:
1855 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
1856 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1857 // CHECK2:       .omp.final.done:
1858 // CHECK2-NEXT:    ret void
1859 //
1860 //
1861 // CHECK2-LABEL: define {{[^@]+}}@main
1862 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] {
1863 // CHECK2-NEXT:  entry:
1864 // CHECK2-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1865 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1866 // CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1867 // CHECK2-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
1868 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
1869 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
1870 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
1871 // CHECK2-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
1872 // CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
1873 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1874 // CHECK2-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 // CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
1876 // CHECK2-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1877 // CHECK2:       omp_offload.failed:
1878 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81() #[[ATTR2]]
1879 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1880 // CHECK2:       omp_offload.cont:
1881 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1882 // CHECK2-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 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
1884 // CHECK2-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
1885 // CHECK2:       omp_offload.failed2:
1886 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90() #[[ATTR2]]
1887 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
1888 // CHECK2:       omp_offload.cont3:
1889 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* @Arg, align 4
1890 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
1891 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
1892 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
1893 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1894 // CHECK2-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
1895 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
1896 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1897 // CHECK2-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
1898 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
1899 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1900 // CHECK2-NEXT:    store i8* null, i8** [[TMP10]], align 8
1901 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1902 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1903 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
1904 // CHECK2-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 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1906 // CHECK2-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
1907 // CHECK2:       omp_offload.failed5:
1908 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99(i64 [[TMP5]]) #[[ATTR2]]
1909 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
1910 // CHECK2:       omp_offload.cont6:
1911 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
1912 // CHECK2-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
1913 // CHECK2-NEXT:    ret i32 [[CALL]]
1914 //
1915 //
1916 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
1917 // CHECK2-SAME: () #[[ATTR1]] {
1918 // CHECK2-NEXT:  entry:
1919 // CHECK2-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 // CHECK2-NEXT:    ret void
1921 //
1922 //
1923 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
1924 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1925 // CHECK2-NEXT:  entry:
1926 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1927 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1928 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1929 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1930 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1931 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1932 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1933 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1934 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
1935 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1936 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1937 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1938 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1939 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1940 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1941 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1942 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1943 // CHECK2-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 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1945 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1946 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1947 // CHECK2:       cond.true:
1948 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1949 // CHECK2:       cond.false:
1950 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1951 // CHECK2-NEXT:    br label [[COND_END]]
1952 // CHECK2:       cond.end:
1953 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1954 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1955 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1956 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1957 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1958 // CHECK2:       omp.inner.for.cond:
1959 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
1960 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !26
1961 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1962 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1963 // CHECK2:       omp.inner.for.body:
1964 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !26
1965 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1966 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !26
1967 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1968 // CHECK2-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 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1970 // CHECK2:       omp.inner.for.inc:
1971 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
1972 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !26
1973 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1974 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
1975 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
1976 // CHECK2:       omp.inner.for.end:
1977 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1978 // CHECK2:       omp.loop.exit:
1979 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1980 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
1981 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
1982 // CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
1983 // CHECK2:       .omp.final.then:
1984 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
1985 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
1986 // CHECK2:       .omp.final.done:
1987 // CHECK2-NEXT:    ret void
1988 //
1989 //
1990 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..5
1991 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1992 // CHECK2-NEXT:  entry:
1993 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1994 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1995 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1996 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1997 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1998 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1999 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2000 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2001 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2002 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2003 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2004 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2005 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2006 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2007 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2008 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2009 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
2010 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2011 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2012 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2013 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2014 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2015 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2016 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2017 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2018 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2019 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2020 // CHECK2-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 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2022 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2023 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2024 // CHECK2:       cond.true:
2025 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2026 // CHECK2:       cond.false:
2027 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2028 // CHECK2-NEXT:    br label [[COND_END]]
2029 // CHECK2:       cond.end:
2030 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2031 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2032 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2033 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2034 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2035 // CHECK2:       omp.inner.for.cond:
2036 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
2037 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !29
2038 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2039 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2040 // CHECK2:       omp.inner.for.body:
2041 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
2042 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2043 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2044 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !29
2045 // CHECK2-NEXT:    call void @_Z3fn4v(), !llvm.access.group !29
2046 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2047 // CHECK2:       omp.body.continue:
2048 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2049 // CHECK2:       omp.inner.for.inc:
2050 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
2051 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2052 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
2053 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
2054 // CHECK2:       omp.inner.for.end:
2055 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2056 // CHECK2:       omp.loop.exit:
2057 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2058 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2059 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2060 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2061 // CHECK2:       .omp.final.then:
2062 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
2063 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2064 // CHECK2:       .omp.final.done:
2065 // CHECK2-NEXT:    ret void
2066 //
2067 //
2068 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90
2069 // CHECK2-SAME: () #[[ATTR1]] {
2070 // CHECK2-NEXT:  entry:
2071 // CHECK2-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 // CHECK2-NEXT:    ret void
2073 //
2074 //
2075 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..6
2076 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2077 // CHECK2-NEXT:  entry:
2078 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2079 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2080 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2081 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2082 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2083 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2084 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2085 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2086 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2087 // CHECK2-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
2088 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2089 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2090 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2091 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2092 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2093 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2094 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2095 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2096 // CHECK2-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 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2098 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2099 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2100 // CHECK2:       cond.true:
2101 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2102 // CHECK2:       cond.false:
2103 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2104 // CHECK2-NEXT:    br label [[COND_END]]
2105 // CHECK2:       cond.end:
2106 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2107 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2108 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2109 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2110 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2111 // CHECK2:       omp.inner.for.cond:
2112 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
2113 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !32
2114 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2115 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2116 // CHECK2:       omp.inner.for.body:
2117 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !32
2118 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2119 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !32
2120 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2121 // CHECK2-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !32
2122 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !32
2123 // CHECK2-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !32
2124 // CHECK2-NEXT:    call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !32
2125 // CHECK2-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !32
2126 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2127 // CHECK2:       omp.inner.for.inc:
2128 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
2129 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !32
2130 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2131 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !32
2132 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP33:![0-9]+]]
2133 // CHECK2:       omp.inner.for.end:
2134 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2135 // CHECK2:       omp.loop.exit:
2136 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2137 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2138 // CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
2139 // CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2140 // CHECK2:       .omp.final.then:
2141 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
2142 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2143 // CHECK2:       .omp.final.done:
2144 // CHECK2-NEXT:    ret void
2145 //
2146 //
2147 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..7
2148 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2149 // CHECK2-NEXT:  entry:
2150 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2151 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2152 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2153 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2154 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2155 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2156 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2157 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2158 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2159 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2160 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2161 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2162 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2163 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2164 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2165 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2166 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
2167 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2168 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2169 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2170 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2171 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2172 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2173 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2174 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2175 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2176 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2177 // CHECK2-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 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2179 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2180 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2181 // CHECK2:       cond.true:
2182 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2183 // CHECK2:       cond.false:
2184 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2185 // CHECK2-NEXT:    br label [[COND_END]]
2186 // CHECK2:       cond.end:
2187 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2188 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2189 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2190 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2191 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2192 // CHECK2:       omp.inner.for.cond:
2193 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
2194 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !35
2195 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2196 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2197 // CHECK2:       omp.inner.for.body:
2198 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
2199 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2200 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2201 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !35
2202 // CHECK2-NEXT:    call void @_Z3fn5v(), !llvm.access.group !35
2203 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2204 // CHECK2:       omp.body.continue:
2205 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2206 // CHECK2:       omp.inner.for.inc:
2207 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
2208 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2209 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
2210 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
2211 // CHECK2:       omp.inner.for.end:
2212 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2213 // CHECK2:       omp.loop.exit:
2214 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2215 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2216 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2217 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2218 // CHECK2:       .omp.final.then:
2219 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
2220 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2221 // CHECK2:       .omp.final.done:
2222 // CHECK2-NEXT:    ret void
2223 //
2224 //
2225 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99
2226 // CHECK2-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
2227 // CHECK2-NEXT:  entry:
2228 // CHECK2-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
2229 // CHECK2-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
2230 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
2231 // CHECK2-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 // CHECK2-NEXT:    ret void
2233 //
2234 //
2235 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..8
2236 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
2237 // CHECK2-NEXT:  entry:
2238 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2239 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2240 // CHECK2-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
2241 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2242 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2243 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2244 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2245 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2246 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2247 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2248 // CHECK2-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
2249 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2250 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2251 // CHECK2-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
2252 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
2253 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2254 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2255 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2256 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2257 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2258 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2259 // CHECK2-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)
2260 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2261 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
2262 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2263 // CHECK2:       cond.true:
2264 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2265 // CHECK2:       cond.false:
2266 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2267 // CHECK2-NEXT:    br label [[COND_END]]
2268 // CHECK2:       cond.end:
2269 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2270 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2271 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2272 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2273 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2274 // CHECK2:       omp.inner.for.cond:
2275 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
2276 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !38
2277 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2278 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2279 // CHECK2:       omp.inner.for.body:
2280 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !38
2281 // CHECK2-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
2282 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !38
2283 // CHECK2-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
2284 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !38
2285 // CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
2286 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2287 // CHECK2:       omp_if.then:
2288 // CHECK2-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
2289 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
2290 // CHECK2:       omp_if.else:
2291 // CHECK2-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !38
2292 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !38
2293 // CHECK2-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !38
2294 // CHECK2-NEXT:    call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !38
2295 // CHECK2-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !38
2296 // CHECK2-NEXT:    br label [[OMP_IF_END]]
2297 // CHECK2:       omp_if.end:
2298 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2299 // CHECK2:       omp.inner.for.inc:
2300 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
2301 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !38
2302 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
2303 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !38
2304 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP39:![0-9]+]]
2305 // CHECK2:       omp.inner.for.end:
2306 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2307 // CHECK2:       omp.loop.exit:
2308 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2309 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2310 // CHECK2-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
2311 // CHECK2-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2312 // CHECK2:       .omp.final.then:
2313 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
2314 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2315 // CHECK2:       .omp.final.done:
2316 // CHECK2-NEXT:    ret void
2317 //
2318 //
2319 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..9
2320 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2321 // CHECK2-NEXT:  entry:
2322 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2323 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2324 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2325 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2326 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2327 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2328 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2329 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2330 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2331 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2332 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2333 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2334 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2335 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2336 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2337 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2338 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
2339 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2340 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2341 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2342 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2343 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2344 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2345 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2346 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2347 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2348 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2349 // CHECK2-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)
2350 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2351 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2352 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2353 // CHECK2:       cond.true:
2354 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2355 // CHECK2:       cond.false:
2356 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2357 // CHECK2-NEXT:    br label [[COND_END]]
2358 // CHECK2:       cond.end:
2359 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2360 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2361 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2362 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2363 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2364 // CHECK2:       omp.inner.for.cond:
2365 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
2366 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !41
2367 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2368 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2369 // CHECK2:       omp.inner.for.body:
2370 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
2371 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2372 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2373 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !41
2374 // CHECK2-NEXT:    call void @_Z3fn6v(), !llvm.access.group !41
2375 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2376 // CHECK2:       omp.body.continue:
2377 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2378 // CHECK2:       omp.inner.for.inc:
2379 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
2380 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2381 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !41
2382 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP42:![0-9]+]]
2383 // CHECK2:       omp.inner.for.end:
2384 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2385 // CHECK2:       omp.loop.exit:
2386 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2387 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2388 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2389 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2390 // CHECK2:       .omp.final.then:
2391 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
2392 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2393 // CHECK2:       .omp.final.done:
2394 // CHECK2-NEXT:    ret void
2395 //
2396 //
2397 // CHECK2-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
2398 // CHECK2-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
2399 // CHECK2-NEXT:  entry:
2400 // CHECK2-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
2401 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2402 // CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
2403 // CHECK2-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
2404 // CHECK2-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
2405 // CHECK2-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
2406 // CHECK2-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
2407 // CHECK2-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
2408 // CHECK2-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
2409 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
2410 // CHECK2-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)
2411 // CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
2412 // CHECK2-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2413 // CHECK2:       omp_offload.failed:
2414 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59() #[[ATTR2]]
2415 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2416 // CHECK2:       omp_offload.cont:
2417 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
2418 // CHECK2-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)
2419 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
2420 // CHECK2-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
2421 // CHECK2:       omp_offload.failed2:
2422 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65() #[[ATTR2]]
2423 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
2424 // CHECK2:       omp_offload.cont3:
2425 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
2426 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
2427 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
2428 // CHECK2-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
2429 // CHECK2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2430 // CHECK2-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
2431 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
2432 // CHECK2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2433 // CHECK2-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
2434 // CHECK2-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
2435 // CHECK2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
2436 // CHECK2-NEXT:    store i8* null, i8** [[TMP10]], align 8
2437 // CHECK2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2438 // CHECK2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2439 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
2440 // CHECK2-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)
2441 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
2442 // CHECK2-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
2443 // CHECK2:       omp_offload.failed5:
2444 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71(i64 [[TMP5]]) #[[ATTR2]]
2445 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
2446 // CHECK2:       omp_offload.cont6:
2447 // CHECK2-NEXT:    ret i32 0
2448 //
2449 //
2450 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59
2451 // CHECK2-SAME: () #[[ATTR1]] {
2452 // CHECK2-NEXT:  entry:
2453 // CHECK2-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*, ...)*))
2454 // CHECK2-NEXT:    ret void
2455 //
2456 //
2457 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..10
2458 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2459 // CHECK2-NEXT:  entry:
2460 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2461 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2462 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2463 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2464 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2465 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2466 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2467 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2468 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2469 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2470 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2471 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2472 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2473 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2474 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2475 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2476 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2477 // CHECK2-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)
2478 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2479 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2480 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2481 // CHECK2:       cond.true:
2482 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2483 // CHECK2:       cond.false:
2484 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2485 // CHECK2-NEXT:    br label [[COND_END]]
2486 // CHECK2:       cond.end:
2487 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2488 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2489 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2490 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2491 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2492 // CHECK2:       omp.inner.for.cond:
2493 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
2494 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !44
2495 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2496 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2497 // CHECK2:       omp.inner.for.body:
2498 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !44
2499 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2500 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !44
2501 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2502 // CHECK2-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
2503 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2504 // CHECK2:       omp.inner.for.inc:
2505 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
2506 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !44
2507 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
2508 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !44
2509 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP45:![0-9]+]]
2510 // CHECK2:       omp.inner.for.end:
2511 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2512 // CHECK2:       omp.loop.exit:
2513 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2514 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2515 // CHECK2-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
2516 // CHECK2-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2517 // CHECK2:       .omp.final.then:
2518 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
2519 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2520 // CHECK2:       .omp.final.done:
2521 // CHECK2-NEXT:    ret void
2522 //
2523 //
2524 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..11
2525 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2526 // CHECK2-NEXT:  entry:
2527 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2528 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2529 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2530 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2531 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2532 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2533 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2534 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2535 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2536 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2537 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2538 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2539 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2540 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2541 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2542 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2543 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
2544 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2545 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2546 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2547 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2548 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2549 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2550 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2551 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2552 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2553 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2554 // CHECK2-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)
2555 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2556 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2557 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2558 // CHECK2:       cond.true:
2559 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2560 // CHECK2:       cond.false:
2561 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2562 // CHECK2-NEXT:    br label [[COND_END]]
2563 // CHECK2:       cond.end:
2564 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2565 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2566 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2567 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2568 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2569 // CHECK2:       omp.inner.for.cond:
2570 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
2571 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !47
2572 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2573 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2574 // CHECK2:       omp.inner.for.body:
2575 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
2576 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2577 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2578 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !47
2579 // CHECK2-NEXT:    call void @_Z3fn1v(), !llvm.access.group !47
2580 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2581 // CHECK2:       omp.body.continue:
2582 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2583 // CHECK2:       omp.inner.for.inc:
2584 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
2585 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2586 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
2587 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
2588 // CHECK2:       omp.inner.for.end:
2589 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2590 // CHECK2:       omp.loop.exit:
2591 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2592 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2593 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2594 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2595 // CHECK2:       .omp.final.then:
2596 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
2597 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2598 // CHECK2:       .omp.final.done:
2599 // CHECK2-NEXT:    ret void
2600 //
2601 //
2602 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65
2603 // CHECK2-SAME: () #[[ATTR1]] {
2604 // CHECK2-NEXT:  entry:
2605 // CHECK2-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*, ...)*))
2606 // CHECK2-NEXT:    ret void
2607 //
2608 //
2609 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..12
2610 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2611 // CHECK2-NEXT:  entry:
2612 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2613 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2614 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2615 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2616 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2617 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2618 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2619 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2620 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2621 // CHECK2-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
2622 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2623 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2624 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2625 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2626 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2627 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2628 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2629 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2630 // CHECK2-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)
2631 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2632 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2633 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2634 // CHECK2:       cond.true:
2635 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2636 // CHECK2:       cond.false:
2637 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2638 // CHECK2-NEXT:    br label [[COND_END]]
2639 // CHECK2:       cond.end:
2640 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
2641 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2642 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2643 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
2644 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2645 // CHECK2:       omp.inner.for.cond:
2646 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
2647 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !50
2648 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
2649 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2650 // CHECK2:       omp.inner.for.body:
2651 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !50
2652 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
2653 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !50
2654 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
2655 // CHECK2-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !50
2656 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !50
2657 // CHECK2-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !50
2658 // CHECK2-NEXT:    call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !50
2659 // CHECK2-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !50
2660 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2661 // CHECK2:       omp.inner.for.inc:
2662 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
2663 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !50
2664 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
2665 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
2666 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
2667 // CHECK2:       omp.inner.for.end:
2668 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2669 // CHECK2:       omp.loop.exit:
2670 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
2671 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2672 // CHECK2-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
2673 // CHECK2-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2674 // CHECK2:       .omp.final.then:
2675 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
2676 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2677 // CHECK2:       .omp.final.done:
2678 // CHECK2-NEXT:    ret void
2679 //
2680 //
2681 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..13
2682 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2683 // CHECK2-NEXT:  entry:
2684 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2685 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2686 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2687 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2688 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2689 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2690 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2691 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2692 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2693 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2694 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2695 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2696 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2697 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2698 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2699 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2700 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
2701 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2702 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2703 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2704 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2705 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2706 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2707 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2708 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2709 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2710 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2711 // CHECK2-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)
2712 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2713 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2714 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2715 // CHECK2:       cond.true:
2716 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2717 // CHECK2:       cond.false:
2718 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2719 // CHECK2-NEXT:    br label [[COND_END]]
2720 // CHECK2:       cond.end:
2721 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2722 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2723 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2724 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2725 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2726 // CHECK2:       omp.inner.for.cond:
2727 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
2728 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !53
2729 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2730 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2731 // CHECK2:       omp.inner.for.body:
2732 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
2733 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2734 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2735 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !53
2736 // CHECK2-NEXT:    call void @_Z3fn2v(), !llvm.access.group !53
2737 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2738 // CHECK2:       omp.body.continue:
2739 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2740 // CHECK2:       omp.inner.for.inc:
2741 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
2742 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2743 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !53
2744 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP54:![0-9]+]]
2745 // CHECK2:       omp.inner.for.end:
2746 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2747 // CHECK2:       omp.loop.exit:
2748 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2749 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2750 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2751 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2752 // CHECK2:       .omp.final.then:
2753 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
2754 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2755 // CHECK2:       .omp.final.done:
2756 // CHECK2-NEXT:    ret void
2757 //
2758 //
2759 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71
2760 // CHECK2-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
2761 // CHECK2-NEXT:  entry:
2762 // CHECK2-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
2763 // CHECK2-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
2764 // CHECK2-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
2765 // CHECK2-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]])
2766 // CHECK2-NEXT:    ret void
2767 //
2768 //
2769 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..14
2770 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
2771 // CHECK2-NEXT:  entry:
2772 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2773 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2774 // CHECK2-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
2775 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2776 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2777 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2778 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2779 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2780 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2781 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2782 // CHECK2-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
2783 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2784 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2785 // CHECK2-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
2786 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
2787 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2788 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2789 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2790 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2791 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2792 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
2793 // CHECK2-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)
2794 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2795 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
2796 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2797 // CHECK2:       cond.true:
2798 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2799 // CHECK2:       cond.false:
2800 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2801 // CHECK2-NEXT:    br label [[COND_END]]
2802 // CHECK2:       cond.end:
2803 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
2804 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
2805 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
2806 // CHECK2-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
2807 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2808 // CHECK2:       omp.inner.for.cond:
2809 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
2810 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !56
2811 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
2812 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2813 // CHECK2:       omp.inner.for.body:
2814 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !56
2815 // CHECK2-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
2816 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !56
2817 // CHECK2-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
2818 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !56
2819 // CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
2820 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
2821 // CHECK2:       omp_if.then:
2822 // CHECK2-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
2823 // CHECK2-NEXT:    br label [[OMP_IF_END:%.*]]
2824 // CHECK2:       omp_if.else:
2825 // CHECK2-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !56
2826 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !56
2827 // CHECK2-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !56
2828 // CHECK2-NEXT:    call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !56
2829 // CHECK2-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !56
2830 // CHECK2-NEXT:    br label [[OMP_IF_END]]
2831 // CHECK2:       omp_if.end:
2832 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2833 // CHECK2:       omp.inner.for.inc:
2834 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
2835 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !56
2836 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
2837 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !56
2838 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP57:![0-9]+]]
2839 // CHECK2:       omp.inner.for.end:
2840 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2841 // CHECK2:       omp.loop.exit:
2842 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
2843 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2844 // CHECK2-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
2845 // CHECK2-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2846 // CHECK2:       .omp.final.then:
2847 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
2848 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2849 // CHECK2:       .omp.final.done:
2850 // CHECK2-NEXT:    ret void
2851 //
2852 //
2853 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..15
2854 // CHECK2-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
2855 // CHECK2-NEXT:  entry:
2856 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2857 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2858 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
2859 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
2860 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2861 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2862 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2863 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2864 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2865 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2866 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2867 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2868 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2869 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2870 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2871 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2872 // CHECK2-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
2873 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
2874 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
2875 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
2876 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
2877 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
2878 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
2879 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2880 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2881 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2882 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
2883 // CHECK2-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)
2884 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2885 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
2886 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2887 // CHECK2:       cond.true:
2888 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2889 // CHECK2:       cond.false:
2890 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2891 // CHECK2-NEXT:    br label [[COND_END]]
2892 // CHECK2:       cond.end:
2893 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
2894 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2895 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2896 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
2897 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2898 // CHECK2:       omp.inner.for.cond:
2899 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
2900 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !59
2901 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
2902 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2903 // CHECK2:       omp.inner.for.body:
2904 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
2905 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
2906 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2907 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !59
2908 // CHECK2-NEXT:    call void @_Z3fn3v(), !llvm.access.group !59
2909 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2910 // CHECK2:       omp.body.continue:
2911 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2912 // CHECK2:       omp.inner.for.inc:
2913 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
2914 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
2915 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
2916 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP60:![0-9]+]]
2917 // CHECK2:       omp.inner.for.end:
2918 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2919 // CHECK2:       omp.loop.exit:
2920 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
2921 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
2922 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
2923 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
2924 // CHECK2:       .omp.final.then:
2925 // CHECK2-NEXT:    store i32 100, i32* [[I]], align 4
2926 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
2927 // CHECK2:       .omp.final.done:
2928 // CHECK2-NEXT:    ret void
2929 //
2930 //
2931 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
2932 // CHECK2-SAME: () #[[ATTR5:[0-9]+]] {
2933 // CHECK2-NEXT:  entry:
2934 // CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
2935 // CHECK2-NEXT:    ret void
2936 //
2937 //
2938 // CHECK3-LABEL: define {{[^@]+}}@_Z9gtid_testv
2939 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
2940 // CHECK3-NEXT:  entry:
2941 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2942 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
2943 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
2944 // 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)
2945 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
2946 // CHECK3-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2947 // CHECK3:       omp_offload.failed:
2948 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43() #[[ATTR2:[0-9]+]]
2949 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2950 // CHECK3:       omp_offload.cont:
2951 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
2952 // 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)
2953 // CHECK3-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
2954 // CHECK3-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
2955 // CHECK3:       omp_offload.failed2:
2956 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48() #[[ATTR2]]
2957 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
2958 // CHECK3:       omp_offload.cont3:
2959 // CHECK3-NEXT:    ret void
2960 //
2961 //
2962 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43
2963 // CHECK3-SAME: () #[[ATTR1:[0-9]+]] {
2964 // CHECK3-NEXT:  entry:
2965 // 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*, ...)*))
2966 // CHECK3-NEXT:    ret void
2967 //
2968 //
2969 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
2970 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
2971 // CHECK3-NEXT:  entry:
2972 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2973 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2974 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2975 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2976 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2977 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2978 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2979 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2980 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2981 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2982 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2983 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
2984 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
2985 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2986 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2987 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2988 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
2989 // 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)
2990 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2991 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
2992 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2993 // CHECK3:       cond.true:
2994 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2995 // CHECK3:       cond.false:
2996 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
2997 // CHECK3-NEXT:    br label [[COND_END]]
2998 // CHECK3:       cond.end:
2999 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
3000 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3001 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3002 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
3003 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3004 // CHECK3:       omp.inner.for.cond:
3005 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
3006 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !11
3007 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3008 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3009 // CHECK3:       omp.inner.for.body:
3010 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !11
3011 // CHECK3-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
3012 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !11
3013 // CHECK3-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
3014 // 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
3015 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3016 // CHECK3:       omp.inner.for.inc:
3017 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
3018 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !11
3019 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
3020 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
3021 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
3022 // CHECK3:       omp.inner.for.end:
3023 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3024 // CHECK3:       omp.loop.exit:
3025 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
3026 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3027 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
3028 // CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3029 // CHECK3:       .omp.final.then:
3030 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
3031 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3032 // CHECK3:       .omp.final.done:
3033 // CHECK3-NEXT:    ret void
3034 //
3035 //
3036 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
3037 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
3038 // CHECK3-NEXT:  entry:
3039 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3040 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3041 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3042 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3043 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3044 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3045 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3046 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3047 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3048 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3049 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3050 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3051 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3052 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3053 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3054 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3055 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3056 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3057 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
3058 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3059 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3060 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3061 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3062 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3063 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3064 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3065 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3066 // 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)
3067 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3068 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3069 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3070 // CHECK3:       cond.true:
3071 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3072 // CHECK3:       cond.false:
3073 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3074 // CHECK3-NEXT:    br label [[COND_END]]
3075 // CHECK3:       cond.end:
3076 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3077 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3078 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3079 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3080 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3081 // CHECK3:       omp.inner.for.cond:
3082 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
3083 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !15
3084 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3085 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3086 // CHECK3:       omp.inner.for.body:
3087 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
3088 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3089 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3090 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !15
3091 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3092 // CHECK3:       omp.body.continue:
3093 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3094 // CHECK3:       omp.inner.for.inc:
3095 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
3096 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3097 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
3098 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
3099 // CHECK3:       omp.inner.for.end:
3100 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3101 // CHECK3:       omp.loop.exit:
3102 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3103 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3104 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3105 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3106 // CHECK3:       .omp.final.then:
3107 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
3108 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3109 // CHECK3:       .omp.final.done:
3110 // CHECK3-NEXT:    ret void
3111 //
3112 //
3113 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48
3114 // CHECK3-SAME: () #[[ATTR1]] {
3115 // CHECK3-NEXT:  entry:
3116 // 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*, ...)*))
3117 // CHECK3-NEXT:    ret void
3118 //
3119 //
3120 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
3121 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3122 // CHECK3-NEXT:  entry:
3123 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3124 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3125 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3126 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3127 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3128 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3129 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3130 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3131 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3132 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
3133 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3134 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3135 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3136 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3137 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3138 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3139 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3140 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3141 // 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)
3142 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3143 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
3144 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3145 // CHECK3:       cond.true:
3146 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3147 // CHECK3:       cond.false:
3148 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3149 // CHECK3-NEXT:    br label [[COND_END]]
3150 // CHECK3:       cond.end:
3151 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
3152 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3153 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3154 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
3155 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3156 // CHECK3:       omp.inner.for.cond:
3157 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
3158 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !20
3159 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3160 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3161 // CHECK3:       omp.inner.for.body:
3162 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !20
3163 // CHECK3-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
3164 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !20
3165 // CHECK3-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
3166 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !20
3167 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !20
3168 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !20
3169 // CHECK3-NEXT:    call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !20
3170 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !20
3171 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3172 // CHECK3:       omp.inner.for.inc:
3173 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
3174 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !20
3175 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3176 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
3177 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
3178 // CHECK3:       omp.inner.for.end:
3179 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3180 // CHECK3:       omp.loop.exit:
3181 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
3182 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3183 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
3184 // CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3185 // CHECK3:       .omp.final.then:
3186 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
3187 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3188 // CHECK3:       .omp.final.done:
3189 // CHECK3-NEXT:    ret void
3190 //
3191 //
3192 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
3193 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
3194 // CHECK3-NEXT:  entry:
3195 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3196 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3197 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3198 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3199 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3200 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3201 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3202 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3203 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3204 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3205 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3206 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3207 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3208 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3209 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3210 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3211 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3212 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3213 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
3214 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3215 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3216 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3217 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3218 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3219 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3220 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3221 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3222 // 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)
3223 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3224 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3225 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3226 // CHECK3:       cond.true:
3227 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3228 // CHECK3:       cond.false:
3229 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3230 // CHECK3-NEXT:    br label [[COND_END]]
3231 // CHECK3:       cond.end:
3232 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3233 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3234 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3235 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3236 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3237 // CHECK3:       omp.inner.for.cond:
3238 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
3239 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !23
3240 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3241 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3242 // CHECK3:       omp.inner.for.body:
3243 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
3244 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3245 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3246 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !23
3247 // CHECK3-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !23
3248 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3249 // CHECK3:       omp.body.continue:
3250 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3251 // CHECK3:       omp.inner.for.inc:
3252 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
3253 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3254 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
3255 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
3256 // CHECK3:       omp.inner.for.end:
3257 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3258 // CHECK3:       omp.loop.exit:
3259 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3260 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3261 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3262 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3263 // CHECK3:       .omp.final.then:
3264 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
3265 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3266 // CHECK3:       .omp.final.done:
3267 // CHECK3-NEXT:    ret void
3268 //
3269 //
3270 // CHECK3-LABEL: define {{[^@]+}}@main
3271 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
3272 // CHECK3-NEXT:  entry:
3273 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
3274 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3275 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
3276 // CHECK3-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
3277 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
3278 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
3279 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
3280 // CHECK3-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
3281 // CHECK3-NEXT:    store i32 0, i32* [[RETVAL]], align 4
3282 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
3283 // 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)
3284 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
3285 // CHECK3-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
3286 // CHECK3:       omp_offload.failed:
3287 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81() #[[ATTR2]]
3288 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
3289 // CHECK3:       omp_offload.cont:
3290 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
3291 // 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)
3292 // CHECK3-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
3293 // CHECK3-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
3294 // CHECK3:       omp_offload.failed2:
3295 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90() #[[ATTR2]]
3296 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
3297 // CHECK3:       omp_offload.cont3:
3298 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* @Arg, align 4
3299 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
3300 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
3301 // CHECK3-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
3302 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3303 // CHECK3-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
3304 // CHECK3-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
3305 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3306 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
3307 // CHECK3-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
3308 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
3309 // CHECK3-NEXT:    store i8* null, i8** [[TMP10]], align 8
3310 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
3311 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
3312 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
3313 // 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)
3314 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
3315 // CHECK3-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
3316 // CHECK3:       omp_offload.failed5:
3317 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99(i64 [[TMP5]]) #[[ATTR2]]
3318 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
3319 // CHECK3:       omp_offload.cont6:
3320 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
3321 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
3322 // CHECK3-NEXT:    ret i32 [[CALL]]
3323 //
3324 //
3325 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
3326 // CHECK3-SAME: () #[[ATTR1]] {
3327 // CHECK3-NEXT:  entry:
3328 // 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*, ...)*))
3329 // CHECK3-NEXT:    ret void
3330 //
3331 //
3332 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
3333 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3334 // CHECK3-NEXT:  entry:
3335 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3336 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3337 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3338 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3339 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3340 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3341 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3342 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3343 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3344 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3345 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3346 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3347 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3348 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3349 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3350 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3351 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3352 // 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)
3353 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3354 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
3355 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3356 // CHECK3:       cond.true:
3357 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3358 // CHECK3:       cond.false:
3359 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3360 // CHECK3-NEXT:    br label [[COND_END]]
3361 // CHECK3:       cond.end:
3362 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
3363 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3364 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3365 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
3366 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3367 // CHECK3:       omp.inner.for.cond:
3368 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
3369 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !26
3370 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3371 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3372 // CHECK3:       omp.inner.for.body:
3373 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !26
3374 // CHECK3-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
3375 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !26
3376 // CHECK3-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
3377 // 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
3378 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3379 // CHECK3:       omp.inner.for.inc:
3380 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
3381 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !26
3382 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
3383 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
3384 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
3385 // CHECK3:       omp.inner.for.end:
3386 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3387 // CHECK3:       omp.loop.exit:
3388 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
3389 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3390 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
3391 // CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3392 // CHECK3:       .omp.final.then:
3393 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
3394 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3395 // CHECK3:       .omp.final.done:
3396 // CHECK3-NEXT:    ret void
3397 //
3398 //
3399 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..5
3400 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
3401 // CHECK3-NEXT:  entry:
3402 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3403 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3404 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3405 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3406 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3407 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3408 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3409 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3410 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3411 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3412 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3413 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3414 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3415 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3416 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3417 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3418 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3419 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3420 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
3421 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3422 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3423 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3424 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3425 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3426 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3427 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3428 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3429 // 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)
3430 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3431 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3432 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3433 // CHECK3:       cond.true:
3434 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3435 // CHECK3:       cond.false:
3436 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3437 // CHECK3-NEXT:    br label [[COND_END]]
3438 // CHECK3:       cond.end:
3439 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3440 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3441 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3442 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3443 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3444 // CHECK3:       omp.inner.for.cond:
3445 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
3446 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !29
3447 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3448 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3449 // CHECK3:       omp.inner.for.body:
3450 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
3451 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3452 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3453 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !29
3454 // CHECK3-NEXT:    call void @_Z3fn4v(), !llvm.access.group !29
3455 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3456 // CHECK3:       omp.body.continue:
3457 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3458 // CHECK3:       omp.inner.for.inc:
3459 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
3460 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3461 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
3462 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
3463 // CHECK3:       omp.inner.for.end:
3464 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3465 // CHECK3:       omp.loop.exit:
3466 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3467 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3468 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3469 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3470 // CHECK3:       .omp.final.then:
3471 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
3472 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3473 // CHECK3:       .omp.final.done:
3474 // CHECK3-NEXT:    ret void
3475 //
3476 //
3477 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90
3478 // CHECK3-SAME: () #[[ATTR1]] {
3479 // CHECK3-NEXT:  entry:
3480 // 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*, ...)*))
3481 // CHECK3-NEXT:    ret void
3482 //
3483 //
3484 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..6
3485 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
3486 // CHECK3-NEXT:  entry:
3487 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3488 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3489 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3490 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3491 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3492 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3493 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3494 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3495 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3496 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
3497 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3498 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3499 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3500 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3501 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3502 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3503 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3504 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3505 // 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)
3506 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3507 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
3508 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3509 // CHECK3:       cond.true:
3510 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3511 // CHECK3:       cond.false:
3512 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3513 // CHECK3-NEXT:    br label [[COND_END]]
3514 // CHECK3:       cond.end:
3515 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
3516 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3517 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3518 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
3519 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3520 // CHECK3:       omp.inner.for.cond:
3521 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3522 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3523 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
3524 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3525 // CHECK3:       omp.inner.for.body:
3526 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3527 // CHECK3-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
3528 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3529 // CHECK3-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
3530 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
3531 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3532 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
3533 // CHECK3-NEXT:    call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
3534 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
3535 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3536 // CHECK3:       omp.inner.for.inc:
3537 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3538 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3539 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
3540 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
3541 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
3542 // CHECK3:       omp.inner.for.end:
3543 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3544 // CHECK3:       omp.loop.exit:
3545 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
3546 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3547 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
3548 // CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3549 // CHECK3:       .omp.final.then:
3550 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
3551 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3552 // CHECK3:       .omp.final.done:
3553 // CHECK3-NEXT:    ret void
3554 //
3555 //
3556 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
3557 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
3558 // CHECK3-NEXT:  entry:
3559 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3560 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3561 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3562 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3563 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3564 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3565 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3566 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3567 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3568 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3569 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3570 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3571 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3572 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3573 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3574 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3575 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3576 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3577 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
3578 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3579 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
3580 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
3581 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
3582 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3583 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3584 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3585 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3586 // 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)
3587 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3588 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3589 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3590 // CHECK3:       cond.true:
3591 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3592 // CHECK3:       cond.false:
3593 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3594 // CHECK3-NEXT:    br label [[COND_END]]
3595 // CHECK3:       cond.end:
3596 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3597 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3598 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3599 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3600 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3601 // CHECK3:       omp.inner.for.cond:
3602 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3603 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3604 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
3605 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3606 // CHECK3:       omp.inner.for.body:
3607 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3608 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
3609 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3610 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
3611 // CHECK3-NEXT:    call void @_Z3fn5v()
3612 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3613 // CHECK3:       omp.body.continue:
3614 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3615 // CHECK3:       omp.inner.for.inc:
3616 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3617 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
3618 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
3619 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
3620 // CHECK3:       omp.inner.for.end:
3621 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3622 // CHECK3:       omp.loop.exit:
3623 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3624 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3625 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
3626 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3627 // CHECK3:       .omp.final.then:
3628 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
3629 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3630 // CHECK3:       .omp.final.done:
3631 // CHECK3-NEXT:    ret void
3632 //
3633 //
3634 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99
3635 // CHECK3-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
3636 // CHECK3-NEXT:  entry:
3637 // CHECK3-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
3638 // CHECK3-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
3639 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
3640 // 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]])
3641 // CHECK3-NEXT:    ret void
3642 //
3643 //
3644 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..8
3645 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
3646 // CHECK3-NEXT:  entry:
3647 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3648 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3649 // CHECK3-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
3650 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
3651 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3652 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3653 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
3654 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
3655 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3656 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3657 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3658 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
3659 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
3660 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__CASTED12:%.*]] = alloca i64, align 8
3661 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR18:%.*]] = alloca i32, align 4
3662 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3663 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3664 // CHECK3-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
3665 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
3666 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
3667 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
3668 // CHECK3-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
3669 // CHECK3-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
3670 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
3671 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
3672 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3673 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3674 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3675 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
3676 // 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)
3677 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3678 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
3679 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3680 // CHECK3:       cond.true:
3681 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3682 // CHECK3:       cond.false:
3683 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3684 // CHECK3-NEXT:    br label [[COND_END]]
3685 // CHECK3:       cond.end:
3686 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
3687 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
3688 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3689 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
3690 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
3691 // CHECK3-NEXT:    [[TOBOOL1:%.*]] = trunc i8 [[TMP7]] to i1
3692 // CHECK3-NEXT:    br i1 [[TOBOOL1]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE7:%.*]]
3693 // CHECK3:       omp_if.then:
3694 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3695 // CHECK3:       omp.inner.for.cond:
3696 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
3697 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
3698 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
3699 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3700 // CHECK3:       omp.inner.for.body:
3701 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
3702 // CHECK3-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
3703 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
3704 // CHECK3-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
3705 // CHECK3-NEXT:    [[TMP14:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !35
3706 // CHECK3-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP14]] to i1
3707 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
3708 // CHECK3-NEXT:    [[FROMBOOL4:%.*]] = zext i1 [[TOBOOL3]] to i8
3709 // CHECK3-NEXT:    store i8 [[FROMBOOL4]], i8* [[CONV]], align 1, !llvm.access.group !35
3710 // CHECK3-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !35
3711 // CHECK3-NEXT:    [[TMP16:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !35
3712 // CHECK3-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP16]] to i1
3713 // CHECK3-NEXT:    br i1 [[TOBOOL5]], label [[OMP_IF_THEN6:%.*]], label [[OMP_IF_ELSE:%.*]]
3714 // CHECK3:       omp_if.then6:
3715 // 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
3716 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
3717 // CHECK3:       omp_if.else:
3718 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]]), !llvm.access.group !35
3719 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !35
3720 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !35
3721 // CHECK3-NEXT:    call void @.omp_outlined..9(i32* [[TMP17]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]]) #[[ATTR2]], !llvm.access.group !35
3722 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]]), !llvm.access.group !35
3723 // CHECK3-NEXT:    br label [[OMP_IF_END]]
3724 // CHECK3:       omp_if.end:
3725 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3726 // CHECK3:       omp.inner.for.inc:
3727 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
3728 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
3729 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
3730 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
3731 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
3732 // CHECK3:       omp.inner.for.end:
3733 // CHECK3-NEXT:    br label [[OMP_IF_END23:%.*]]
3734 // CHECK3:       omp_if.else7:
3735 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
3736 // CHECK3:       omp.inner.for.cond8:
3737 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3738 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3739 // CHECK3-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
3740 // CHECK3-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END22:%.*]]
3741 // CHECK3:       omp.inner.for.body10:
3742 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
3743 // CHECK3-NEXT:    [[TMP23:%.*]] = zext i32 [[TMP22]] to i64
3744 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
3745 // CHECK3-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP24]] to i64
3746 // CHECK3-NEXT:    [[TMP26:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
3747 // CHECK3-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP26]] to i1
3748 // CHECK3-NEXT:    [[CONV13:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED12]] to i8*
3749 // CHECK3-NEXT:    [[FROMBOOL14:%.*]] = zext i1 [[TOBOOL11]] to i8
3750 // CHECK3-NEXT:    store i8 [[FROMBOOL14]], i8* [[CONV13]], align 1
3751 // CHECK3-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED12]], align 8
3752 // CHECK3-NEXT:    [[TMP28:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
3753 // CHECK3-NEXT:    [[TOBOOL15:%.*]] = trunc i8 [[TMP28]] to i1
3754 // CHECK3-NEXT:    br i1 [[TOBOOL15]], label [[OMP_IF_THEN16:%.*]], label [[OMP_IF_ELSE17:%.*]]
3755 // CHECK3:       omp_if.then16:
3756 // 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]])
3757 // CHECK3-NEXT:    br label [[OMP_IF_END19:%.*]]
3758 // CHECK3:       omp_if.else17:
3759 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
3760 // CHECK3-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3761 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR18]], align 4
3762 // CHECK3-NEXT:    call void @.omp_outlined..10(i32* [[TMP29]], i32* [[DOTBOUND_ZERO_ADDR18]], i64 [[TMP23]], i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR2]]
3763 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
3764 // CHECK3-NEXT:    br label [[OMP_IF_END19]]
3765 // CHECK3:       omp_if.end19:
3766 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC20:%.*]]
3767 // CHECK3:       omp.inner.for.inc20:
3768 // CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3769 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3770 // CHECK3-NEXT:    [[ADD21:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
3771 // CHECK3-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
3772 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP38:![0-9]+]]
3773 // CHECK3:       omp.inner.for.end22:
3774 // CHECK3-NEXT:    br label [[OMP_IF_END23]]
3775 // CHECK3:       omp_if.end23:
3776 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3777 // CHECK3:       omp.loop.exit:
3778 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
3779 // CHECK3-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3780 // CHECK3-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
3781 // CHECK3-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3782 // CHECK3:       .omp.final.then:
3783 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
3784 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3785 // CHECK3:       .omp.final.done:
3786 // CHECK3-NEXT:    ret void
3787 //
3788 //
3789 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..9
3790 // 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]] {
3791 // CHECK3-NEXT:  entry:
3792 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3793 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3794 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3795 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3796 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3797 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3798 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3799 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3800 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3801 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3802 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3803 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3804 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3805 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3806 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3807 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3808 // CHECK3-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3809 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
3810 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3811 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3812 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3813 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP0]] to i32
3814 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3815 // CHECK3-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
3816 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
3817 // CHECK3-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
3818 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3819 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3820 // CHECK3-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
3821 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
3822 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3823 // CHECK3:       omp_if.then:
3824 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3825 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
3826 // 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)
3827 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3828 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
3829 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3830 // CHECK3:       cond.true:
3831 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3832 // CHECK3:       cond.false:
3833 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3834 // CHECK3-NEXT:    br label [[COND_END]]
3835 // CHECK3:       cond.end:
3836 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
3837 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3838 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3839 // CHECK3-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
3840 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3841 // CHECK3:       omp.inner.for.cond:
3842 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
3843 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
3844 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
3845 // CHECK3-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3846 // CHECK3:       omp.inner.for.body:
3847 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
3848 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
3849 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3850 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
3851 // CHECK3-NEXT:    call void @_Z3fn6v(), !llvm.access.group !39
3852 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3853 // CHECK3:       omp.body.continue:
3854 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3855 // CHECK3:       omp.inner.for.inc:
3856 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
3857 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
3858 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
3859 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
3860 // CHECK3:       omp.inner.for.end:
3861 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
3862 // CHECK3:       omp_if.else:
3863 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3864 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
3865 // 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)
3866 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3867 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP14]], 99
3868 // CHECK3-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
3869 // CHECK3:       cond.true6:
3870 // CHECK3-NEXT:    br label [[COND_END8:%.*]]
3871 // CHECK3:       cond.false7:
3872 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3873 // CHECK3-NEXT:    br label [[COND_END8]]
3874 // CHECK3:       cond.end8:
3875 // CHECK3-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP15]], [[COND_FALSE7]] ]
3876 // CHECK3-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_UB]], align 4
3877 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3878 // CHECK3-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
3879 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND10:%.*]]
3880 // CHECK3:       omp.inner.for.cond10:
3881 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3882 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3883 // CHECK3-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
3884 // CHECK3-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY12:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
3885 // CHECK3:       omp.inner.for.body12:
3886 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3887 // CHECK3-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP19]], 1
3888 // CHECK3-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
3889 // CHECK3-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
3890 // CHECK3-NEXT:    call void @_Z3fn6v()
3891 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
3892 // CHECK3:       omp.body.continue15:
3893 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
3894 // CHECK3:       omp.inner.for.inc16:
3895 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3896 // CHECK3-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
3897 // CHECK3-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
3898 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND10]], !llvm.loop [[LOOP42:![0-9]+]]
3899 // CHECK3:       omp.inner.for.end18:
3900 // CHECK3-NEXT:    br label [[OMP_IF_END]]
3901 // CHECK3:       omp_if.end:
3902 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
3903 // CHECK3:       omp.loop.exit:
3904 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3905 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
3906 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
3907 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
3908 // CHECK3-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
3909 // CHECK3-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
3910 // CHECK3:       .omp.final.then:
3911 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
3912 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
3913 // CHECK3:       .omp.final.done:
3914 // CHECK3-NEXT:    ret void
3915 //
3916 //
3917 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..10
3918 // 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]] {
3919 // CHECK3-NEXT:  entry:
3920 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3921 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3922 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
3923 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
3924 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
3925 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3926 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3927 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3928 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3929 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3930 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3931 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3932 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3933 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3934 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3935 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3936 // CHECK3-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
3937 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
3938 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3939 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
3940 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
3941 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP0]] to i32
3942 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
3943 // CHECK3-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
3944 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
3945 // CHECK3-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
3946 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3947 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3948 // CHECK3-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
3949 // CHECK3-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
3950 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
3951 // CHECK3:       omp_if.then:
3952 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3953 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
3954 // 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)
3955 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3956 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
3957 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3958 // CHECK3:       cond.true:
3959 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3960 // CHECK3:       cond.false:
3961 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3962 // CHECK3-NEXT:    br label [[COND_END]]
3963 // CHECK3:       cond.end:
3964 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
3965 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3966 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3967 // CHECK3-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
3968 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3969 // CHECK3:       omp.inner.for.cond:
3970 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
3971 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !43
3972 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
3973 // CHECK3-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3974 // CHECK3:       omp.inner.for.body:
3975 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
3976 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
3977 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
3978 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !43
3979 // CHECK3-NEXT:    call void @_Z3fn6v(), !llvm.access.group !43
3980 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3981 // CHECK3:       omp.body.continue:
3982 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3983 // CHECK3:       omp.inner.for.inc:
3984 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
3985 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
3986 // CHECK3-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
3987 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
3988 // CHECK3:       omp.inner.for.end:
3989 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
3990 // CHECK3:       omp_if.else:
3991 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3992 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
3993 // 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)
3994 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3995 // CHECK3-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP14]], 99
3996 // CHECK3-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
3997 // CHECK3:       cond.true6:
3998 // CHECK3-NEXT:    br label [[COND_END8:%.*]]
3999 // CHECK3:       cond.false7:
4000 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4001 // CHECK3-NEXT:    br label [[COND_END8]]
4002 // CHECK3:       cond.end8:
4003 // CHECK3-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP15]], [[COND_FALSE7]] ]
4004 // CHECK3-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_UB]], align 4
4005 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4006 // CHECK3-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
4007 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND10:%.*]]
4008 // CHECK3:       omp.inner.for.cond10:
4009 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4010 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4011 // CHECK3-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
4012 // CHECK3-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY12:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
4013 // CHECK3:       omp.inner.for.body12:
4014 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4015 // CHECK3-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP19]], 1
4016 // CHECK3-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
4017 // CHECK3-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
4018 // CHECK3-NEXT:    call void @_Z3fn6v()
4019 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
4020 // CHECK3:       omp.body.continue15:
4021 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
4022 // CHECK3:       omp.inner.for.inc16:
4023 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4024 // CHECK3-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
4025 // CHECK3-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
4026 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND10]], !llvm.loop [[LOOP46:![0-9]+]]
4027 // CHECK3:       omp.inner.for.end18:
4028 // CHECK3-NEXT:    br label [[OMP_IF_END]]
4029 // CHECK3:       omp_if.end:
4030 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4031 // CHECK3:       omp.loop.exit:
4032 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4033 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
4034 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
4035 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4036 // CHECK3-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
4037 // CHECK3-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4038 // CHECK3:       .omp.final.then:
4039 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
4040 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4041 // CHECK3:       .omp.final.done:
4042 // CHECK3-NEXT:    ret void
4043 //
4044 //
4045 // CHECK3-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
4046 // CHECK3-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
4047 // CHECK3-NEXT:  entry:
4048 // CHECK3-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
4049 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4050 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
4051 // CHECK3-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
4052 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
4053 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
4054 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
4055 // CHECK3-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
4056 // CHECK3-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
4057 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4058 // 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)
4059 // CHECK3-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
4060 // CHECK3-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4061 // CHECK3:       omp_offload.failed:
4062 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59() #[[ATTR2]]
4063 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4064 // CHECK3:       omp_offload.cont:
4065 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4066 // 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)
4067 // CHECK3-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
4068 // CHECK3-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
4069 // CHECK3:       omp_offload.failed2:
4070 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65() #[[ATTR2]]
4071 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
4072 // CHECK3:       omp_offload.cont3:
4073 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
4074 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
4075 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
4076 // CHECK3-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
4077 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4078 // CHECK3-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
4079 // CHECK3-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
4080 // CHECK3-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4081 // CHECK3-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
4082 // CHECK3-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
4083 // CHECK3-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4084 // CHECK3-NEXT:    store i8* null, i8** [[TMP10]], align 8
4085 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4086 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4087 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4088 // 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)
4089 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4090 // CHECK3-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
4091 // CHECK3:       omp_offload.failed5:
4092 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71(i64 [[TMP5]]) #[[ATTR2]]
4093 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
4094 // CHECK3:       omp_offload.cont6:
4095 // CHECK3-NEXT:    ret i32 0
4096 //
4097 //
4098 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59
4099 // CHECK3-SAME: () #[[ATTR1]] {
4100 // CHECK3-NEXT:  entry:
4101 // 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*, ...)*))
4102 // CHECK3-NEXT:    ret void
4103 //
4104 //
4105 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..11
4106 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4107 // CHECK3-NEXT:  entry:
4108 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4109 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4110 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4111 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4112 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4113 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4114 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4115 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4116 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4117 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4118 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4119 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4120 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4121 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4122 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4123 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4124 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4125 // 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)
4126 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4127 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4128 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4129 // CHECK3:       cond.true:
4130 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4131 // CHECK3:       cond.false:
4132 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4133 // CHECK3-NEXT:    br label [[COND_END]]
4134 // CHECK3:       cond.end:
4135 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4136 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4137 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4138 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4139 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4140 // CHECK3:       omp.inner.for.cond:
4141 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
4142 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
4143 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4144 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4145 // CHECK3:       omp.inner.for.body:
4146 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !47
4147 // CHECK3-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4148 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
4149 // CHECK3-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4150 // 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
4151 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4152 // CHECK3:       omp.inner.for.inc:
4153 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
4154 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !47
4155 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
4156 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
4157 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
4158 // CHECK3:       omp.inner.for.end:
4159 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4160 // CHECK3:       omp.loop.exit:
4161 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4162 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4163 // CHECK3-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4164 // CHECK3-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4165 // CHECK3:       .omp.final.then:
4166 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
4167 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4168 // CHECK3:       .omp.final.done:
4169 // CHECK3-NEXT:    ret void
4170 //
4171 //
4172 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..12
4173 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4174 // CHECK3-NEXT:  entry:
4175 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4176 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4177 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4178 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4179 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4180 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4181 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4182 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4183 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4184 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4185 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4186 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4187 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4188 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4189 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4190 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4191 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
4192 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4193 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4194 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4195 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4196 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4197 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4198 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4199 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4200 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4201 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4202 // 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)
4203 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4204 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4205 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4206 // CHECK3:       cond.true:
4207 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4208 // CHECK3:       cond.false:
4209 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4210 // CHECK3-NEXT:    br label [[COND_END]]
4211 // CHECK3:       cond.end:
4212 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4213 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4214 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4215 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4216 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4217 // CHECK3:       omp.inner.for.cond:
4218 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
4219 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !50
4220 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4221 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4222 // CHECK3:       omp.inner.for.body:
4223 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
4224 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4225 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4226 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !50
4227 // CHECK3-NEXT:    call void @_Z3fn1v(), !llvm.access.group !50
4228 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4229 // CHECK3:       omp.body.continue:
4230 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4231 // CHECK3:       omp.inner.for.inc:
4232 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
4233 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4234 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
4235 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
4236 // CHECK3:       omp.inner.for.end:
4237 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4238 // CHECK3:       omp.loop.exit:
4239 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4240 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4241 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4242 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4243 // CHECK3:       .omp.final.then:
4244 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
4245 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4246 // CHECK3:       .omp.final.done:
4247 // CHECK3-NEXT:    ret void
4248 //
4249 //
4250 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65
4251 // CHECK3-SAME: () #[[ATTR1]] {
4252 // CHECK3-NEXT:  entry:
4253 // 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*, ...)*))
4254 // CHECK3-NEXT:    ret void
4255 //
4256 //
4257 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..13
4258 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4259 // CHECK3-NEXT:  entry:
4260 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4261 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4262 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4263 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4264 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4265 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4266 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4267 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4268 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4269 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
4270 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4271 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4272 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4273 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4274 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4275 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4276 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4277 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4278 // 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)
4279 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4280 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4281 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4282 // CHECK3:       cond.true:
4283 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4284 // CHECK3:       cond.false:
4285 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4286 // CHECK3-NEXT:    br label [[COND_END]]
4287 // CHECK3:       cond.end:
4288 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4289 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4290 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4291 // CHECK3-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4292 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4293 // CHECK3:       omp.inner.for.cond:
4294 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4295 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4296 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4297 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4298 // CHECK3:       omp.inner.for.body:
4299 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4300 // CHECK3-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4301 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4302 // CHECK3-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4303 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
4304 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4305 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
4306 // CHECK3-NEXT:    call void @.omp_outlined..14(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
4307 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
4308 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4309 // CHECK3:       omp.inner.for.inc:
4310 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4311 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4312 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4313 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4314 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP53:![0-9]+]]
4315 // CHECK3:       omp.inner.for.end:
4316 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4317 // CHECK3:       omp.loop.exit:
4318 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4319 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4320 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
4321 // CHECK3-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4322 // CHECK3:       .omp.final.then:
4323 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
4324 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4325 // CHECK3:       .omp.final.done:
4326 // CHECK3-NEXT:    ret void
4327 //
4328 //
4329 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..14
4330 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4331 // CHECK3-NEXT:  entry:
4332 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4333 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4334 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4335 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4336 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4337 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4338 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4339 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4340 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4341 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4342 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4343 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4344 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4345 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4346 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4347 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4348 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
4349 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4350 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4351 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4352 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4353 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4354 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4355 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4356 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4357 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4358 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4359 // 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)
4360 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4361 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4362 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4363 // CHECK3:       cond.true:
4364 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4365 // CHECK3:       cond.false:
4366 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4367 // CHECK3-NEXT:    br label [[COND_END]]
4368 // CHECK3:       cond.end:
4369 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4370 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4371 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4372 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4373 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4374 // CHECK3:       omp.inner.for.cond:
4375 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4376 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4377 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4378 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4379 // CHECK3:       omp.inner.for.body:
4380 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4381 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4382 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4383 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
4384 // CHECK3-NEXT:    call void @_Z3fn2v()
4385 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4386 // CHECK3:       omp.body.continue:
4387 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4388 // CHECK3:       omp.inner.for.inc:
4389 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4390 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4391 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
4392 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP54:![0-9]+]]
4393 // CHECK3:       omp.inner.for.end:
4394 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4395 // CHECK3:       omp.loop.exit:
4396 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4397 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4398 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4399 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4400 // CHECK3:       .omp.final.then:
4401 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
4402 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4403 // CHECK3:       .omp.final.done:
4404 // CHECK3-NEXT:    ret void
4405 //
4406 //
4407 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71
4408 // CHECK3-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
4409 // CHECK3-NEXT:  entry:
4410 // CHECK3-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
4411 // CHECK3-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
4412 // CHECK3-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
4413 // 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]])
4414 // CHECK3-NEXT:    ret void
4415 //
4416 //
4417 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..15
4418 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
4419 // CHECK3-NEXT:  entry:
4420 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4421 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4422 // CHECK3-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
4423 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4424 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4425 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4426 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4427 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4428 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4429 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4430 // CHECK3-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
4431 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4432 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4433 // CHECK3-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
4434 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
4435 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4436 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4437 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4438 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4439 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4440 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
4441 // 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)
4442 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4443 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
4444 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4445 // CHECK3:       cond.true:
4446 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4447 // CHECK3:       cond.false:
4448 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4449 // CHECK3-NEXT:    br label [[COND_END]]
4450 // CHECK3:       cond.end:
4451 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
4452 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4453 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4454 // CHECK3-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
4455 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4456 // CHECK3:       omp.inner.for.cond:
4457 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
4458 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !55
4459 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
4460 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4461 // CHECK3:       omp.inner.for.body:
4462 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !55
4463 // CHECK3-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
4464 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !55
4465 // CHECK3-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
4466 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !55
4467 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
4468 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
4469 // CHECK3:       omp_if.then:
4470 // 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
4471 // CHECK3-NEXT:    br label [[OMP_IF_END:%.*]]
4472 // CHECK3:       omp_if.else:
4473 // CHECK3-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !55
4474 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !55
4475 // CHECK3-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !55
4476 // CHECK3-NEXT:    call void @.omp_outlined..16(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !55
4477 // CHECK3-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !55
4478 // CHECK3-NEXT:    br label [[OMP_IF_END]]
4479 // CHECK3:       omp_if.end:
4480 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4481 // CHECK3:       omp.inner.for.inc:
4482 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
4483 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !55
4484 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
4485 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
4486 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP56:![0-9]+]]
4487 // CHECK3:       omp.inner.for.end:
4488 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4489 // CHECK3:       omp.loop.exit:
4490 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
4491 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4492 // CHECK3-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
4493 // CHECK3-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4494 // CHECK3:       .omp.final.then:
4495 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
4496 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4497 // CHECK3:       .omp.final.done:
4498 // CHECK3-NEXT:    ret void
4499 //
4500 //
4501 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..16
4502 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4503 // CHECK3-NEXT:  entry:
4504 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4505 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4506 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4507 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4508 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4509 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4510 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4511 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4512 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4513 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4514 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
4515 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4516 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4517 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4518 // CHECK3-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4519 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4520 // CHECK3-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
4521 // CHECK3-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4522 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4523 // CHECK3-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4524 // CHECK3-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4525 // CHECK3-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4526 // CHECK3-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4527 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4528 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4529 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4530 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4531 // 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)
4532 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4533 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4534 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4535 // CHECK3:       cond.true:
4536 // CHECK3-NEXT:    br label [[COND_END:%.*]]
4537 // CHECK3:       cond.false:
4538 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4539 // CHECK3-NEXT:    br label [[COND_END]]
4540 // CHECK3:       cond.end:
4541 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4542 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4543 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4544 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4545 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4546 // CHECK3:       omp.inner.for.cond:
4547 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
4548 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !58
4549 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4550 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4551 // CHECK3:       omp.inner.for.body:
4552 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
4553 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4554 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4555 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !58
4556 // CHECK3-NEXT:    call void @_Z3fn3v(), !llvm.access.group !58
4557 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4558 // CHECK3:       omp.body.continue:
4559 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4560 // CHECK3:       omp.inner.for.inc:
4561 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
4562 // CHECK3-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4563 // CHECK3-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
4564 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP59:![0-9]+]]
4565 // CHECK3:       omp.inner.for.end:
4566 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4567 // CHECK3:       omp.loop.exit:
4568 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4569 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4570 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4571 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4572 // CHECK3:       .omp.final.then:
4573 // CHECK3-NEXT:    store i32 100, i32* [[I]], align 4
4574 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4575 // CHECK3:       .omp.final.done:
4576 // CHECK3-NEXT:    ret void
4577 //
4578 //
4579 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
4580 // CHECK3-SAME: () #[[ATTR5:[0-9]+]] {
4581 // CHECK3-NEXT:  entry:
4582 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
4583 // CHECK3-NEXT:    ret void
4584 //
4585 //
4586 // CHECK4-LABEL: define {{[^@]+}}@_Z9gtid_testv
4587 // CHECK4-SAME: () #[[ATTR0:[0-9]+]] {
4588 // CHECK4-NEXT:  entry:
4589 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4590 // CHECK4-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
4591 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
4592 // CHECK4-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)
4593 // CHECK4-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
4594 // CHECK4-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4595 // CHECK4:       omp_offload.failed:
4596 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43() #[[ATTR2:[0-9]+]]
4597 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4598 // CHECK4:       omp_offload.cont:
4599 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4600 // CHECK4-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)
4601 // CHECK4-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
4602 // CHECK4-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
4603 // CHECK4:       omp_offload.failed2:
4604 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48() #[[ATTR2]]
4605 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
4606 // CHECK4:       omp_offload.cont3:
4607 // CHECK4-NEXT:    ret void
4608 //
4609 //
4610 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43
4611 // CHECK4-SAME: () #[[ATTR1:[0-9]+]] {
4612 // CHECK4-NEXT:  entry:
4613 // CHECK4-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*, ...)*))
4614 // CHECK4-NEXT:    ret void
4615 //
4616 //
4617 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
4618 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4619 // CHECK4-NEXT:  entry:
4620 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4621 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4622 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4623 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4624 // CHECK4-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4625 // CHECK4-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4626 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4627 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4628 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
4629 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4630 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4631 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4632 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4633 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4634 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4635 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4636 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4637 // CHECK4-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)
4638 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4639 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4640 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4641 // CHECK4:       cond.true:
4642 // CHECK4-NEXT:    br label [[COND_END:%.*]]
4643 // CHECK4:       cond.false:
4644 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4645 // CHECK4-NEXT:    br label [[COND_END]]
4646 // CHECK4:       cond.end:
4647 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4648 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4649 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4650 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4651 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4652 // CHECK4:       omp.inner.for.cond:
4653 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
4654 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !11
4655 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4656 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4657 // CHECK4:       omp.inner.for.body:
4658 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !11
4659 // CHECK4-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4660 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !11
4661 // CHECK4-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4662 // CHECK4-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
4663 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4664 // CHECK4:       omp.inner.for.inc:
4665 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
4666 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !11
4667 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
4668 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !11
4669 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
4670 // CHECK4:       omp.inner.for.end:
4671 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4672 // CHECK4:       omp.loop.exit:
4673 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4674 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4675 // CHECK4-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4676 // CHECK4-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4677 // CHECK4:       .omp.final.then:
4678 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
4679 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4680 // CHECK4:       .omp.final.done:
4681 // CHECK4-NEXT:    ret void
4682 //
4683 //
4684 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
4685 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4686 // CHECK4-NEXT:  entry:
4687 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4688 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4689 // CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4690 // CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4691 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4692 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4693 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4694 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4695 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4696 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4697 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
4698 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4699 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4700 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4701 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4702 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4703 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
4704 // CHECK4-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4705 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4706 // CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4707 // CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4708 // CHECK4-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4709 // CHECK4-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4710 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4711 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4712 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4713 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4714 // CHECK4-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)
4715 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4716 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4717 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4718 // CHECK4:       cond.true:
4719 // CHECK4-NEXT:    br label [[COND_END:%.*]]
4720 // CHECK4:       cond.false:
4721 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4722 // CHECK4-NEXT:    br label [[COND_END]]
4723 // CHECK4:       cond.end:
4724 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4725 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4726 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4727 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4728 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4729 // CHECK4:       omp.inner.for.cond:
4730 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
4731 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !15
4732 // CHECK4-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4733 // CHECK4-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4734 // CHECK4:       omp.inner.for.body:
4735 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
4736 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4737 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4738 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !15
4739 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4740 // CHECK4:       omp.body.continue:
4741 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4742 // CHECK4:       omp.inner.for.inc:
4743 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
4744 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4745 // CHECK4-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
4746 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
4747 // CHECK4:       omp.inner.for.end:
4748 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4749 // CHECK4:       omp.loop.exit:
4750 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4751 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4752 // CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4753 // CHECK4-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4754 // CHECK4:       .omp.final.then:
4755 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
4756 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4757 // CHECK4:       .omp.final.done:
4758 // CHECK4-NEXT:    ret void
4759 //
4760 //
4761 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48
4762 // CHECK4-SAME: () #[[ATTR1]] {
4763 // CHECK4-NEXT:  entry:
4764 // CHECK4-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*, ...)*))
4765 // CHECK4-NEXT:    ret void
4766 //
4767 //
4768 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
4769 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4770 // CHECK4-NEXT:  entry:
4771 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4772 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4773 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4774 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4775 // CHECK4-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4776 // CHECK4-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4777 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4778 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4779 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
4780 // CHECK4-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
4781 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4782 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4783 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4784 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4785 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4786 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4787 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4788 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
4789 // CHECK4-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)
4790 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4791 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
4792 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4793 // CHECK4:       cond.true:
4794 // CHECK4-NEXT:    br label [[COND_END:%.*]]
4795 // CHECK4:       cond.false:
4796 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
4797 // CHECK4-NEXT:    br label [[COND_END]]
4798 // CHECK4:       cond.end:
4799 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
4800 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
4801 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
4802 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
4803 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4804 // CHECK4:       omp.inner.for.cond:
4805 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
4806 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !20
4807 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
4808 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4809 // CHECK4:       omp.inner.for.body:
4810 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !20
4811 // CHECK4-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
4812 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !20
4813 // CHECK4-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
4814 // CHECK4-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !20
4815 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !20
4816 // CHECK4-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !20
4817 // CHECK4-NEXT:    call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !20
4818 // CHECK4-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !20
4819 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4820 // CHECK4:       omp.inner.for.inc:
4821 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
4822 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !20
4823 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
4824 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !20
4825 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP21:![0-9]+]]
4826 // CHECK4:       omp.inner.for.end:
4827 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4828 // CHECK4:       omp.loop.exit:
4829 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
4830 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4831 // CHECK4-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
4832 // CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4833 // CHECK4:       .omp.final.then:
4834 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
4835 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4836 // CHECK4:       .omp.final.done:
4837 // CHECK4-NEXT:    ret void
4838 //
4839 //
4840 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3
4841 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
4842 // CHECK4-NEXT:  entry:
4843 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4844 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4845 // CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
4846 // CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
4847 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4848 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4849 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4850 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4851 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4852 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4853 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
4854 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4855 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4856 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4857 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4858 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4859 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
4860 // CHECK4-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
4861 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
4862 // CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
4863 // CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
4864 // CHECK4-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
4865 // CHECK4-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
4866 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4867 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4868 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4869 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
4870 // CHECK4-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)
4871 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4872 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
4873 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4874 // CHECK4:       cond.true:
4875 // CHECK4-NEXT:    br label [[COND_END:%.*]]
4876 // CHECK4:       cond.false:
4877 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4878 // CHECK4-NEXT:    br label [[COND_END]]
4879 // CHECK4:       cond.end:
4880 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
4881 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4882 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4883 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
4884 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4885 // CHECK4:       omp.inner.for.cond:
4886 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
4887 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !23
4888 // CHECK4-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
4889 // CHECK4-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4890 // CHECK4:       omp.inner.for.body:
4891 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
4892 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
4893 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
4894 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !23
4895 // CHECK4-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !23
4896 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4897 // CHECK4:       omp.body.continue:
4898 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4899 // CHECK4:       omp.inner.for.inc:
4900 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
4901 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
4902 // CHECK4-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !23
4903 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP24:![0-9]+]]
4904 // CHECK4:       omp.inner.for.end:
4905 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4906 // CHECK4:       omp.loop.exit:
4907 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
4908 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
4909 // CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
4910 // CHECK4-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
4911 // CHECK4:       .omp.final.then:
4912 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
4913 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
4914 // CHECK4:       .omp.final.done:
4915 // CHECK4-NEXT:    ret void
4916 //
4917 //
4918 // CHECK4-LABEL: define {{[^@]+}}@main
4919 // CHECK4-SAME: () #[[ATTR3:[0-9]+]] {
4920 // CHECK4-NEXT:  entry:
4921 // CHECK4-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
4922 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4923 // CHECK4-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
4924 // CHECK4-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
4925 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
4926 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
4927 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
4928 // CHECK4-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
4929 // CHECK4-NEXT:    store i32 0, i32* [[RETVAL]], align 4
4930 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4931 // CHECK4-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)
4932 // CHECK4-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
4933 // CHECK4-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
4934 // CHECK4:       omp_offload.failed:
4935 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81() #[[ATTR2]]
4936 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
4937 // CHECK4:       omp_offload.cont:
4938 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4939 // CHECK4-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)
4940 // CHECK4-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
4941 // CHECK4-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
4942 // CHECK4:       omp_offload.failed2:
4943 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90() #[[ATTR2]]
4944 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
4945 // CHECK4:       omp_offload.cont3:
4946 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* @Arg, align 4
4947 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
4948 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
4949 // CHECK4-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
4950 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4951 // CHECK4-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
4952 // CHECK4-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
4953 // CHECK4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4954 // CHECK4-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
4955 // CHECK4-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
4956 // CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
4957 // CHECK4-NEXT:    store i8* null, i8** [[TMP10]], align 8
4958 // CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
4959 // CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
4960 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
4961 // CHECK4-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)
4962 // CHECK4-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
4963 // CHECK4-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
4964 // CHECK4:       omp_offload.failed5:
4965 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99(i64 [[TMP5]]) #[[ATTR2]]
4966 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
4967 // CHECK4:       omp_offload.cont6:
4968 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
4969 // CHECK4-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
4970 // CHECK4-NEXT:    ret i32 [[CALL]]
4971 //
4972 //
4973 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
4974 // CHECK4-SAME: () #[[ATTR1]] {
4975 // CHECK4-NEXT:  entry:
4976 // CHECK4-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*, ...)*))
4977 // CHECK4-NEXT:    ret void
4978 //
4979 //
4980 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..4
4981 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
4982 // CHECK4-NEXT:  entry:
4983 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4984 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4985 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4986 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4987 // CHECK4-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
4988 // CHECK4-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
4989 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4990 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4991 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
4992 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4993 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4994 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
4995 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
4996 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4997 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4998 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4999 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5000 // CHECK4-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)
5001 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5002 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5003 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5004 // CHECK4:       cond.true:
5005 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5006 // CHECK4:       cond.false:
5007 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5008 // CHECK4-NEXT:    br label [[COND_END]]
5009 // CHECK4:       cond.end:
5010 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5011 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5012 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5013 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5014 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5015 // CHECK4:       omp.inner.for.cond:
5016 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
5017 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !26
5018 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5019 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5020 // CHECK4:       omp.inner.for.body:
5021 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !26
5022 // CHECK4-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5023 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !26
5024 // CHECK4-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5025 // CHECK4-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
5026 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5027 // CHECK4:       omp.inner.for.inc:
5028 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
5029 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !26
5030 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
5031 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !26
5032 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP27:![0-9]+]]
5033 // CHECK4:       omp.inner.for.end:
5034 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5035 // CHECK4:       omp.loop.exit:
5036 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5037 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5038 // CHECK4-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
5039 // CHECK4-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5040 // CHECK4:       .omp.final.then:
5041 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
5042 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5043 // CHECK4:       .omp.final.done:
5044 // CHECK4-NEXT:    ret void
5045 //
5046 //
5047 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..5
5048 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5049 // CHECK4-NEXT:  entry:
5050 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5051 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5052 // CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5053 // CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5054 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5055 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5056 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5057 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5058 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5059 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5060 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5061 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5062 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5063 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5064 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5065 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5066 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
5067 // CHECK4-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5068 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5069 // CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5070 // CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5071 // CHECK4-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5072 // CHECK4-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5073 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5074 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5075 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5076 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5077 // CHECK4-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)
5078 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5079 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5080 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5081 // CHECK4:       cond.true:
5082 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5083 // CHECK4:       cond.false:
5084 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5085 // CHECK4-NEXT:    br label [[COND_END]]
5086 // CHECK4:       cond.end:
5087 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5088 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5089 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5090 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5091 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5092 // CHECK4:       omp.inner.for.cond:
5093 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
5094 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !29
5095 // CHECK4-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5096 // CHECK4-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5097 // CHECK4:       omp.inner.for.body:
5098 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
5099 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5100 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5101 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !29
5102 // CHECK4-NEXT:    call void @_Z3fn4v(), !llvm.access.group !29
5103 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5104 // CHECK4:       omp.body.continue:
5105 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5106 // CHECK4:       omp.inner.for.inc:
5107 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
5108 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5109 // CHECK4-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !29
5110 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP30:![0-9]+]]
5111 // CHECK4:       omp.inner.for.end:
5112 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5113 // CHECK4:       omp.loop.exit:
5114 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5115 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5116 // CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
5117 // CHECK4-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5118 // CHECK4:       .omp.final.then:
5119 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
5120 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5121 // CHECK4:       .omp.final.done:
5122 // CHECK4-NEXT:    ret void
5123 //
5124 //
5125 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90
5126 // CHECK4-SAME: () #[[ATTR1]] {
5127 // CHECK4-NEXT:  entry:
5128 // CHECK4-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*, ...)*))
5129 // CHECK4-NEXT:    ret void
5130 //
5131 //
5132 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..6
5133 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
5134 // CHECK4-NEXT:  entry:
5135 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5136 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5137 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5138 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5139 // CHECK4-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5140 // CHECK4-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5141 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5142 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5143 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5144 // CHECK4-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
5145 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5146 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5147 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5148 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5149 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5150 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5151 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5152 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5153 // CHECK4-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)
5154 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5155 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5156 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5157 // CHECK4:       cond.true:
5158 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5159 // CHECK4:       cond.false:
5160 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5161 // CHECK4-NEXT:    br label [[COND_END]]
5162 // CHECK4:       cond.end:
5163 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5164 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5165 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5166 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5167 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5168 // CHECK4:       omp.inner.for.cond:
5169 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5170 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5171 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5172 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5173 // CHECK4:       omp.inner.for.body:
5174 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5175 // CHECK4-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5176 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5177 // CHECK4-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5178 // CHECK4-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
5179 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5180 // CHECK4-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
5181 // CHECK4-NEXT:    call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
5182 // CHECK4-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
5183 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5184 // CHECK4:       omp.inner.for.inc:
5185 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5186 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5187 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5188 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5189 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP32:![0-9]+]]
5190 // CHECK4:       omp.inner.for.end:
5191 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5192 // CHECK4:       omp.loop.exit:
5193 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5194 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5195 // CHECK4-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
5196 // CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5197 // CHECK4:       .omp.final.then:
5198 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
5199 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5200 // CHECK4:       .omp.final.done:
5201 // CHECK4-NEXT:    ret void
5202 //
5203 //
5204 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..7
5205 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5206 // CHECK4-NEXT:  entry:
5207 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5208 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5209 // CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5210 // CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5211 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5212 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5213 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5214 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5215 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5216 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5217 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5218 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5219 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5220 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5221 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5222 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5223 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
5224 // CHECK4-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5225 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5226 // CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5227 // CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5228 // CHECK4-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5229 // CHECK4-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5230 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5231 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5232 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5233 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5234 // CHECK4-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)
5235 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5236 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5237 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5238 // CHECK4:       cond.true:
5239 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5240 // CHECK4:       cond.false:
5241 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5242 // CHECK4-NEXT:    br label [[COND_END]]
5243 // CHECK4:       cond.end:
5244 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5245 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5246 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5247 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5248 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5249 // CHECK4:       omp.inner.for.cond:
5250 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5251 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5252 // CHECK4-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5253 // CHECK4-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5254 // CHECK4:       omp.inner.for.body:
5255 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5256 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5257 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5258 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
5259 // CHECK4-NEXT:    call void @_Z3fn5v()
5260 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5261 // CHECK4:       omp.body.continue:
5262 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5263 // CHECK4:       omp.inner.for.inc:
5264 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5265 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5266 // CHECK4-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
5267 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
5268 // CHECK4:       omp.inner.for.end:
5269 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5270 // CHECK4:       omp.loop.exit:
5271 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5272 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5273 // CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
5274 // CHECK4-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5275 // CHECK4:       .omp.final.then:
5276 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
5277 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5278 // CHECK4:       .omp.final.done:
5279 // CHECK4-NEXT:    ret void
5280 //
5281 //
5282 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99
5283 // CHECK4-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
5284 // CHECK4-NEXT:  entry:
5285 // CHECK4-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
5286 // CHECK4-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
5287 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
5288 // CHECK4-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]])
5289 // CHECK4-NEXT:    ret void
5290 //
5291 //
5292 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..8
5293 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
5294 // CHECK4-NEXT:  entry:
5295 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5296 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5297 // CHECK4-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
5298 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
5299 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5300 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5301 // CHECK4-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5302 // CHECK4-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5303 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5304 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5305 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5306 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
5307 // CHECK4-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
5308 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__CASTED12:%.*]] = alloca i64, align 8
5309 // CHECK4-NEXT:    [[DOTBOUND_ZERO_ADDR18:%.*]] = alloca i32, align 4
5310 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5311 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5312 // CHECK4-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
5313 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
5314 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5315 // CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
5316 // CHECK4-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
5317 // CHECK4-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
5318 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5319 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5320 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5321 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5322 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5323 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5324 // CHECK4-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)
5325 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5326 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5327 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5328 // CHECK4:       cond.true:
5329 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5330 // CHECK4:       cond.false:
5331 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5332 // CHECK4-NEXT:    br label [[COND_END]]
5333 // CHECK4:       cond.end:
5334 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5335 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5336 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5337 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5338 // CHECK4-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
5339 // CHECK4-NEXT:    [[TOBOOL1:%.*]] = trunc i8 [[TMP7]] to i1
5340 // CHECK4-NEXT:    br i1 [[TOBOOL1]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE7:%.*]]
5341 // CHECK4:       omp_if.then:
5342 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5343 // CHECK4:       omp.inner.for.cond:
5344 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
5345 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
5346 // CHECK4-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
5347 // CHECK4-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5348 // CHECK4:       omp.inner.for.body:
5349 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !35
5350 // CHECK4-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
5351 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !35
5352 // CHECK4-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
5353 // CHECK4-NEXT:    [[TMP14:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !35
5354 // CHECK4-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP14]] to i1
5355 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
5356 // CHECK4-NEXT:    [[FROMBOOL4:%.*]] = zext i1 [[TOBOOL3]] to i8
5357 // CHECK4-NEXT:    store i8 [[FROMBOOL4]], i8* [[CONV]], align 1, !llvm.access.group !35
5358 // CHECK4-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !35
5359 // CHECK4-NEXT:    [[TMP16:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !35
5360 // CHECK4-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP16]] to i1
5361 // CHECK4-NEXT:    br i1 [[TOBOOL5]], label [[OMP_IF_THEN6:%.*]], label [[OMP_IF_ELSE:%.*]]
5362 // CHECK4:       omp_if.then6:
5363 // CHECK4-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
5364 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
5365 // CHECK4:       omp_if.else:
5366 // CHECK4-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]]), !llvm.access.group !35
5367 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !35
5368 // CHECK4-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !35
5369 // CHECK4-NEXT:    call void @.omp_outlined..9(i32* [[TMP17]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]]) #[[ATTR2]], !llvm.access.group !35
5370 // CHECK4-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]]), !llvm.access.group !35
5371 // CHECK4-NEXT:    br label [[OMP_IF_END]]
5372 // CHECK4:       omp_if.end:
5373 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5374 // CHECK4:       omp.inner.for.inc:
5375 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
5376 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !35
5377 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
5378 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !35
5379 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
5380 // CHECK4:       omp.inner.for.end:
5381 // CHECK4-NEXT:    br label [[OMP_IF_END23:%.*]]
5382 // CHECK4:       omp_if.else7:
5383 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
5384 // CHECK4:       omp.inner.for.cond8:
5385 // CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5386 // CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5387 // CHECK4-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
5388 // CHECK4-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END22:%.*]]
5389 // CHECK4:       omp.inner.for.body10:
5390 // CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5391 // CHECK4-NEXT:    [[TMP23:%.*]] = zext i32 [[TMP22]] to i64
5392 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5393 // CHECK4-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP24]] to i64
5394 // CHECK4-NEXT:    [[TMP26:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
5395 // CHECK4-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP26]] to i1
5396 // CHECK4-NEXT:    [[CONV13:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED12]] to i8*
5397 // CHECK4-NEXT:    [[FROMBOOL14:%.*]] = zext i1 [[TOBOOL11]] to i8
5398 // CHECK4-NEXT:    store i8 [[FROMBOOL14]], i8* [[CONV13]], align 1
5399 // CHECK4-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED12]], align 8
5400 // CHECK4-NEXT:    [[TMP28:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
5401 // CHECK4-NEXT:    [[TOBOOL15:%.*]] = trunc i8 [[TMP28]] to i1
5402 // CHECK4-NEXT:    br i1 [[TOBOOL15]], label [[OMP_IF_THEN16:%.*]], label [[OMP_IF_ELSE17:%.*]]
5403 // CHECK4:       omp_if.then16:
5404 // CHECK4-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]])
5405 // CHECK4-NEXT:    br label [[OMP_IF_END19:%.*]]
5406 // CHECK4:       omp_if.else17:
5407 // CHECK4-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
5408 // CHECK4-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5409 // CHECK4-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR18]], align 4
5410 // CHECK4-NEXT:    call void @.omp_outlined..10(i32* [[TMP29]], i32* [[DOTBOUND_ZERO_ADDR18]], i64 [[TMP23]], i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR2]]
5411 // CHECK4-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
5412 // CHECK4-NEXT:    br label [[OMP_IF_END19]]
5413 // CHECK4:       omp_if.end19:
5414 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC20:%.*]]
5415 // CHECK4:       omp.inner.for.inc20:
5416 // CHECK4-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5417 // CHECK4-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5418 // CHECK4-NEXT:    [[ADD21:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
5419 // CHECK4-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
5420 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP38:![0-9]+]]
5421 // CHECK4:       omp.inner.for.end22:
5422 // CHECK4-NEXT:    br label [[OMP_IF_END23]]
5423 // CHECK4:       omp_if.end23:
5424 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5425 // CHECK4:       omp.loop.exit:
5426 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5427 // CHECK4-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5428 // CHECK4-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
5429 // CHECK4-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5430 // CHECK4:       .omp.final.then:
5431 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
5432 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5433 // CHECK4:       .omp.final.done:
5434 // CHECK4-NEXT:    ret void
5435 //
5436 //
5437 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..9
5438 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
5439 // CHECK4-NEXT:  entry:
5440 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5441 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5442 // CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5443 // CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5444 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
5445 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5446 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5447 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5448 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5449 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5450 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5451 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5452 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5453 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5454 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5455 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5456 // CHECK4-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
5457 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
5458 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5459 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
5460 // CHECK4-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5461 // CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP0]] to i32
5462 // CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5463 // CHECK4-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
5464 // CHECK4-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
5465 // CHECK4-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
5466 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5467 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5468 // CHECK4-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
5469 // CHECK4-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
5470 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
5471 // CHECK4:       omp_if.then:
5472 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5473 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
5474 // CHECK4-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)
5475 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5476 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
5477 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5478 // CHECK4:       cond.true:
5479 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5480 // CHECK4:       cond.false:
5481 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5482 // CHECK4-NEXT:    br label [[COND_END]]
5483 // CHECK4:       cond.end:
5484 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
5485 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5486 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5487 // CHECK4-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
5488 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5489 // CHECK4:       omp.inner.for.cond:
5490 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
5491 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
5492 // CHECK4-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
5493 // CHECK4-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5494 // CHECK4:       omp.inner.for.body:
5495 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
5496 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
5497 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5498 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
5499 // CHECK4-NEXT:    call void @_Z3fn6v(), !llvm.access.group !39
5500 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5501 // CHECK4:       omp.body.continue:
5502 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5503 // CHECK4:       omp.inner.for.inc:
5504 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
5505 // CHECK4-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
5506 // CHECK4-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
5507 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
5508 // CHECK4:       omp.inner.for.end:
5509 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
5510 // CHECK4:       omp_if.else:
5511 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5512 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
5513 // CHECK4-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)
5514 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5515 // CHECK4-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP14]], 99
5516 // CHECK4-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
5517 // CHECK4:       cond.true6:
5518 // CHECK4-NEXT:    br label [[COND_END8:%.*]]
5519 // CHECK4:       cond.false7:
5520 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5521 // CHECK4-NEXT:    br label [[COND_END8]]
5522 // CHECK4:       cond.end8:
5523 // CHECK4-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP15]], [[COND_FALSE7]] ]
5524 // CHECK4-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_UB]], align 4
5525 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5526 // CHECK4-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
5527 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND10:%.*]]
5528 // CHECK4:       omp.inner.for.cond10:
5529 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5530 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5531 // CHECK4-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
5532 // CHECK4-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY12:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
5533 // CHECK4:       omp.inner.for.body12:
5534 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5535 // CHECK4-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP19]], 1
5536 // CHECK4-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
5537 // CHECK4-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
5538 // CHECK4-NEXT:    call void @_Z3fn6v()
5539 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
5540 // CHECK4:       omp.body.continue15:
5541 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
5542 // CHECK4:       omp.inner.for.inc16:
5543 // CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5544 // CHECK4-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
5545 // CHECK4-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
5546 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND10]], !llvm.loop [[LOOP42:![0-9]+]]
5547 // CHECK4:       omp.inner.for.end18:
5548 // CHECK4-NEXT:    br label [[OMP_IF_END]]
5549 // CHECK4:       omp_if.end:
5550 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5551 // CHECK4:       omp.loop.exit:
5552 // CHECK4-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5553 // CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
5554 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
5555 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5556 // CHECK4-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
5557 // CHECK4-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5558 // CHECK4:       .omp.final.then:
5559 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
5560 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5561 // CHECK4:       .omp.final.done:
5562 // CHECK4-NEXT:    ret void
5563 //
5564 //
5565 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..10
5566 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
5567 // CHECK4-NEXT:  entry:
5568 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5569 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5570 // CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5571 // CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5572 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
5573 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5574 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5575 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5576 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5577 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5578 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5579 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5580 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5581 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5582 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5583 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5584 // CHECK4-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
5585 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
5586 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5587 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
5588 // CHECK4-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5589 // CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP0]] to i32
5590 // CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5591 // CHECK4-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
5592 // CHECK4-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
5593 // CHECK4-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
5594 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5595 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5596 // CHECK4-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
5597 // CHECK4-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
5598 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
5599 // CHECK4:       omp_if.then:
5600 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5601 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
5602 // CHECK4-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)
5603 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5604 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
5605 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5606 // CHECK4:       cond.true:
5607 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5608 // CHECK4:       cond.false:
5609 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5610 // CHECK4-NEXT:    br label [[COND_END]]
5611 // CHECK4:       cond.end:
5612 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
5613 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5614 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5615 // CHECK4-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
5616 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5617 // CHECK4:       omp.inner.for.cond:
5618 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
5619 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !43
5620 // CHECK4-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
5621 // CHECK4-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5622 // CHECK4:       omp.inner.for.body:
5623 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
5624 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
5625 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5626 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !43
5627 // CHECK4-NEXT:    call void @_Z3fn6v(), !llvm.access.group !43
5628 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5629 // CHECK4:       omp.body.continue:
5630 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5631 // CHECK4:       omp.inner.for.inc:
5632 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
5633 // CHECK4-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
5634 // CHECK4-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
5635 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
5636 // CHECK4:       omp.inner.for.end:
5637 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
5638 // CHECK4:       omp_if.else:
5639 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5640 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
5641 // CHECK4-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)
5642 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5643 // CHECK4-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP14]], 99
5644 // CHECK4-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
5645 // CHECK4:       cond.true6:
5646 // CHECK4-NEXT:    br label [[COND_END8:%.*]]
5647 // CHECK4:       cond.false7:
5648 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5649 // CHECK4-NEXT:    br label [[COND_END8]]
5650 // CHECK4:       cond.end8:
5651 // CHECK4-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP15]], [[COND_FALSE7]] ]
5652 // CHECK4-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_UB]], align 4
5653 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5654 // CHECK4-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
5655 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND10:%.*]]
5656 // CHECK4:       omp.inner.for.cond10:
5657 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5658 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5659 // CHECK4-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
5660 // CHECK4-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY12:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
5661 // CHECK4:       omp.inner.for.body12:
5662 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5663 // CHECK4-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP19]], 1
5664 // CHECK4-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
5665 // CHECK4-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
5666 // CHECK4-NEXT:    call void @_Z3fn6v()
5667 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
5668 // CHECK4:       omp.body.continue15:
5669 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
5670 // CHECK4:       omp.inner.for.inc16:
5671 // CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5672 // CHECK4-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
5673 // CHECK4-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
5674 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND10]], !llvm.loop [[LOOP46:![0-9]+]]
5675 // CHECK4:       omp.inner.for.end18:
5676 // CHECK4-NEXT:    br label [[OMP_IF_END]]
5677 // CHECK4:       omp_if.end:
5678 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5679 // CHECK4:       omp.loop.exit:
5680 // CHECK4-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5681 // CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
5682 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
5683 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5684 // CHECK4-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
5685 // CHECK4-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5686 // CHECK4:       .omp.final.then:
5687 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
5688 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5689 // CHECK4:       .omp.final.done:
5690 // CHECK4-NEXT:    ret void
5691 //
5692 //
5693 // CHECK4-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
5694 // CHECK4-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
5695 // CHECK4-NEXT:  entry:
5696 // CHECK4-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
5697 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5698 // CHECK4-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
5699 // CHECK4-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
5700 // CHECK4-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
5701 // CHECK4-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
5702 // CHECK4-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
5703 // CHECK4-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
5704 // CHECK4-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
5705 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
5706 // CHECK4-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)
5707 // CHECK4-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
5708 // CHECK4-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
5709 // CHECK4:       omp_offload.failed:
5710 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59() #[[ATTR2]]
5711 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT]]
5712 // CHECK4:       omp_offload.cont:
5713 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
5714 // CHECK4-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)
5715 // CHECK4-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
5716 // CHECK4-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
5717 // CHECK4:       omp_offload.failed2:
5718 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65() #[[ATTR2]]
5719 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
5720 // CHECK4:       omp_offload.cont3:
5721 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
5722 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
5723 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
5724 // CHECK4-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
5725 // CHECK4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5726 // CHECK4-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
5727 // CHECK4-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
5728 // CHECK4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5729 // CHECK4-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
5730 // CHECK4-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
5731 // CHECK4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
5732 // CHECK4-NEXT:    store i8* null, i8** [[TMP10]], align 8
5733 // CHECK4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
5734 // CHECK4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
5735 // CHECK4-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
5736 // CHECK4-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)
5737 // CHECK4-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
5738 // CHECK4-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
5739 // CHECK4:       omp_offload.failed5:
5740 // CHECK4-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71(i64 [[TMP5]]) #[[ATTR2]]
5741 // CHECK4-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
5742 // CHECK4:       omp_offload.cont6:
5743 // CHECK4-NEXT:    ret i32 0
5744 //
5745 //
5746 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59
5747 // CHECK4-SAME: () #[[ATTR1]] {
5748 // CHECK4-NEXT:  entry:
5749 // CHECK4-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*, ...)*))
5750 // CHECK4-NEXT:    ret void
5751 //
5752 //
5753 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..11
5754 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
5755 // CHECK4-NEXT:  entry:
5756 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5757 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5758 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5759 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5760 // CHECK4-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5761 // CHECK4-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5762 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5763 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5764 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5765 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5766 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5767 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5768 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5769 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5770 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5771 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5772 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5773 // CHECK4-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)
5774 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5775 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5776 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5777 // CHECK4:       cond.true:
5778 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5779 // CHECK4:       cond.false:
5780 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5781 // CHECK4-NEXT:    br label [[COND_END]]
5782 // CHECK4:       cond.end:
5783 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5784 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5785 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5786 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5787 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5788 // CHECK4:       omp.inner.for.cond:
5789 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
5790 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
5791 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5792 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5793 // CHECK4:       omp.inner.for.body:
5794 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !47
5795 // CHECK4-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5796 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !47
5797 // CHECK4-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5798 // CHECK4-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
5799 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5800 // CHECK4:       omp.inner.for.inc:
5801 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
5802 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !47
5803 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
5804 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
5805 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
5806 // CHECK4:       omp.inner.for.end:
5807 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5808 // CHECK4:       omp.loop.exit:
5809 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5810 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5811 // CHECK4-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
5812 // CHECK4-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5813 // CHECK4:       .omp.final.then:
5814 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
5815 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5816 // CHECK4:       .omp.final.done:
5817 // CHECK4-NEXT:    ret void
5818 //
5819 //
5820 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..12
5821 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5822 // CHECK4-NEXT:  entry:
5823 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5824 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5825 // CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5826 // CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5827 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5828 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5829 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5830 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5831 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5832 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5833 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5834 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5835 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5836 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5837 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5838 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5839 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
5840 // CHECK4-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5841 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5842 // CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5843 // CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
5844 // CHECK4-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
5845 // CHECK4-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
5846 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5847 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5848 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5849 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
5850 // CHECK4-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)
5851 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5852 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
5853 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5854 // CHECK4:       cond.true:
5855 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5856 // CHECK4:       cond.false:
5857 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
5858 // CHECK4-NEXT:    br label [[COND_END]]
5859 // CHECK4:       cond.end:
5860 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
5861 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
5862 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
5863 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
5864 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5865 // CHECK4:       omp.inner.for.cond:
5866 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
5867 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !50
5868 // CHECK4-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
5869 // CHECK4-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5870 // CHECK4:       omp.inner.for.body:
5871 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
5872 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
5873 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
5874 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !50
5875 // CHECK4-NEXT:    call void @_Z3fn1v(), !llvm.access.group !50
5876 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
5877 // CHECK4:       omp.body.continue:
5878 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5879 // CHECK4:       omp.inner.for.inc:
5880 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
5881 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
5882 // CHECK4-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !50
5883 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP51:![0-9]+]]
5884 // CHECK4:       omp.inner.for.end:
5885 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5886 // CHECK4:       omp.loop.exit:
5887 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
5888 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5889 // CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
5890 // CHECK4-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5891 // CHECK4:       .omp.final.then:
5892 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
5893 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5894 // CHECK4:       .omp.final.done:
5895 // CHECK4-NEXT:    ret void
5896 //
5897 //
5898 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65
5899 // CHECK4-SAME: () #[[ATTR1]] {
5900 // CHECK4-NEXT:  entry:
5901 // CHECK4-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*, ...)*))
5902 // CHECK4-NEXT:    ret void
5903 //
5904 //
5905 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..13
5906 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
5907 // CHECK4-NEXT:  entry:
5908 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5909 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5910 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5911 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5912 // CHECK4-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
5913 // CHECK4-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
5914 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5915 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5916 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5917 // CHECK4-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
5918 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5919 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5920 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
5921 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
5922 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
5923 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
5924 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5925 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
5926 // CHECK4-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)
5927 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5928 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
5929 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
5930 // CHECK4:       cond.true:
5931 // CHECK4-NEXT:    br label [[COND_END:%.*]]
5932 // CHECK4:       cond.false:
5933 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5934 // CHECK4-NEXT:    br label [[COND_END]]
5935 // CHECK4:       cond.end:
5936 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
5937 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
5938 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5939 // CHECK4-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
5940 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
5941 // CHECK4:       omp.inner.for.cond:
5942 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5943 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5944 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
5945 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
5946 // CHECK4:       omp.inner.for.body:
5947 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
5948 // CHECK4-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
5949 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
5950 // CHECK4-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
5951 // CHECK4-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
5952 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
5953 // CHECK4-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
5954 // CHECK4-NEXT:    call void @.omp_outlined..14(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
5955 // CHECK4-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
5956 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
5957 // CHECK4:       omp.inner.for.inc:
5958 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
5959 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
5960 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
5961 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
5962 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP53:![0-9]+]]
5963 // CHECK4:       omp.inner.for.end:
5964 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
5965 // CHECK4:       omp.loop.exit:
5966 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
5967 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
5968 // CHECK4-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
5969 // CHECK4-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
5970 // CHECK4:       .omp.final.then:
5971 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
5972 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
5973 // CHECK4:       .omp.final.done:
5974 // CHECK4-NEXT:    ret void
5975 //
5976 //
5977 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..14
5978 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
5979 // CHECK4-NEXT:  entry:
5980 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5981 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5982 // CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
5983 // CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
5984 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5985 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5986 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5987 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5988 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5989 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5990 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
5991 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5992 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5993 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5994 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
5995 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
5996 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
5997 // CHECK4-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
5998 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
5999 // CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6000 // CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
6001 // CHECK4-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
6002 // CHECK4-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
6003 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6004 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6005 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6006 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
6007 // CHECK4-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)
6008 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6009 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
6010 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6011 // CHECK4:       cond.true:
6012 // CHECK4-NEXT:    br label [[COND_END:%.*]]
6013 // CHECK4:       cond.false:
6014 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6015 // CHECK4-NEXT:    br label [[COND_END]]
6016 // CHECK4:       cond.end:
6017 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6018 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6019 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6020 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6021 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6022 // CHECK4:       omp.inner.for.cond:
6023 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6024 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6025 // CHECK4-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6026 // CHECK4-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6027 // CHECK4:       omp.inner.for.body:
6028 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6029 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6030 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6031 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
6032 // CHECK4-NEXT:    call void @_Z3fn2v()
6033 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6034 // CHECK4:       omp.body.continue:
6035 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6036 // CHECK4:       omp.inner.for.inc:
6037 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6038 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
6039 // CHECK4-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
6040 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP54:![0-9]+]]
6041 // CHECK4:       omp.inner.for.end:
6042 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6043 // CHECK4:       omp.loop.exit:
6044 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6045 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6046 // CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
6047 // CHECK4-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6048 // CHECK4:       .omp.final.then:
6049 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
6050 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6051 // CHECK4:       .omp.final.done:
6052 // CHECK4-NEXT:    ret void
6053 //
6054 //
6055 // CHECK4-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71
6056 // CHECK4-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
6057 // CHECK4-NEXT:  entry:
6058 // CHECK4-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
6059 // CHECK4-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
6060 // CHECK4-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
6061 // CHECK4-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]])
6062 // CHECK4-NEXT:    ret void
6063 //
6064 //
6065 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..15
6066 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
6067 // CHECK4-NEXT:  entry:
6068 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6069 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6070 // CHECK4-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
6071 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6072 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6073 // CHECK4-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
6074 // CHECK4-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
6075 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6076 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6077 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
6078 // CHECK4-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
6079 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6080 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6081 // CHECK4-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
6082 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
6083 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
6084 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
6085 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6086 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6087 // CHECK4-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6088 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
6089 // CHECK4-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)
6090 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6091 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
6092 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6093 // CHECK4:       cond.true:
6094 // CHECK4-NEXT:    br label [[COND_END:%.*]]
6095 // CHECK4:       cond.false:
6096 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
6097 // CHECK4-NEXT:    br label [[COND_END]]
6098 // CHECK4:       cond.end:
6099 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
6100 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
6101 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
6102 // CHECK4-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
6103 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6104 // CHECK4:       omp.inner.for.cond:
6105 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
6106 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !55
6107 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6108 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6109 // CHECK4:       omp.inner.for.body:
6110 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !55
6111 // CHECK4-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
6112 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !55
6113 // CHECK4-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
6114 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !55
6115 // CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
6116 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6117 // CHECK4:       omp_if.then:
6118 // CHECK4-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
6119 // CHECK4-NEXT:    br label [[OMP_IF_END:%.*]]
6120 // CHECK4:       omp_if.else:
6121 // CHECK4-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !55
6122 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !55
6123 // CHECK4-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !55
6124 // CHECK4-NEXT:    call void @.omp_outlined..16(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !55
6125 // CHECK4-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !55
6126 // CHECK4-NEXT:    br label [[OMP_IF_END]]
6127 // CHECK4:       omp_if.end:
6128 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6129 // CHECK4:       omp.inner.for.inc:
6130 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
6131 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !55
6132 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
6133 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !55
6134 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP56:![0-9]+]]
6135 // CHECK4:       omp.inner.for.end:
6136 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6137 // CHECK4:       omp.loop.exit:
6138 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
6139 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6140 // CHECK4-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
6141 // CHECK4-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6142 // CHECK4:       .omp.final.then:
6143 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
6144 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6145 // CHECK4:       .omp.final.done:
6146 // CHECK4-NEXT:    ret void
6147 //
6148 //
6149 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..16
6150 // CHECK4-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
6151 // CHECK4-NEXT:  entry:
6152 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6153 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6154 // CHECK4-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
6155 // CHECK4-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
6156 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6157 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6158 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6159 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6160 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6161 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6162 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
6163 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6164 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6165 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6166 // CHECK4-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6167 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6168 // CHECK4-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6169 // CHECK4-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
6170 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
6171 // CHECK4-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
6172 // CHECK4-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
6173 // CHECK4-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
6174 // CHECK4-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
6175 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6176 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6177 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6178 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
6179 // CHECK4-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)
6180 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6181 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
6182 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6183 // CHECK4:       cond.true:
6184 // CHECK4-NEXT:    br label [[COND_END:%.*]]
6185 // CHECK4:       cond.false:
6186 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6187 // CHECK4-NEXT:    br label [[COND_END]]
6188 // CHECK4:       cond.end:
6189 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
6190 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6191 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6192 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
6193 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6194 // CHECK4:       omp.inner.for.cond:
6195 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
6196 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !58
6197 // CHECK4-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
6198 // CHECK4-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6199 // CHECK4:       omp.inner.for.body:
6200 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
6201 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
6202 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6203 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !58
6204 // CHECK4-NEXT:    call void @_Z3fn3v(), !llvm.access.group !58
6205 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6206 // CHECK4:       omp.body.continue:
6207 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6208 // CHECK4:       omp.inner.for.inc:
6209 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
6210 // CHECK4-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
6211 // CHECK4-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !58
6212 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP59:![0-9]+]]
6213 // CHECK4:       omp.inner.for.end:
6214 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6215 // CHECK4:       omp.loop.exit:
6216 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
6217 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
6218 // CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
6219 // CHECK4-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
6220 // CHECK4:       .omp.final.then:
6221 // CHECK4-NEXT:    store i32 100, i32* [[I]], align 4
6222 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
6223 // CHECK4:       .omp.final.done:
6224 // CHECK4-NEXT:    ret void
6225 //
6226 //
6227 // CHECK4-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
6228 // CHECK4-SAME: () #[[ATTR5:[0-9]+]] {
6229 // CHECK4-NEXT:  entry:
6230 // CHECK4-NEXT:    call void @__tgt_register_requires(i64 1)
6231 // CHECK4-NEXT:    ret void
6232 //
6233 //
6234 // CHECK5-LABEL: define {{[^@]+}}@_Z9gtid_testv
6235 // CHECK5-SAME: () #[[ATTR0:[0-9]+]] {
6236 // CHECK5-NEXT:  entry:
6237 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6238 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6239 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6240 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6241 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
6242 // CHECK5-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
6243 // CHECK5-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
6244 // CHECK5-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
6245 // CHECK5-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
6246 // CHECK5-NEXT:    [[I6:%.*]] = alloca i32, align 4
6247 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6248 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6249 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6250 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
6251 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6252 // CHECK5:       omp.inner.for.cond:
6253 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6254 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
6255 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
6256 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6257 // CHECK5:       omp.inner.for.body:
6258 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6259 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
6260 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6261 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
6262 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6263 // CHECK5:       omp.body.continue:
6264 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6265 // CHECK5:       omp.inner.for.inc:
6266 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6267 // CHECK5-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
6268 // CHECK5-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6269 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
6270 // CHECK5:       omp.inner.for.end:
6271 // CHECK5-NEXT:    store i32 100, i32* [[I]], align 4
6272 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
6273 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
6274 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
6275 // CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
6276 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
6277 // CHECK5:       omp.inner.for.cond7:
6278 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
6279 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !6
6280 // CHECK5-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6281 // CHECK5-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
6282 // CHECK5:       omp.inner.for.body9:
6283 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
6284 // CHECK5-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
6285 // CHECK5-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
6286 // CHECK5-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !6
6287 // CHECK5-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !6
6288 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
6289 // CHECK5:       omp.body.continue12:
6290 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
6291 // CHECK5:       omp.inner.for.inc13:
6292 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
6293 // CHECK5-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
6294 // CHECK5-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
6295 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP7:![0-9]+]]
6296 // CHECK5:       omp.inner.for.end15:
6297 // CHECK5-NEXT:    store i32 100, i32* [[I6]], align 4
6298 // CHECK5-NEXT:    ret void
6299 //
6300 //
6301 // CHECK5-LABEL: define {{[^@]+}}@main
6302 // CHECK5-SAME: () #[[ATTR1:[0-9]+]] {
6303 // CHECK5-NEXT:  entry:
6304 // CHECK5-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
6305 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6306 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6307 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6308 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6309 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
6310 // CHECK5-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
6311 // CHECK5-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
6312 // CHECK5-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
6313 // CHECK5-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
6314 // CHECK5-NEXT:    [[I6:%.*]] = alloca i32, align 4
6315 // CHECK5-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
6316 // CHECK5-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
6317 // CHECK5-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
6318 // CHECK5-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
6319 // CHECK5-NEXT:    [[I20:%.*]] = alloca i32, align 4
6320 // CHECK5-NEXT:    store i32 0, i32* [[RETVAL]], align 4
6321 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6322 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6323 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6324 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
6325 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6326 // CHECK5:       omp.inner.for.cond:
6327 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
6328 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
6329 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
6330 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6331 // CHECK5:       omp.inner.for.body:
6332 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
6333 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
6334 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6335 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
6336 // CHECK5-NEXT:    call void @_Z3fn4v(), !llvm.access.group !9
6337 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6338 // CHECK5:       omp.body.continue:
6339 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6340 // CHECK5:       omp.inner.for.inc:
6341 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
6342 // CHECK5-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
6343 // CHECK5-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
6344 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
6345 // CHECK5:       omp.inner.for.end:
6346 // CHECK5-NEXT:    store i32 100, i32* [[I]], align 4
6347 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
6348 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
6349 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
6350 // CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
6351 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
6352 // CHECK5:       omp.inner.for.cond7:
6353 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !12
6354 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !12
6355 // CHECK5-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6356 // CHECK5-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
6357 // CHECK5:       omp.inner.for.body9:
6358 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !12
6359 // CHECK5-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
6360 // CHECK5-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
6361 // CHECK5-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !12
6362 // CHECK5-NEXT:    call void @_Z3fn5v(), !llvm.access.group !12
6363 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
6364 // CHECK5:       omp.body.continue12:
6365 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
6366 // CHECK5:       omp.inner.for.inc13:
6367 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !12
6368 // CHECK5-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
6369 // CHECK5-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !12
6370 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP13:![0-9]+]]
6371 // CHECK5:       omp.inner.for.end15:
6372 // CHECK5-NEXT:    store i32 100, i32* [[I6]], align 4
6373 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
6374 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
6375 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
6376 // CHECK5-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
6377 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
6378 // CHECK5:       omp.inner.for.cond21:
6379 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !15
6380 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !15
6381 // CHECK5-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
6382 // CHECK5-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
6383 // CHECK5:       omp.inner.for.body23:
6384 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !15
6385 // CHECK5-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
6386 // CHECK5-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
6387 // CHECK5-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !15
6388 // CHECK5-NEXT:    call void @_Z3fn6v(), !llvm.access.group !15
6389 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
6390 // CHECK5:       omp.body.continue26:
6391 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
6392 // CHECK5:       omp.inner.for.inc27:
6393 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !15
6394 // CHECK5-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
6395 // CHECK5-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !15
6396 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP16:![0-9]+]]
6397 // CHECK5:       omp.inner.for.end29:
6398 // CHECK5-NEXT:    store i32 100, i32* [[I20]], align 4
6399 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
6400 // CHECK5-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
6401 // CHECK5-NEXT:    ret i32 [[CALL]]
6402 //
6403 //
6404 // CHECK5-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
6405 // CHECK5-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
6406 // CHECK5-NEXT:  entry:
6407 // CHECK5-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
6408 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6409 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6410 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6411 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6412 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
6413 // CHECK5-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
6414 // CHECK5-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
6415 // CHECK5-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
6416 // CHECK5-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
6417 // CHECK5-NEXT:    [[I6:%.*]] = alloca i32, align 4
6418 // CHECK5-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
6419 // CHECK5-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
6420 // CHECK5-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
6421 // CHECK5-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
6422 // CHECK5-NEXT:    [[I20:%.*]] = alloca i32, align 4
6423 // CHECK5-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
6424 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6425 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6426 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6427 // CHECK5-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
6428 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6429 // CHECK5:       omp.inner.for.cond:
6430 // CHECK5-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
6431 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
6432 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
6433 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6434 // CHECK5:       omp.inner.for.body:
6435 // CHECK5-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
6436 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
6437 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6438 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !18
6439 // CHECK5-NEXT:    call void @_Z3fn1v(), !llvm.access.group !18
6440 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6441 // CHECK5:       omp.body.continue:
6442 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6443 // CHECK5:       omp.inner.for.inc:
6444 // CHECK5-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
6445 // CHECK5-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
6446 // CHECK5-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
6447 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
6448 // CHECK5:       omp.inner.for.end:
6449 // CHECK5-NEXT:    store i32 100, i32* [[I]], align 4
6450 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
6451 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
6452 // CHECK5-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
6453 // CHECK5-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
6454 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
6455 // CHECK5:       omp.inner.for.cond7:
6456 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
6457 // CHECK5-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !21
6458 // CHECK5-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6459 // CHECK5-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
6460 // CHECK5:       omp.inner.for.body9:
6461 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
6462 // CHECK5-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
6463 // CHECK5-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
6464 // CHECK5-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !21
6465 // CHECK5-NEXT:    call void @_Z3fn2v(), !llvm.access.group !21
6466 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
6467 // CHECK5:       omp.body.continue12:
6468 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
6469 // CHECK5:       omp.inner.for.inc13:
6470 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
6471 // CHECK5-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
6472 // CHECK5-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
6473 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP22:![0-9]+]]
6474 // CHECK5:       omp.inner.for.end15:
6475 // CHECK5-NEXT:    store i32 100, i32* [[I6]], align 4
6476 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
6477 // CHECK5-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
6478 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
6479 // CHECK5-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
6480 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
6481 // CHECK5:       omp.inner.for.cond21:
6482 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !24
6483 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !24
6484 // CHECK5-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
6485 // CHECK5-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
6486 // CHECK5:       omp.inner.for.body23:
6487 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !24
6488 // CHECK5-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
6489 // CHECK5-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
6490 // CHECK5-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !24
6491 // CHECK5-NEXT:    call void @_Z3fn3v(), !llvm.access.group !24
6492 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
6493 // CHECK5:       omp.body.continue26:
6494 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
6495 // CHECK5:       omp.inner.for.inc27:
6496 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !24
6497 // CHECK5-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
6498 // CHECK5-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !24
6499 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP25:![0-9]+]]
6500 // CHECK5:       omp.inner.for.end29:
6501 // CHECK5-NEXT:    store i32 100, i32* [[I20]], align 4
6502 // CHECK5-NEXT:    ret i32 0
6503 //
6504 //
6505 // CHECK6-LABEL: define {{[^@]+}}@_Z9gtid_testv
6506 // CHECK6-SAME: () #[[ATTR0:[0-9]+]] {
6507 // CHECK6-NEXT:  entry:
6508 // CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6509 // CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6510 // CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6511 // CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6512 // CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
6513 // CHECK6-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
6514 // CHECK6-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
6515 // CHECK6-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
6516 // CHECK6-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
6517 // CHECK6-NEXT:    [[I6:%.*]] = alloca i32, align 4
6518 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6519 // CHECK6-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6520 // CHECK6-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6521 // CHECK6-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
6522 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6523 // CHECK6:       omp.inner.for.cond:
6524 // CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6525 // CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
6526 // CHECK6-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
6527 // CHECK6-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6528 // CHECK6:       omp.inner.for.body:
6529 // CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6530 // CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
6531 // CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6532 // CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
6533 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6534 // CHECK6:       omp.body.continue:
6535 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6536 // CHECK6:       omp.inner.for.inc:
6537 // CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6538 // CHECK6-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
6539 // CHECK6-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6540 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
6541 // CHECK6:       omp.inner.for.end:
6542 // CHECK6-NEXT:    store i32 100, i32* [[I]], align 4
6543 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
6544 // CHECK6-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
6545 // CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
6546 // CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
6547 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
6548 // CHECK6:       omp.inner.for.cond7:
6549 // CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
6550 // CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !6
6551 // CHECK6-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6552 // CHECK6-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
6553 // CHECK6:       omp.inner.for.body9:
6554 // CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
6555 // CHECK6-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
6556 // CHECK6-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
6557 // CHECK6-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !6
6558 // CHECK6-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !6
6559 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
6560 // CHECK6:       omp.body.continue12:
6561 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
6562 // CHECK6:       omp.inner.for.inc13:
6563 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
6564 // CHECK6-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
6565 // CHECK6-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
6566 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP7:![0-9]+]]
6567 // CHECK6:       omp.inner.for.end15:
6568 // CHECK6-NEXT:    store i32 100, i32* [[I6]], align 4
6569 // CHECK6-NEXT:    ret void
6570 //
6571 //
6572 // CHECK6-LABEL: define {{[^@]+}}@main
6573 // CHECK6-SAME: () #[[ATTR1:[0-9]+]] {
6574 // CHECK6-NEXT:  entry:
6575 // CHECK6-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
6576 // CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6577 // CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6578 // CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6579 // CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6580 // CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
6581 // CHECK6-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
6582 // CHECK6-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
6583 // CHECK6-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
6584 // CHECK6-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
6585 // CHECK6-NEXT:    [[I6:%.*]] = alloca i32, align 4
6586 // CHECK6-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
6587 // CHECK6-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
6588 // CHECK6-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
6589 // CHECK6-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
6590 // CHECK6-NEXT:    [[I20:%.*]] = alloca i32, align 4
6591 // CHECK6-NEXT:    store i32 0, i32* [[RETVAL]], align 4
6592 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6593 // CHECK6-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6594 // CHECK6-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6595 // CHECK6-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
6596 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6597 // CHECK6:       omp.inner.for.cond:
6598 // CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
6599 // CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
6600 // CHECK6-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
6601 // CHECK6-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6602 // CHECK6:       omp.inner.for.body:
6603 // CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
6604 // CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
6605 // CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6606 // CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
6607 // CHECK6-NEXT:    call void @_Z3fn4v(), !llvm.access.group !9
6608 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6609 // CHECK6:       omp.body.continue:
6610 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6611 // CHECK6:       omp.inner.for.inc:
6612 // CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
6613 // CHECK6-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
6614 // CHECK6-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
6615 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
6616 // CHECK6:       omp.inner.for.end:
6617 // CHECK6-NEXT:    store i32 100, i32* [[I]], align 4
6618 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
6619 // CHECK6-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
6620 // CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
6621 // CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
6622 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
6623 // CHECK6:       omp.inner.for.cond7:
6624 // CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !12
6625 // CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !12
6626 // CHECK6-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6627 // CHECK6-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
6628 // CHECK6:       omp.inner.for.body9:
6629 // CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !12
6630 // CHECK6-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
6631 // CHECK6-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
6632 // CHECK6-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !12
6633 // CHECK6-NEXT:    call void @_Z3fn5v(), !llvm.access.group !12
6634 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
6635 // CHECK6:       omp.body.continue12:
6636 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
6637 // CHECK6:       omp.inner.for.inc13:
6638 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !12
6639 // CHECK6-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
6640 // CHECK6-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !12
6641 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP13:![0-9]+]]
6642 // CHECK6:       omp.inner.for.end15:
6643 // CHECK6-NEXT:    store i32 100, i32* [[I6]], align 4
6644 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
6645 // CHECK6-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
6646 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
6647 // CHECK6-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
6648 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
6649 // CHECK6:       omp.inner.for.cond21:
6650 // CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !15
6651 // CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !15
6652 // CHECK6-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
6653 // CHECK6-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
6654 // CHECK6:       omp.inner.for.body23:
6655 // CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !15
6656 // CHECK6-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
6657 // CHECK6-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
6658 // CHECK6-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !15
6659 // CHECK6-NEXT:    call void @_Z3fn6v(), !llvm.access.group !15
6660 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
6661 // CHECK6:       omp.body.continue26:
6662 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
6663 // CHECK6:       omp.inner.for.inc27:
6664 // CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !15
6665 // CHECK6-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
6666 // CHECK6-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !15
6667 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP16:![0-9]+]]
6668 // CHECK6:       omp.inner.for.end29:
6669 // CHECK6-NEXT:    store i32 100, i32* [[I20]], align 4
6670 // CHECK6-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
6671 // CHECK6-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
6672 // CHECK6-NEXT:    ret i32 [[CALL]]
6673 //
6674 //
6675 // CHECK6-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
6676 // CHECK6-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
6677 // CHECK6-NEXT:  entry:
6678 // CHECK6-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
6679 // CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6680 // CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6681 // CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6682 // CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6683 // CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
6684 // CHECK6-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
6685 // CHECK6-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
6686 // CHECK6-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
6687 // CHECK6-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
6688 // CHECK6-NEXT:    [[I6:%.*]] = alloca i32, align 4
6689 // CHECK6-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
6690 // CHECK6-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
6691 // CHECK6-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
6692 // CHECK6-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
6693 // CHECK6-NEXT:    [[I20:%.*]] = alloca i32, align 4
6694 // CHECK6-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
6695 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6696 // CHECK6-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6697 // CHECK6-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6698 // CHECK6-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
6699 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6700 // CHECK6:       omp.inner.for.cond:
6701 // CHECK6-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
6702 // CHECK6-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
6703 // CHECK6-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
6704 // CHECK6-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6705 // CHECK6:       omp.inner.for.body:
6706 // CHECK6-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
6707 // CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
6708 // CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6709 // CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !18
6710 // CHECK6-NEXT:    call void @_Z3fn1v(), !llvm.access.group !18
6711 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6712 // CHECK6:       omp.body.continue:
6713 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6714 // CHECK6:       omp.inner.for.inc:
6715 // CHECK6-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
6716 // CHECK6-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
6717 // CHECK6-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
6718 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
6719 // CHECK6:       omp.inner.for.end:
6720 // CHECK6-NEXT:    store i32 100, i32* [[I]], align 4
6721 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
6722 // CHECK6-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
6723 // CHECK6-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
6724 // CHECK6-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
6725 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
6726 // CHECK6:       omp.inner.for.cond7:
6727 // CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
6728 // CHECK6-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !21
6729 // CHECK6-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6730 // CHECK6-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
6731 // CHECK6:       omp.inner.for.body9:
6732 // CHECK6-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
6733 // CHECK6-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
6734 // CHECK6-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
6735 // CHECK6-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !21
6736 // CHECK6-NEXT:    call void @_Z3fn2v(), !llvm.access.group !21
6737 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
6738 // CHECK6:       omp.body.continue12:
6739 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
6740 // CHECK6:       omp.inner.for.inc13:
6741 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
6742 // CHECK6-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
6743 // CHECK6-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !21
6744 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP22:![0-9]+]]
6745 // CHECK6:       omp.inner.for.end15:
6746 // CHECK6-NEXT:    store i32 100, i32* [[I6]], align 4
6747 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
6748 // CHECK6-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
6749 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
6750 // CHECK6-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
6751 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
6752 // CHECK6:       omp.inner.for.cond21:
6753 // CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !24
6754 // CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !24
6755 // CHECK6-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
6756 // CHECK6-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
6757 // CHECK6:       omp.inner.for.body23:
6758 // CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !24
6759 // CHECK6-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
6760 // CHECK6-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
6761 // CHECK6-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !24
6762 // CHECK6-NEXT:    call void @_Z3fn3v(), !llvm.access.group !24
6763 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
6764 // CHECK6:       omp.body.continue26:
6765 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
6766 // CHECK6:       omp.inner.for.inc27:
6767 // CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !24
6768 // CHECK6-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
6769 // CHECK6-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !24
6770 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP25:![0-9]+]]
6771 // CHECK6:       omp.inner.for.end29:
6772 // CHECK6-NEXT:    store i32 100, i32* [[I20]], align 4
6773 // CHECK6-NEXT:    ret i32 0
6774 //
6775 //
6776 // CHECK7-LABEL: define {{[^@]+}}@_Z9gtid_testv
6777 // CHECK7-SAME: () #[[ATTR0:[0-9]+]] {
6778 // CHECK7-NEXT:  entry:
6779 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6780 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6781 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6782 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6783 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
6784 // CHECK7-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
6785 // CHECK7-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
6786 // CHECK7-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
6787 // CHECK7-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
6788 // CHECK7-NEXT:    [[I6:%.*]] = alloca i32, align 4
6789 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6790 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6791 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6792 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
6793 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6794 // CHECK7:       omp.inner.for.cond:
6795 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6796 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
6797 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
6798 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6799 // CHECK7:       omp.inner.for.body:
6800 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6801 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
6802 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6803 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
6804 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6805 // CHECK7:       omp.body.continue:
6806 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6807 // CHECK7:       omp.inner.for.inc:
6808 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6809 // CHECK7-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
6810 // CHECK7-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
6811 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
6812 // CHECK7:       omp.inner.for.end:
6813 // CHECK7-NEXT:    store i32 100, i32* [[I]], align 4
6814 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
6815 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
6816 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
6817 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
6818 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
6819 // CHECK7:       omp.inner.for.cond7:
6820 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
6821 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !6
6822 // CHECK7-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6823 // CHECK7-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
6824 // CHECK7:       omp.inner.for.body9:
6825 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
6826 // CHECK7-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
6827 // CHECK7-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
6828 // CHECK7-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !6
6829 // CHECK7-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !6
6830 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
6831 // CHECK7:       omp.body.continue12:
6832 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
6833 // CHECK7:       omp.inner.for.inc13:
6834 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
6835 // CHECK7-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
6836 // CHECK7-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
6837 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP7:![0-9]+]]
6838 // CHECK7:       omp.inner.for.end15:
6839 // CHECK7-NEXT:    store i32 100, i32* [[I6]], align 4
6840 // CHECK7-NEXT:    ret void
6841 //
6842 //
6843 // CHECK7-LABEL: define {{[^@]+}}@main
6844 // CHECK7-SAME: () #[[ATTR1:[0-9]+]] {
6845 // CHECK7-NEXT:  entry:
6846 // CHECK7-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
6847 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6848 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6849 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6850 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6851 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
6852 // CHECK7-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
6853 // CHECK7-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
6854 // CHECK7-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
6855 // CHECK7-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
6856 // CHECK7-NEXT:    [[I6:%.*]] = alloca i32, align 4
6857 // CHECK7-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
6858 // CHECK7-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
6859 // CHECK7-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
6860 // CHECK7-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
6861 // CHECK7-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
6862 // CHECK7-NEXT:    [[I20:%.*]] = alloca i32, align 4
6863 // CHECK7-NEXT:    store i32 0, i32* [[RETVAL]], align 4
6864 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6865 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
6866 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6867 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
6868 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6869 // CHECK7:       omp.inner.for.cond:
6870 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
6871 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
6872 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
6873 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6874 // CHECK7:       omp.inner.for.body:
6875 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
6876 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
6877 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
6878 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
6879 // CHECK7-NEXT:    call void @_Z3fn4v(), !llvm.access.group !9
6880 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6881 // CHECK7:       omp.body.continue:
6882 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6883 // CHECK7:       omp.inner.for.inc:
6884 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
6885 // CHECK7-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
6886 // CHECK7-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
6887 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
6888 // CHECK7:       omp.inner.for.end:
6889 // CHECK7-NEXT:    store i32 100, i32* [[I]], align 4
6890 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
6891 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
6892 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
6893 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
6894 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
6895 // CHECK7:       omp.inner.for.cond7:
6896 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
6897 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4
6898 // CHECK7-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
6899 // CHECK7-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
6900 // CHECK7:       omp.inner.for.body9:
6901 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
6902 // CHECK7-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
6903 // CHECK7-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
6904 // CHECK7-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4
6905 // CHECK7-NEXT:    call void @_Z3fn5v()
6906 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
6907 // CHECK7:       omp.body.continue12:
6908 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
6909 // CHECK7:       omp.inner.for.inc13:
6910 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
6911 // CHECK7-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
6912 // CHECK7-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4
6913 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP12:![0-9]+]]
6914 // CHECK7:       omp.inner.for.end15:
6915 // CHECK7-NEXT:    store i32 100, i32* [[I6]], align 4
6916 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* @Arg, align 4
6917 // CHECK7-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP10]], 0
6918 // CHECK7-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
6919 // CHECK7-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
6920 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
6921 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
6922 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
6923 // CHECK7-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV19]], align 4
6924 // CHECK7-NEXT:    [[TMP12:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
6925 // CHECK7-NEXT:    [[TOBOOL21:%.*]] = trunc i8 [[TMP12]] to i1
6926 // CHECK7-NEXT:    br i1 [[TOBOOL21]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
6927 // CHECK7:       omp_if.then:
6928 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND22:%.*]]
6929 // CHECK7:       omp.inner.for.cond22:
6930 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !14
6931 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !14
6932 // CHECK7-NEXT:    [[CMP23:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
6933 // CHECK7-NEXT:    br i1 [[CMP23]], label [[OMP_INNER_FOR_BODY24:%.*]], label [[OMP_INNER_FOR_END30:%.*]]
6934 // CHECK7:       omp.inner.for.body24:
6935 // CHECK7-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !14
6936 // CHECK7-NEXT:    [[MUL25:%.*]] = mul nsw i32 [[TMP15]], 1
6937 // CHECK7-NEXT:    [[ADD26:%.*]] = add nsw i32 0, [[MUL25]]
6938 // CHECK7-NEXT:    store i32 [[ADD26]], i32* [[I20]], align 4, !llvm.access.group !14
6939 // CHECK7-NEXT:    call void @_Z3fn6v(), !llvm.access.group !14
6940 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE27:%.*]]
6941 // CHECK7:       omp.body.continue27:
6942 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC28:%.*]]
6943 // CHECK7:       omp.inner.for.inc28:
6944 // CHECK7-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !14
6945 // CHECK7-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP16]], 1
6946 // CHECK7-NEXT:    store i32 [[ADD29]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !14
6947 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND22]], !llvm.loop [[LOOP15:![0-9]+]]
6948 // CHECK7:       omp.inner.for.end30:
6949 // CHECK7-NEXT:    br label [[OMP_IF_END:%.*]]
6950 // CHECK7:       omp_if.else:
6951 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND31:%.*]]
6952 // CHECK7:       omp.inner.for.cond31:
6953 // CHECK7-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
6954 // CHECK7-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4
6955 // CHECK7-NEXT:    [[CMP32:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
6956 // CHECK7-NEXT:    br i1 [[CMP32]], label [[OMP_INNER_FOR_BODY33:%.*]], label [[OMP_INNER_FOR_END39:%.*]]
6957 // CHECK7:       omp.inner.for.body33:
6958 // CHECK7-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
6959 // CHECK7-NEXT:    [[MUL34:%.*]] = mul nsw i32 [[TMP19]], 1
6960 // CHECK7-NEXT:    [[ADD35:%.*]] = add nsw i32 0, [[MUL34]]
6961 // CHECK7-NEXT:    store i32 [[ADD35]], i32* [[I20]], align 4
6962 // CHECK7-NEXT:    call void @_Z3fn6v()
6963 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE36:%.*]]
6964 // CHECK7:       omp.body.continue36:
6965 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC37:%.*]]
6966 // CHECK7:       omp.inner.for.inc37:
6967 // CHECK7-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
6968 // CHECK7-NEXT:    [[ADD38:%.*]] = add nsw i32 [[TMP20]], 1
6969 // CHECK7-NEXT:    store i32 [[ADD38]], i32* [[DOTOMP_IV19]], align 4
6970 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND31]], !llvm.loop [[LOOP17:![0-9]+]]
6971 // CHECK7:       omp.inner.for.end39:
6972 // CHECK7-NEXT:    br label [[OMP_IF_END]]
6973 // CHECK7:       omp_if.end:
6974 // CHECK7-NEXT:    store i32 100, i32* [[I20]], align 4
6975 // CHECK7-NEXT:    [[TMP21:%.*]] = load i32, i32* @Arg, align 4
6976 // CHECK7-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP21]])
6977 // CHECK7-NEXT:    ret i32 [[CALL]]
6978 //
6979 //
6980 // CHECK7-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
6981 // CHECK7-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
6982 // CHECK7-NEXT:  entry:
6983 // CHECK7-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
6984 // CHECK7-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6985 // CHECK7-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6986 // CHECK7-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6987 // CHECK7-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6988 // CHECK7-NEXT:    [[I:%.*]] = alloca i32, align 4
6989 // CHECK7-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
6990 // CHECK7-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
6991 // CHECK7-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
6992 // CHECK7-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
6993 // CHECK7-NEXT:    [[I6:%.*]] = alloca i32, align 4
6994 // CHECK7-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
6995 // CHECK7-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
6996 // CHECK7-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
6997 // CHECK7-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
6998 // CHECK7-NEXT:    [[I20:%.*]] = alloca i32, align 4
6999 // CHECK7-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
7000 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7001 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
7002 // CHECK7-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7003 // CHECK7-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
7004 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7005 // CHECK7:       omp.inner.for.cond:
7006 // CHECK7-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
7007 // CHECK7-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
7008 // CHECK7-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
7009 // CHECK7-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7010 // CHECK7:       omp.inner.for.body:
7011 // CHECK7-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
7012 // CHECK7-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
7013 // CHECK7-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7014 // CHECK7-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !18
7015 // CHECK7-NEXT:    call void @_Z3fn1v(), !llvm.access.group !18
7016 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7017 // CHECK7:       omp.body.continue:
7018 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7019 // CHECK7:       omp.inner.for.inc:
7020 // CHECK7-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
7021 // CHECK7-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
7022 // CHECK7-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
7023 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
7024 // CHECK7:       omp.inner.for.end:
7025 // CHECK7-NEXT:    store i32 100, i32* [[I]], align 4
7026 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
7027 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
7028 // CHECK7-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
7029 // CHECK7-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
7030 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
7031 // CHECK7:       omp.inner.for.cond7:
7032 // CHECK7-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7033 // CHECK7-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4
7034 // CHECK7-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7035 // CHECK7-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
7036 // CHECK7:       omp.inner.for.body9:
7037 // CHECK7-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7038 // CHECK7-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
7039 // CHECK7-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
7040 // CHECK7-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4
7041 // CHECK7-NEXT:    call void @_Z3fn2v()
7042 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
7043 // CHECK7:       omp.body.continue12:
7044 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
7045 // CHECK7:       omp.inner.for.inc13:
7046 // CHECK7-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7047 // CHECK7-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
7048 // CHECK7-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4
7049 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP21:![0-9]+]]
7050 // CHECK7:       omp.inner.for.end15:
7051 // CHECK7-NEXT:    store i32 100, i32* [[I6]], align 4
7052 // CHECK7-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
7053 // CHECK7-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
7054 // CHECK7-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
7055 // CHECK7-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
7056 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
7057 // CHECK7:       omp.inner.for.cond21:
7058 // CHECK7-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !22
7059 // CHECK7-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !22
7060 // CHECK7-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
7061 // CHECK7-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
7062 // CHECK7:       omp.inner.for.body23:
7063 // CHECK7-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !22
7064 // CHECK7-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
7065 // CHECK7-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
7066 // CHECK7-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !22
7067 // CHECK7-NEXT:    call void @_Z3fn3v(), !llvm.access.group !22
7068 // CHECK7-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
7069 // CHECK7:       omp.body.continue26:
7070 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
7071 // CHECK7:       omp.inner.for.inc27:
7072 // CHECK7-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !22
7073 // CHECK7-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
7074 // CHECK7-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !22
7075 // CHECK7-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP23:![0-9]+]]
7076 // CHECK7:       omp.inner.for.end29:
7077 // CHECK7-NEXT:    store i32 100, i32* [[I20]], align 4
7078 // CHECK7-NEXT:    ret i32 0
7079 //
7080 //
7081 // CHECK8-LABEL: define {{[^@]+}}@_Z9gtid_testv
7082 // CHECK8-SAME: () #[[ATTR0:[0-9]+]] {
7083 // CHECK8-NEXT:  entry:
7084 // CHECK8-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7085 // CHECK8-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7086 // CHECK8-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7087 // CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7088 // CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
7089 // CHECK8-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
7090 // CHECK8-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
7091 // CHECK8-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
7092 // CHECK8-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
7093 // CHECK8-NEXT:    [[I6:%.*]] = alloca i32, align 4
7094 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7095 // CHECK8-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
7096 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7097 // CHECK8-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
7098 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7099 // CHECK8:       omp.inner.for.cond:
7100 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
7101 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !2
7102 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
7103 // CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7104 // CHECK8:       omp.inner.for.body:
7105 // CHECK8-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
7106 // CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
7107 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7108 // CHECK8-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !2
7109 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7110 // CHECK8:       omp.body.continue:
7111 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7112 // CHECK8:       omp.inner.for.inc:
7113 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
7114 // CHECK8-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
7115 // CHECK8-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !2
7116 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
7117 // CHECK8:       omp.inner.for.end:
7118 // CHECK8-NEXT:    store i32 100, i32* [[I]], align 4
7119 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
7120 // CHECK8-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
7121 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
7122 // CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
7123 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
7124 // CHECK8:       omp.inner.for.cond7:
7125 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
7126 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !6
7127 // CHECK8-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7128 // CHECK8-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
7129 // CHECK8:       omp.inner.for.body9:
7130 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
7131 // CHECK8-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
7132 // CHECK8-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
7133 // CHECK8-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !6
7134 // CHECK8-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !6
7135 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
7136 // CHECK8:       omp.body.continue12:
7137 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
7138 // CHECK8:       omp.inner.for.inc13:
7139 // CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
7140 // CHECK8-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
7141 // CHECK8-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !6
7142 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP7:![0-9]+]]
7143 // CHECK8:       omp.inner.for.end15:
7144 // CHECK8-NEXT:    store i32 100, i32* [[I6]], align 4
7145 // CHECK8-NEXT:    ret void
7146 //
7147 //
7148 // CHECK8-LABEL: define {{[^@]+}}@main
7149 // CHECK8-SAME: () #[[ATTR1:[0-9]+]] {
7150 // CHECK8-NEXT:  entry:
7151 // CHECK8-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
7152 // CHECK8-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7153 // CHECK8-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7154 // CHECK8-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7155 // CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7156 // CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
7157 // CHECK8-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
7158 // CHECK8-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
7159 // CHECK8-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
7160 // CHECK8-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
7161 // CHECK8-NEXT:    [[I6:%.*]] = alloca i32, align 4
7162 // CHECK8-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
7163 // CHECK8-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
7164 // CHECK8-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
7165 // CHECK8-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
7166 // CHECK8-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
7167 // CHECK8-NEXT:    [[I20:%.*]] = alloca i32, align 4
7168 // CHECK8-NEXT:    store i32 0, i32* [[RETVAL]], align 4
7169 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7170 // CHECK8-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
7171 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7172 // CHECK8-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
7173 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7174 // CHECK8:       omp.inner.for.cond:
7175 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
7176 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !9
7177 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
7178 // CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7179 // CHECK8:       omp.inner.for.body:
7180 // CHECK8-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
7181 // CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
7182 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7183 // CHECK8-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !9
7184 // CHECK8-NEXT:    call void @_Z3fn4v(), !llvm.access.group !9
7185 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7186 // CHECK8:       omp.body.continue:
7187 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7188 // CHECK8:       omp.inner.for.inc:
7189 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
7190 // CHECK8-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
7191 // CHECK8-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !9
7192 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP10:![0-9]+]]
7193 // CHECK8:       omp.inner.for.end:
7194 // CHECK8-NEXT:    store i32 100, i32* [[I]], align 4
7195 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
7196 // CHECK8-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
7197 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
7198 // CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
7199 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
7200 // CHECK8:       omp.inner.for.cond7:
7201 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7202 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4
7203 // CHECK8-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7204 // CHECK8-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
7205 // CHECK8:       omp.inner.for.body9:
7206 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7207 // CHECK8-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
7208 // CHECK8-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
7209 // CHECK8-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4
7210 // CHECK8-NEXT:    call void @_Z3fn5v()
7211 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
7212 // CHECK8:       omp.body.continue12:
7213 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
7214 // CHECK8:       omp.inner.for.inc13:
7215 // CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7216 // CHECK8-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
7217 // CHECK8-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4
7218 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP12:![0-9]+]]
7219 // CHECK8:       omp.inner.for.end15:
7220 // CHECK8-NEXT:    store i32 100, i32* [[I6]], align 4
7221 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* @Arg, align 4
7222 // CHECK8-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP10]], 0
7223 // CHECK8-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
7224 // CHECK8-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
7225 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
7226 // CHECK8-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
7227 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
7228 // CHECK8-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV19]], align 4
7229 // CHECK8-NEXT:    [[TMP12:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
7230 // CHECK8-NEXT:    [[TOBOOL21:%.*]] = trunc i8 [[TMP12]] to i1
7231 // CHECK8-NEXT:    br i1 [[TOBOOL21]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
7232 // CHECK8:       omp_if.then:
7233 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND22:%.*]]
7234 // CHECK8:       omp.inner.for.cond22:
7235 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !14
7236 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !14
7237 // CHECK8-NEXT:    [[CMP23:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
7238 // CHECK8-NEXT:    br i1 [[CMP23]], label [[OMP_INNER_FOR_BODY24:%.*]], label [[OMP_INNER_FOR_END30:%.*]]
7239 // CHECK8:       omp.inner.for.body24:
7240 // CHECK8-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !14
7241 // CHECK8-NEXT:    [[MUL25:%.*]] = mul nsw i32 [[TMP15]], 1
7242 // CHECK8-NEXT:    [[ADD26:%.*]] = add nsw i32 0, [[MUL25]]
7243 // CHECK8-NEXT:    store i32 [[ADD26]], i32* [[I20]], align 4, !llvm.access.group !14
7244 // CHECK8-NEXT:    call void @_Z3fn6v(), !llvm.access.group !14
7245 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE27:%.*]]
7246 // CHECK8:       omp.body.continue27:
7247 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC28:%.*]]
7248 // CHECK8:       omp.inner.for.inc28:
7249 // CHECK8-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !14
7250 // CHECK8-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP16]], 1
7251 // CHECK8-NEXT:    store i32 [[ADD29]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !14
7252 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND22]], !llvm.loop [[LOOP15:![0-9]+]]
7253 // CHECK8:       omp.inner.for.end30:
7254 // CHECK8-NEXT:    br label [[OMP_IF_END:%.*]]
7255 // CHECK8:       omp_if.else:
7256 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND31:%.*]]
7257 // CHECK8:       omp.inner.for.cond31:
7258 // CHECK8-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
7259 // CHECK8-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4
7260 // CHECK8-NEXT:    [[CMP32:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
7261 // CHECK8-NEXT:    br i1 [[CMP32]], label [[OMP_INNER_FOR_BODY33:%.*]], label [[OMP_INNER_FOR_END39:%.*]]
7262 // CHECK8:       omp.inner.for.body33:
7263 // CHECK8-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
7264 // CHECK8-NEXT:    [[MUL34:%.*]] = mul nsw i32 [[TMP19]], 1
7265 // CHECK8-NEXT:    [[ADD35:%.*]] = add nsw i32 0, [[MUL34]]
7266 // CHECK8-NEXT:    store i32 [[ADD35]], i32* [[I20]], align 4
7267 // CHECK8-NEXT:    call void @_Z3fn6v()
7268 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE36:%.*]]
7269 // CHECK8:       omp.body.continue36:
7270 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC37:%.*]]
7271 // CHECK8:       omp.inner.for.inc37:
7272 // CHECK8-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
7273 // CHECK8-NEXT:    [[ADD38:%.*]] = add nsw i32 [[TMP20]], 1
7274 // CHECK8-NEXT:    store i32 [[ADD38]], i32* [[DOTOMP_IV19]], align 4
7275 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND31]], !llvm.loop [[LOOP17:![0-9]+]]
7276 // CHECK8:       omp.inner.for.end39:
7277 // CHECK8-NEXT:    br label [[OMP_IF_END]]
7278 // CHECK8:       omp_if.end:
7279 // CHECK8-NEXT:    store i32 100, i32* [[I20]], align 4
7280 // CHECK8-NEXT:    [[TMP21:%.*]] = load i32, i32* @Arg, align 4
7281 // CHECK8-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP21]])
7282 // CHECK8-NEXT:    ret i32 [[CALL]]
7283 //
7284 //
7285 // CHECK8-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
7286 // CHECK8-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
7287 // CHECK8-NEXT:  entry:
7288 // CHECK8-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
7289 // CHECK8-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7290 // CHECK8-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7291 // CHECK8-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7292 // CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7293 // CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
7294 // CHECK8-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
7295 // CHECK8-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
7296 // CHECK8-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
7297 // CHECK8-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
7298 // CHECK8-NEXT:    [[I6:%.*]] = alloca i32, align 4
7299 // CHECK8-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
7300 // CHECK8-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
7301 // CHECK8-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
7302 // CHECK8-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
7303 // CHECK8-NEXT:    [[I20:%.*]] = alloca i32, align 4
7304 // CHECK8-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
7305 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7306 // CHECK8-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
7307 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7308 // CHECK8-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
7309 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7310 // CHECK8:       omp.inner.for.cond:
7311 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
7312 // CHECK8-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !18
7313 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
7314 // CHECK8-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7315 // CHECK8:       omp.inner.for.body:
7316 // CHECK8-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
7317 // CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
7318 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7319 // CHECK8-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !18
7320 // CHECK8-NEXT:    call void @_Z3fn1v(), !llvm.access.group !18
7321 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7322 // CHECK8:       omp.body.continue:
7323 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7324 // CHECK8:       omp.inner.for.inc:
7325 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
7326 // CHECK8-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
7327 // CHECK8-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !18
7328 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP19:![0-9]+]]
7329 // CHECK8:       omp.inner.for.end:
7330 // CHECK8-NEXT:    store i32 100, i32* [[I]], align 4
7331 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
7332 // CHECK8-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
7333 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
7334 // CHECK8-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
7335 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
7336 // CHECK8:       omp.inner.for.cond7:
7337 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7338 // CHECK8-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4
7339 // CHECK8-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
7340 // CHECK8-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
7341 // CHECK8:       omp.inner.for.body9:
7342 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7343 // CHECK8-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
7344 // CHECK8-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
7345 // CHECK8-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4
7346 // CHECK8-NEXT:    call void @_Z3fn2v()
7347 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
7348 // CHECK8:       omp.body.continue12:
7349 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
7350 // CHECK8:       omp.inner.for.inc13:
7351 // CHECK8-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
7352 // CHECK8-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
7353 // CHECK8-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4
7354 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP21:![0-9]+]]
7355 // CHECK8:       omp.inner.for.end15:
7356 // CHECK8-NEXT:    store i32 100, i32* [[I6]], align 4
7357 // CHECK8-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
7358 // CHECK8-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
7359 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
7360 // CHECK8-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
7361 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
7362 // CHECK8:       omp.inner.for.cond21:
7363 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !22
7364 // CHECK8-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !22
7365 // CHECK8-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
7366 // CHECK8-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
7367 // CHECK8:       omp.inner.for.body23:
7368 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !22
7369 // CHECK8-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
7370 // CHECK8-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
7371 // CHECK8-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !22
7372 // CHECK8-NEXT:    call void @_Z3fn3v(), !llvm.access.group !22
7373 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
7374 // CHECK8:       omp.body.continue26:
7375 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
7376 // CHECK8:       omp.inner.for.inc27:
7377 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !22
7378 // CHECK8-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
7379 // CHECK8-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !22
7380 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP23:![0-9]+]]
7381 // CHECK8:       omp.inner.for.end29:
7382 // CHECK8-NEXT:    store i32 100, i32* [[I20]], align 4
7383 // CHECK8-NEXT:    ret i32 0
7384 //
7385 //
7386 // CHECK9-LABEL: define {{[^@]+}}@_Z9gtid_testv
7387 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
7388 // CHECK9-NEXT:  entry:
7389 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7390 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
7391 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
7392 // 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)
7393 // CHECK9-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
7394 // CHECK9-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
7395 // CHECK9:       omp_offload.failed:
7396 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43() #[[ATTR2:[0-9]+]]
7397 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
7398 // CHECK9:       omp_offload.cont:
7399 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
7400 // 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)
7401 // CHECK9-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
7402 // CHECK9-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
7403 // CHECK9:       omp_offload.failed2:
7404 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48() #[[ATTR2]]
7405 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
7406 // CHECK9:       omp_offload.cont3:
7407 // CHECK9-NEXT:    ret void
7408 //
7409 //
7410 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43
7411 // CHECK9-SAME: () #[[ATTR1:[0-9]+]] {
7412 // CHECK9-NEXT:  entry:
7413 // 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*, ...)*))
7414 // CHECK9-NEXT:    ret void
7415 //
7416 //
7417 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
7418 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
7419 // CHECK9-NEXT:  entry:
7420 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7421 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7422 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7423 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7424 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7425 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7426 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7427 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7428 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7429 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7430 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7431 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7432 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
7433 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7434 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7435 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7436 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7437 // 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)
7438 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7439 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
7440 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7441 // CHECK9:       cond.true:
7442 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7443 // CHECK9:       cond.false:
7444 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7445 // CHECK9-NEXT:    br label [[COND_END]]
7446 // CHECK9:       cond.end:
7447 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7448 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7449 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7450 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7451 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7452 // CHECK9:       omp.inner.for.cond:
7453 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
7454 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
7455 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7456 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7457 // CHECK9:       omp.inner.for.body:
7458 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !15
7459 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
7460 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
7461 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
7462 // 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
7463 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7464 // CHECK9:       omp.inner.for.inc:
7465 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
7466 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !15
7467 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
7468 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
7469 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
7470 // CHECK9:       omp.inner.for.end:
7471 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7472 // CHECK9:       omp.loop.exit:
7473 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7474 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7475 // CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
7476 // CHECK9-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7477 // CHECK9:       .omp.final.then:
7478 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
7479 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
7480 // CHECK9:       .omp.final.done:
7481 // CHECK9-NEXT:    ret void
7482 //
7483 //
7484 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
7485 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
7486 // CHECK9-NEXT:  entry:
7487 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7488 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7489 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
7490 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
7491 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7492 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7493 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7494 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7495 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7496 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7497 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7498 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7499 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7500 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7501 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7502 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7503 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
7504 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7505 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
7506 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7507 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
7508 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
7509 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
7510 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7511 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7512 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7513 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
7514 // 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)
7515 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7516 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
7517 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7518 // CHECK9:       cond.true:
7519 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7520 // CHECK9:       cond.false:
7521 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7522 // CHECK9-NEXT:    br label [[COND_END]]
7523 // CHECK9:       cond.end:
7524 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7525 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7526 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7527 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
7528 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7529 // CHECK9:       omp.inner.for.cond:
7530 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
7531 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
7532 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7533 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7534 // CHECK9:       omp.inner.for.body:
7535 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
7536 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7537 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7538 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !19
7539 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7540 // CHECK9:       omp.body.continue:
7541 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7542 // CHECK9:       omp.inner.for.inc:
7543 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
7544 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
7545 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
7546 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
7547 // CHECK9:       omp.inner.for.end:
7548 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7549 // CHECK9:       omp.loop.exit:
7550 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
7551 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7552 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
7553 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7554 // CHECK9:       .omp.final.then:
7555 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
7556 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
7557 // CHECK9:       .omp.final.done:
7558 // CHECK9-NEXT:    ret void
7559 //
7560 //
7561 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48
7562 // CHECK9-SAME: () #[[ATTR1]] {
7563 // CHECK9-NEXT:  entry:
7564 // 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*, ...)*))
7565 // CHECK9-NEXT:    ret void
7566 //
7567 //
7568 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..2
7569 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
7570 // CHECK9-NEXT:  entry:
7571 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7572 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7573 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7574 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7575 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7576 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7577 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7578 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7579 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7580 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
7581 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7582 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7583 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7584 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
7585 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7586 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7587 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7588 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7589 // 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)
7590 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7591 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
7592 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7593 // CHECK9:       cond.true:
7594 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7595 // CHECK9:       cond.false:
7596 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7597 // CHECK9-NEXT:    br label [[COND_END]]
7598 // CHECK9:       cond.end:
7599 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7600 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7601 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7602 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7603 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7604 // CHECK9:       omp.inner.for.cond:
7605 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
7606 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
7607 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7608 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7609 // CHECK9:       omp.inner.for.body:
7610 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
7611 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
7612 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
7613 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
7614 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !24
7615 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !24
7616 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !24
7617 // CHECK9-NEXT:    call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !24
7618 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !24
7619 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7620 // CHECK9:       omp.inner.for.inc:
7621 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
7622 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
7623 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
7624 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
7625 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
7626 // CHECK9:       omp.inner.for.end:
7627 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7628 // CHECK9:       omp.loop.exit:
7629 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7630 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7631 // CHECK9-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
7632 // CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7633 // CHECK9:       .omp.final.then:
7634 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
7635 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
7636 // CHECK9:       .omp.final.done:
7637 // CHECK9-NEXT:    ret void
7638 //
7639 //
7640 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..3
7641 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
7642 // CHECK9-NEXT:  entry:
7643 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7644 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7645 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
7646 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
7647 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7648 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7649 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7650 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7651 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7652 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7653 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7654 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7655 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7656 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7657 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7658 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7659 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
7660 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7661 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
7662 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7663 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
7664 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
7665 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
7666 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7667 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7668 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7669 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
7670 // 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)
7671 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7672 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
7673 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7674 // CHECK9:       cond.true:
7675 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7676 // CHECK9:       cond.false:
7677 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7678 // CHECK9-NEXT:    br label [[COND_END]]
7679 // CHECK9:       cond.end:
7680 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7681 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7682 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7683 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
7684 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7685 // CHECK9:       omp.inner.for.cond:
7686 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
7687 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
7688 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7689 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7690 // CHECK9:       omp.inner.for.body:
7691 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
7692 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7693 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7694 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !27
7695 // CHECK9-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !27
7696 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7697 // CHECK9:       omp.body.continue:
7698 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7699 // CHECK9:       omp.inner.for.inc:
7700 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
7701 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
7702 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
7703 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
7704 // CHECK9:       omp.inner.for.end:
7705 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7706 // CHECK9:       omp.loop.exit:
7707 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
7708 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7709 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
7710 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7711 // CHECK9:       .omp.final.then:
7712 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
7713 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
7714 // CHECK9:       .omp.final.done:
7715 // CHECK9-NEXT:    ret void
7716 //
7717 //
7718 // CHECK9-LABEL: define {{[^@]+}}@main
7719 // CHECK9-SAME: () #[[ATTR3:[0-9]+]] {
7720 // CHECK9-NEXT:  entry:
7721 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
7722 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7723 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
7724 // CHECK9-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
7725 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
7726 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
7727 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
7728 // CHECK9-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
7729 // CHECK9-NEXT:    store i32 0, i32* [[RETVAL]], align 4
7730 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
7731 // 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)
7732 // CHECK9-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
7733 // CHECK9-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
7734 // CHECK9:       omp_offload.failed:
7735 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81() #[[ATTR2]]
7736 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
7737 // CHECK9:       omp_offload.cont:
7738 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
7739 // 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)
7740 // CHECK9-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
7741 // CHECK9-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
7742 // CHECK9:       omp_offload.failed2:
7743 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90() #[[ATTR2]]
7744 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
7745 // CHECK9:       omp_offload.cont3:
7746 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* @Arg, align 4
7747 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
7748 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
7749 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
7750 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7751 // CHECK9-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
7752 // CHECK9-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
7753 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7754 // CHECK9-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
7755 // CHECK9-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
7756 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
7757 // CHECK9-NEXT:    store i8* null, i8** [[TMP10]], align 8
7758 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
7759 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
7760 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
7761 // 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)
7762 // CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
7763 // CHECK9-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
7764 // CHECK9:       omp_offload.failed5:
7765 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99(i64 [[TMP5]]) #[[ATTR2]]
7766 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
7767 // CHECK9:       omp_offload.cont6:
7768 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
7769 // CHECK9-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
7770 // CHECK9-NEXT:    ret i32 [[CALL]]
7771 //
7772 //
7773 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
7774 // CHECK9-SAME: () #[[ATTR1]] {
7775 // CHECK9-NEXT:  entry:
7776 // 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*, ...)*))
7777 // CHECK9-NEXT:    ret void
7778 //
7779 //
7780 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..4
7781 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
7782 // CHECK9-NEXT:  entry:
7783 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7784 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7785 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7786 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7787 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7788 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7789 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7790 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7791 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7792 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7793 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7794 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7795 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
7796 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7797 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7798 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7799 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7800 // 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)
7801 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7802 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
7803 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7804 // CHECK9:       cond.true:
7805 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7806 // CHECK9:       cond.false:
7807 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7808 // CHECK9-NEXT:    br label [[COND_END]]
7809 // CHECK9:       cond.end:
7810 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7811 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7812 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7813 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7814 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7815 // CHECK9:       omp.inner.for.cond:
7816 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
7817 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
7818 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7819 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7820 // CHECK9:       omp.inner.for.body:
7821 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
7822 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
7823 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
7824 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
7825 // 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
7826 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7827 // CHECK9:       omp.inner.for.inc:
7828 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
7829 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
7830 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
7831 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
7832 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
7833 // CHECK9:       omp.inner.for.end:
7834 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7835 // CHECK9:       omp.loop.exit:
7836 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7837 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7838 // CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
7839 // CHECK9-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7840 // CHECK9:       .omp.final.then:
7841 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
7842 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
7843 // CHECK9:       .omp.final.done:
7844 // CHECK9-NEXT:    ret void
7845 //
7846 //
7847 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..5
7848 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
7849 // CHECK9-NEXT:  entry:
7850 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7851 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7852 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
7853 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
7854 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7855 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7856 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7857 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7858 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7859 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7860 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7861 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7862 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7863 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7864 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7865 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7866 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
7867 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
7868 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
7869 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
7870 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
7871 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
7872 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
7873 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7874 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7875 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7876 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
7877 // 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)
7878 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7879 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
7880 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7881 // CHECK9:       cond.true:
7882 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7883 // CHECK9:       cond.false:
7884 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7885 // CHECK9-NEXT:    br label [[COND_END]]
7886 // CHECK9:       cond.end:
7887 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
7888 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7889 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7890 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
7891 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7892 // CHECK9:       omp.inner.for.cond:
7893 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
7894 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
7895 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
7896 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7897 // CHECK9:       omp.inner.for.body:
7898 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
7899 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
7900 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7901 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
7902 // CHECK9-NEXT:    call void @_Z3fn4v(), !llvm.access.group !33
7903 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7904 // CHECK9:       omp.body.continue:
7905 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7906 // CHECK9:       omp.inner.for.inc:
7907 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
7908 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
7909 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
7910 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
7911 // CHECK9:       omp.inner.for.end:
7912 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7913 // CHECK9:       omp.loop.exit:
7914 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
7915 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7916 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
7917 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7918 // CHECK9:       .omp.final.then:
7919 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
7920 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
7921 // CHECK9:       .omp.final.done:
7922 // CHECK9-NEXT:    ret void
7923 //
7924 //
7925 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90
7926 // CHECK9-SAME: () #[[ATTR1]] {
7927 // CHECK9-NEXT:  entry:
7928 // 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*, ...)*))
7929 // CHECK9-NEXT:    ret void
7930 //
7931 //
7932 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..6
7933 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
7934 // CHECK9-NEXT:  entry:
7935 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7936 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7937 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7938 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7939 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
7940 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
7941 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7942 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7943 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
7944 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
7945 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7946 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7947 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
7948 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
7949 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7950 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7951 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7952 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
7953 // 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)
7954 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7955 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
7956 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7957 // CHECK9:       cond.true:
7958 // CHECK9-NEXT:    br label [[COND_END:%.*]]
7959 // CHECK9:       cond.false:
7960 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
7961 // CHECK9-NEXT:    br label [[COND_END]]
7962 // CHECK9:       cond.end:
7963 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
7964 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
7965 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
7966 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
7967 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7968 // CHECK9:       omp.inner.for.cond:
7969 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
7970 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
7971 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
7972 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7973 // CHECK9:       omp.inner.for.body:
7974 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !36
7975 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
7976 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
7977 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
7978 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !36
7979 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !36
7980 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !36
7981 // CHECK9-NEXT:    call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !36
7982 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !36
7983 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7984 // CHECK9:       omp.inner.for.inc:
7985 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
7986 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !36
7987 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
7988 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
7989 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
7990 // CHECK9:       omp.inner.for.end:
7991 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7992 // CHECK9:       omp.loop.exit:
7993 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
7994 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
7995 // CHECK9-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
7996 // CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
7997 // CHECK9:       .omp.final.then:
7998 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
7999 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8000 // CHECK9:       .omp.final.done:
8001 // CHECK9-NEXT:    ret void
8002 //
8003 //
8004 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..7
8005 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
8006 // CHECK9-NEXT:  entry:
8007 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8008 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8009 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8010 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8011 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8012 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8013 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8014 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8015 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8016 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8017 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8018 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8019 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8020 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8021 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8022 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8023 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
8024 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8025 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8026 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8027 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8028 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8029 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8030 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8031 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8032 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8033 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8034 // 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)
8035 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8036 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8037 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8038 // CHECK9:       cond.true:
8039 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8040 // CHECK9:       cond.false:
8041 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8042 // CHECK9-NEXT:    br label [[COND_END]]
8043 // CHECK9:       cond.end:
8044 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8045 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8046 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8047 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
8048 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8049 // CHECK9:       omp.inner.for.cond:
8050 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
8051 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
8052 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
8053 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8054 // CHECK9:       omp.inner.for.body:
8055 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
8056 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
8057 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8058 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
8059 // CHECK9-NEXT:    call void @_Z3fn5v(), !llvm.access.group !39
8060 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8061 // CHECK9:       omp.body.continue:
8062 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8063 // CHECK9:       omp.inner.for.inc:
8064 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
8065 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
8066 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
8067 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
8068 // CHECK9:       omp.inner.for.end:
8069 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8070 // CHECK9:       omp.loop.exit:
8071 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8072 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8073 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
8074 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8075 // CHECK9:       .omp.final.then:
8076 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
8077 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8078 // CHECK9:       .omp.final.done:
8079 // CHECK9-NEXT:    ret void
8080 //
8081 //
8082 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99
8083 // CHECK9-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
8084 // CHECK9-NEXT:  entry:
8085 // CHECK9-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
8086 // CHECK9-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
8087 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
8088 // 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]])
8089 // CHECK9-NEXT:    ret void
8090 //
8091 //
8092 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..8
8093 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
8094 // CHECK9-NEXT:  entry:
8095 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8096 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8097 // CHECK9-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
8098 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8099 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8100 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8101 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8102 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8103 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8104 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8105 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
8106 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8107 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8108 // CHECK9-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
8109 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
8110 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8111 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
8112 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8113 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8114 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8115 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
8116 // 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)
8117 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8118 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
8119 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8120 // CHECK9:       cond.true:
8121 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8122 // CHECK9:       cond.false:
8123 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8124 // CHECK9-NEXT:    br label [[COND_END]]
8125 // CHECK9:       cond.end:
8126 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
8127 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8128 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8129 // CHECK9-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
8130 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8131 // CHECK9:       omp.inner.for.cond:
8132 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
8133 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !42
8134 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
8135 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8136 // CHECK9:       omp.inner.for.body:
8137 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !42
8138 // CHECK9-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
8139 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !42
8140 // CHECK9-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
8141 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !42
8142 // CHECK9-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
8143 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
8144 // CHECK9:       omp_if.then:
8145 // 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
8146 // CHECK9-NEXT:    br label [[OMP_IF_END:%.*]]
8147 // CHECK9:       omp_if.else:
8148 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !42
8149 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !42
8150 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !42
8151 // CHECK9-NEXT:    call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !42
8152 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !42
8153 // CHECK9-NEXT:    br label [[OMP_IF_END]]
8154 // CHECK9:       omp_if.end:
8155 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8156 // CHECK9:       omp.inner.for.inc:
8157 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
8158 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !42
8159 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
8160 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
8161 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP43:![0-9]+]]
8162 // CHECK9:       omp.inner.for.end:
8163 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8164 // CHECK9:       omp.loop.exit:
8165 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
8166 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8167 // CHECK9-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
8168 // CHECK9-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8169 // CHECK9:       .omp.final.then:
8170 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
8171 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8172 // CHECK9:       .omp.final.done:
8173 // CHECK9-NEXT:    ret void
8174 //
8175 //
8176 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..9
8177 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
8178 // CHECK9-NEXT:  entry:
8179 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8180 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8181 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8182 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8183 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8184 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8185 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8186 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8187 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8188 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8189 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8190 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8191 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8192 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8193 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8194 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8195 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
8196 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8197 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8198 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8199 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8200 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8201 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8202 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8203 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8204 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8205 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8206 // 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)
8207 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8208 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8209 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8210 // CHECK9:       cond.true:
8211 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8212 // CHECK9:       cond.false:
8213 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8214 // CHECK9-NEXT:    br label [[COND_END]]
8215 // CHECK9:       cond.end:
8216 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8217 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8218 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8219 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
8220 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8221 // CHECK9:       omp.inner.for.cond:
8222 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
8223 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !45
8224 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
8225 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8226 // CHECK9:       omp.inner.for.body:
8227 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
8228 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
8229 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8230 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !45
8231 // CHECK9-NEXT:    call void @_Z3fn6v(), !llvm.access.group !45
8232 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8233 // CHECK9:       omp.body.continue:
8234 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8235 // CHECK9:       omp.inner.for.inc:
8236 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
8237 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
8238 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
8239 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP46:![0-9]+]]
8240 // CHECK9:       omp.inner.for.end:
8241 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8242 // CHECK9:       omp.loop.exit:
8243 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8244 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8245 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
8246 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8247 // CHECK9:       .omp.final.then:
8248 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
8249 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8250 // CHECK9:       .omp.final.done:
8251 // CHECK9-NEXT:    ret void
8252 //
8253 //
8254 // CHECK9-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
8255 // CHECK9-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
8256 // CHECK9-NEXT:  entry:
8257 // CHECK9-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
8258 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8259 // CHECK9-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
8260 // CHECK9-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
8261 // CHECK9-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
8262 // CHECK9-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
8263 // CHECK9-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
8264 // CHECK9-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
8265 // CHECK9-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
8266 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
8267 // 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)
8268 // CHECK9-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
8269 // CHECK9-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8270 // CHECK9:       omp_offload.failed:
8271 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59() #[[ATTR2]]
8272 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT]]
8273 // CHECK9:       omp_offload.cont:
8274 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
8275 // 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)
8276 // CHECK9-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
8277 // CHECK9-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
8278 // CHECK9:       omp_offload.failed2:
8279 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65() #[[ATTR2]]
8280 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
8281 // CHECK9:       omp_offload.cont3:
8282 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
8283 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
8284 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
8285 // CHECK9-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
8286 // CHECK9-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8287 // CHECK9-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
8288 // CHECK9-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
8289 // CHECK9-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8290 // CHECK9-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
8291 // CHECK9-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
8292 // CHECK9-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
8293 // CHECK9-NEXT:    store i8* null, i8** [[TMP10]], align 8
8294 // CHECK9-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
8295 // CHECK9-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
8296 // CHECK9-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
8297 // 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)
8298 // CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
8299 // CHECK9-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
8300 // CHECK9:       omp_offload.failed5:
8301 // CHECK9-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71(i64 [[TMP5]]) #[[ATTR2]]
8302 // CHECK9-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
8303 // CHECK9:       omp_offload.cont6:
8304 // CHECK9-NEXT:    ret i32 0
8305 //
8306 //
8307 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59
8308 // CHECK9-SAME: () #[[ATTR1]] {
8309 // CHECK9-NEXT:  entry:
8310 // 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*, ...)*))
8311 // CHECK9-NEXT:    ret void
8312 //
8313 //
8314 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..10
8315 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
8316 // CHECK9-NEXT:  entry:
8317 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8318 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8319 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8320 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8321 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8322 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8323 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8324 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8325 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8326 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8327 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8328 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8329 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
8330 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8331 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8332 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8333 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8334 // 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)
8335 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8336 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
8337 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8338 // CHECK9:       cond.true:
8339 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8340 // CHECK9:       cond.false:
8341 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8342 // CHECK9-NEXT:    br label [[COND_END]]
8343 // CHECK9:       cond.end:
8344 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8345 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8346 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8347 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8348 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8349 // CHECK9:       omp.inner.for.cond:
8350 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !48
8351 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !48
8352 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8353 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8354 // CHECK9:       omp.inner.for.body:
8355 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !48
8356 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
8357 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !48
8358 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
8359 // 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
8360 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8361 // CHECK9:       omp.inner.for.inc:
8362 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !48
8363 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !48
8364 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
8365 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !48
8366 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP49:![0-9]+]]
8367 // CHECK9:       omp.inner.for.end:
8368 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8369 // CHECK9:       omp.loop.exit:
8370 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8371 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8372 // CHECK9-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
8373 // CHECK9-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8374 // CHECK9:       .omp.final.then:
8375 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
8376 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8377 // CHECK9:       .omp.final.done:
8378 // CHECK9-NEXT:    ret void
8379 //
8380 //
8381 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..11
8382 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
8383 // CHECK9-NEXT:  entry:
8384 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8385 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8386 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8387 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8388 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8389 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8390 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8391 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8392 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8393 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8394 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8395 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8396 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8397 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8398 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8399 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8400 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
8401 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8402 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8403 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8404 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8405 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8406 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8407 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8408 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8409 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8410 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8411 // 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)
8412 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8413 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8414 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8415 // CHECK9:       cond.true:
8416 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8417 // CHECK9:       cond.false:
8418 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8419 // CHECK9-NEXT:    br label [[COND_END]]
8420 // CHECK9:       cond.end:
8421 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8422 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8423 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8424 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
8425 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8426 // CHECK9:       omp.inner.for.cond:
8427 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
8428 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !51
8429 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
8430 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8431 // CHECK9:       omp.inner.for.body:
8432 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
8433 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
8434 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8435 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !51
8436 // CHECK9-NEXT:    call void @_Z3fn1v(), !llvm.access.group !51
8437 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8438 // CHECK9:       omp.body.continue:
8439 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8440 // CHECK9:       omp.inner.for.inc:
8441 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
8442 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
8443 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
8444 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP52:![0-9]+]]
8445 // CHECK9:       omp.inner.for.end:
8446 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8447 // CHECK9:       omp.loop.exit:
8448 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8449 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8450 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
8451 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8452 // CHECK9:       .omp.final.then:
8453 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
8454 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8455 // CHECK9:       .omp.final.done:
8456 // CHECK9-NEXT:    ret void
8457 //
8458 //
8459 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65
8460 // CHECK9-SAME: () #[[ATTR1]] {
8461 // CHECK9-NEXT:  entry:
8462 // 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*, ...)*))
8463 // CHECK9-NEXT:    ret void
8464 //
8465 //
8466 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..12
8467 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
8468 // CHECK9-NEXT:  entry:
8469 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8470 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8471 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8472 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8473 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8474 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8475 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8476 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8477 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8478 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
8479 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8480 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8481 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8482 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
8483 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8484 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8485 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8486 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8487 // 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)
8488 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8489 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
8490 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8491 // CHECK9:       cond.true:
8492 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8493 // CHECK9:       cond.false:
8494 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8495 // CHECK9-NEXT:    br label [[COND_END]]
8496 // CHECK9:       cond.end:
8497 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8498 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8499 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8500 // CHECK9-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8501 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8502 // CHECK9:       omp.inner.for.cond:
8503 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
8504 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !54
8505 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8506 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8507 // CHECK9:       omp.inner.for.body:
8508 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !54
8509 // CHECK9-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
8510 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !54
8511 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
8512 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !54
8513 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !54
8514 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !54
8515 // CHECK9-NEXT:    call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !54
8516 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !54
8517 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8518 // CHECK9:       omp.inner.for.inc:
8519 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
8520 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !54
8521 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
8522 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
8523 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP55:![0-9]+]]
8524 // CHECK9:       omp.inner.for.end:
8525 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8526 // CHECK9:       omp.loop.exit:
8527 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8528 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8529 // CHECK9-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
8530 // CHECK9-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8531 // CHECK9:       .omp.final.then:
8532 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
8533 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8534 // CHECK9:       .omp.final.done:
8535 // CHECK9-NEXT:    ret void
8536 //
8537 //
8538 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..13
8539 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
8540 // CHECK9-NEXT:  entry:
8541 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8542 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8543 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8544 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8545 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8546 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8547 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8548 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8549 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8550 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8551 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8552 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8553 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8554 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8555 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8556 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8557 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
8558 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8559 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8560 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8561 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8562 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8563 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8564 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8565 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8566 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8567 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8568 // 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)
8569 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8570 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8571 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8572 // CHECK9:       cond.true:
8573 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8574 // CHECK9:       cond.false:
8575 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8576 // CHECK9-NEXT:    br label [[COND_END]]
8577 // CHECK9:       cond.end:
8578 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8579 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8580 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8581 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
8582 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8583 // CHECK9:       omp.inner.for.cond:
8584 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !57
8585 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !57
8586 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
8587 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8588 // CHECK9:       omp.inner.for.body:
8589 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !57
8590 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
8591 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8592 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !57
8593 // CHECK9-NEXT:    call void @_Z3fn2v(), !llvm.access.group !57
8594 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8595 // CHECK9:       omp.body.continue:
8596 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8597 // CHECK9:       omp.inner.for.inc:
8598 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !57
8599 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
8600 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !57
8601 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP58:![0-9]+]]
8602 // CHECK9:       omp.inner.for.end:
8603 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8604 // CHECK9:       omp.loop.exit:
8605 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8606 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8607 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
8608 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8609 // CHECK9:       .omp.final.then:
8610 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
8611 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8612 // CHECK9:       .omp.final.done:
8613 // CHECK9-NEXT:    ret void
8614 //
8615 //
8616 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71
8617 // CHECK9-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
8618 // CHECK9-NEXT:  entry:
8619 // CHECK9-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
8620 // CHECK9-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
8621 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
8622 // 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]])
8623 // CHECK9-NEXT:    ret void
8624 //
8625 //
8626 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..14
8627 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
8628 // CHECK9-NEXT:  entry:
8629 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8630 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8631 // CHECK9-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
8632 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8633 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8634 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8635 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8636 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8637 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8638 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8639 // CHECK9-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
8640 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8641 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8642 // CHECK9-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
8643 // CHECK9-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
8644 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8645 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
8646 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8647 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8648 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8649 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
8650 // 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)
8651 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8652 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
8653 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8654 // CHECK9:       cond.true:
8655 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8656 // CHECK9:       cond.false:
8657 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8658 // CHECK9-NEXT:    br label [[COND_END]]
8659 // CHECK9:       cond.end:
8660 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
8661 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8662 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8663 // CHECK9-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
8664 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8665 // CHECK9:       omp.inner.for.cond:
8666 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !60
8667 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !60
8668 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
8669 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8670 // CHECK9:       omp.inner.for.body:
8671 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !60
8672 // CHECK9-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
8673 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !60
8674 // CHECK9-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
8675 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !60
8676 // CHECK9-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
8677 // CHECK9-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
8678 // CHECK9:       omp_if.then:
8679 // 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
8680 // CHECK9-NEXT:    br label [[OMP_IF_END:%.*]]
8681 // CHECK9:       omp_if.else:
8682 // CHECK9-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !60
8683 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !60
8684 // CHECK9-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !60
8685 // CHECK9-NEXT:    call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !60
8686 // CHECK9-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !60
8687 // CHECK9-NEXT:    br label [[OMP_IF_END]]
8688 // CHECK9:       omp_if.end:
8689 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8690 // CHECK9:       omp.inner.for.inc:
8691 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !60
8692 // CHECK9-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !60
8693 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
8694 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !60
8695 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP61:![0-9]+]]
8696 // CHECK9:       omp.inner.for.end:
8697 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8698 // CHECK9:       omp.loop.exit:
8699 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
8700 // CHECK9-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8701 // CHECK9-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
8702 // CHECK9-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8703 // CHECK9:       .omp.final.then:
8704 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
8705 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8706 // CHECK9:       .omp.final.done:
8707 // CHECK9-NEXT:    ret void
8708 //
8709 //
8710 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..15
8711 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
8712 // CHECK9-NEXT:  entry:
8713 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8714 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8715 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8716 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8717 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8718 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8719 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8720 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8721 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8722 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8723 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
8724 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8725 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8726 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8727 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8728 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8729 // CHECK9-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
8730 // CHECK9-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8731 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8732 // CHECK9-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8733 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8734 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8735 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8736 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8737 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8738 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8739 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8740 // 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)
8741 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8742 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8743 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8744 // CHECK9:       cond.true:
8745 // CHECK9-NEXT:    br label [[COND_END:%.*]]
8746 // CHECK9:       cond.false:
8747 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8748 // CHECK9-NEXT:    br label [[COND_END]]
8749 // CHECK9:       cond.end:
8750 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8751 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8752 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8753 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
8754 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8755 // CHECK9:       omp.inner.for.cond:
8756 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !63
8757 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !63
8758 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
8759 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8760 // CHECK9:       omp.inner.for.body:
8761 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !63
8762 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
8763 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8764 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !63
8765 // CHECK9-NEXT:    call void @_Z3fn3v(), !llvm.access.group !63
8766 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8767 // CHECK9:       omp.body.continue:
8768 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8769 // CHECK9:       omp.inner.for.inc:
8770 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !63
8771 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
8772 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !63
8773 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP64:![0-9]+]]
8774 // CHECK9:       omp.inner.for.end:
8775 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8776 // CHECK9:       omp.loop.exit:
8777 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8778 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8779 // CHECK9-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
8780 // CHECK9-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8781 // CHECK9:       .omp.final.then:
8782 // CHECK9-NEXT:    store i32 100, i32* [[I]], align 4
8783 // CHECK9-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8784 // CHECK9:       .omp.final.done:
8785 // CHECK9-NEXT:    ret void
8786 //
8787 //
8788 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
8789 // CHECK9-SAME: () #[[ATTR5:[0-9]+]] {
8790 // CHECK9-NEXT:  entry:
8791 // CHECK9-NEXT:    call void @__tgt_register_requires(i64 1)
8792 // CHECK9-NEXT:    ret void
8793 //
8794 //
8795 // CHECK10-LABEL: define {{[^@]+}}@_Z9gtid_testv
8796 // CHECK10-SAME: () #[[ATTR0:[0-9]+]] {
8797 // CHECK10-NEXT:  entry:
8798 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8799 // CHECK10-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
8800 // CHECK10-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
8801 // CHECK10-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)
8802 // CHECK10-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
8803 // CHECK10-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
8804 // CHECK10:       omp_offload.failed:
8805 // CHECK10-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43() #[[ATTR2:[0-9]+]]
8806 // CHECK10-NEXT:    br label [[OMP_OFFLOAD_CONT]]
8807 // CHECK10:       omp_offload.cont:
8808 // CHECK10-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
8809 // CHECK10-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)
8810 // CHECK10-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
8811 // CHECK10-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
8812 // CHECK10:       omp_offload.failed2:
8813 // CHECK10-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48() #[[ATTR2]]
8814 // CHECK10-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
8815 // CHECK10:       omp_offload.cont3:
8816 // CHECK10-NEXT:    ret void
8817 //
8818 //
8819 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43
8820 // CHECK10-SAME: () #[[ATTR1:[0-9]+]] {
8821 // CHECK10-NEXT:  entry:
8822 // CHECK10-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*, ...)*))
8823 // CHECK10-NEXT:    ret void
8824 //
8825 //
8826 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined.
8827 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
8828 // CHECK10-NEXT:  entry:
8829 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8830 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8831 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8832 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8833 // CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8834 // CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8835 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8836 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8837 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
8838 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8839 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8840 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8841 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
8842 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8843 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8844 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8845 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8846 // CHECK10-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)
8847 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8848 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
8849 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8850 // CHECK10:       cond.true:
8851 // CHECK10-NEXT:    br label [[COND_END:%.*]]
8852 // CHECK10:       cond.false:
8853 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
8854 // CHECK10-NEXT:    br label [[COND_END]]
8855 // CHECK10:       cond.end:
8856 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
8857 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
8858 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
8859 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
8860 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8861 // CHECK10:       omp.inner.for.cond:
8862 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
8863 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
8864 // CHECK10-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
8865 // CHECK10-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8866 // CHECK10:       omp.inner.for.body:
8867 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !15
8868 // CHECK10-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
8869 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
8870 // CHECK10-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
8871 // CHECK10-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
8872 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8873 // CHECK10:       omp.inner.for.inc:
8874 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
8875 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !15
8876 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
8877 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
8878 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
8879 // CHECK10:       omp.inner.for.end:
8880 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8881 // CHECK10:       omp.loop.exit:
8882 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
8883 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8884 // CHECK10-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
8885 // CHECK10-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8886 // CHECK10:       .omp.final.then:
8887 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
8888 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8889 // CHECK10:       .omp.final.done:
8890 // CHECK10-NEXT:    ret void
8891 //
8892 //
8893 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..1
8894 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
8895 // CHECK10-NEXT:  entry:
8896 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8897 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8898 // CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
8899 // CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
8900 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8901 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8902 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
8903 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
8904 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8905 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8906 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
8907 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8908 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8909 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8910 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8911 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
8912 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
8913 // CHECK10-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
8914 // CHECK10-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
8915 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
8916 // CHECK10-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
8917 // CHECK10-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
8918 // CHECK10-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
8919 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8920 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8921 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8922 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
8923 // CHECK10-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)
8924 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8925 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
8926 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8927 // CHECK10:       cond.true:
8928 // CHECK10-NEXT:    br label [[COND_END:%.*]]
8929 // CHECK10:       cond.false:
8930 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
8931 // CHECK10-NEXT:    br label [[COND_END]]
8932 // CHECK10:       cond.end:
8933 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
8934 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
8935 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
8936 // CHECK10-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
8937 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8938 // CHECK10:       omp.inner.for.cond:
8939 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
8940 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
8941 // CHECK10-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
8942 // CHECK10-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8943 // CHECK10:       omp.inner.for.body:
8944 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
8945 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
8946 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
8947 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !19
8948 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8949 // CHECK10:       omp.body.continue:
8950 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8951 // CHECK10:       omp.inner.for.inc:
8952 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
8953 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
8954 // CHECK10-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
8955 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
8956 // CHECK10:       omp.inner.for.end:
8957 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8958 // CHECK10:       omp.loop.exit:
8959 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
8960 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
8961 // CHECK10-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
8962 // CHECK10-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
8963 // CHECK10:       .omp.final.then:
8964 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
8965 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
8966 // CHECK10:       .omp.final.done:
8967 // CHECK10-NEXT:    ret void
8968 //
8969 //
8970 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48
8971 // CHECK10-SAME: () #[[ATTR1]] {
8972 // CHECK10-NEXT:  entry:
8973 // CHECK10-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*, ...)*))
8974 // CHECK10-NEXT:    ret void
8975 //
8976 //
8977 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..2
8978 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
8979 // CHECK10-NEXT:  entry:
8980 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
8981 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
8982 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
8983 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
8984 // CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
8985 // CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
8986 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
8987 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8988 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
8989 // CHECK10-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
8990 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8991 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8992 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
8993 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
8994 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
8995 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8996 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8997 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
8998 // CHECK10-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)
8999 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9000 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
9001 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9002 // CHECK10:       cond.true:
9003 // CHECK10-NEXT:    br label [[COND_END:%.*]]
9004 // CHECK10:       cond.false:
9005 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9006 // CHECK10-NEXT:    br label [[COND_END]]
9007 // CHECK10:       cond.end:
9008 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9009 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9010 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9011 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9012 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9013 // CHECK10:       omp.inner.for.cond:
9014 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
9015 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
9016 // CHECK10-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9017 // CHECK10-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9018 // CHECK10:       omp.inner.for.body:
9019 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
9020 // CHECK10-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
9021 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
9022 // CHECK10-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
9023 // CHECK10-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !24
9024 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !24
9025 // CHECK10-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !24
9026 // CHECK10-NEXT:    call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !24
9027 // CHECK10-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !24
9028 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9029 // CHECK10:       omp.inner.for.inc:
9030 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
9031 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
9032 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
9033 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
9034 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
9035 // CHECK10:       omp.inner.for.end:
9036 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9037 // CHECK10:       omp.loop.exit:
9038 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9039 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9040 // CHECK10-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
9041 // CHECK10-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9042 // CHECK10:       .omp.final.then:
9043 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
9044 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9045 // CHECK10:       .omp.final.done:
9046 // CHECK10-NEXT:    ret void
9047 //
9048 //
9049 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..3
9050 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
9051 // CHECK10-NEXT:  entry:
9052 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9053 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9054 // CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9055 // CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9056 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9057 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9058 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9059 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9060 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9061 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9062 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
9063 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9064 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9065 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9066 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9067 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9068 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
9069 // CHECK10-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9070 // CHECK10-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
9071 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9072 // CHECK10-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9073 // CHECK10-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9074 // CHECK10-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9075 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9076 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9077 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9078 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9079 // CHECK10-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)
9080 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9081 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
9082 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9083 // CHECK10:       cond.true:
9084 // CHECK10-NEXT:    br label [[COND_END:%.*]]
9085 // CHECK10:       cond.false:
9086 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9087 // CHECK10-NEXT:    br label [[COND_END]]
9088 // CHECK10:       cond.end:
9089 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9090 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9091 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9092 // CHECK10-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9093 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9094 // CHECK10:       omp.inner.for.cond:
9095 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
9096 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
9097 // CHECK10-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9098 // CHECK10-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9099 // CHECK10:       omp.inner.for.body:
9100 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
9101 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
9102 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9103 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !27
9104 // CHECK10-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !27
9105 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9106 // CHECK10:       omp.body.continue:
9107 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9108 // CHECK10:       omp.inner.for.inc:
9109 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
9110 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
9111 // CHECK10-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
9112 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
9113 // CHECK10:       omp.inner.for.end:
9114 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9115 // CHECK10:       omp.loop.exit:
9116 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9117 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9118 // CHECK10-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
9119 // CHECK10-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9120 // CHECK10:       .omp.final.then:
9121 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
9122 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9123 // CHECK10:       .omp.final.done:
9124 // CHECK10-NEXT:    ret void
9125 //
9126 //
9127 // CHECK10-LABEL: define {{[^@]+}}@main
9128 // CHECK10-SAME: () #[[ATTR3:[0-9]+]] {
9129 // CHECK10-NEXT:  entry:
9130 // CHECK10-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
9131 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9132 // CHECK10-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
9133 // CHECK10-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
9134 // CHECK10-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
9135 // CHECK10-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
9136 // CHECK10-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
9137 // CHECK10-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
9138 // CHECK10-NEXT:    store i32 0, i32* [[RETVAL]], align 4
9139 // CHECK10-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
9140 // CHECK10-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)
9141 // CHECK10-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
9142 // CHECK10-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
9143 // CHECK10:       omp_offload.failed:
9144 // CHECK10-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81() #[[ATTR2]]
9145 // CHECK10-NEXT:    br label [[OMP_OFFLOAD_CONT]]
9146 // CHECK10:       omp_offload.cont:
9147 // CHECK10-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
9148 // CHECK10-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)
9149 // CHECK10-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
9150 // CHECK10-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
9151 // CHECK10:       omp_offload.failed2:
9152 // CHECK10-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90() #[[ATTR2]]
9153 // CHECK10-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
9154 // CHECK10:       omp_offload.cont3:
9155 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* @Arg, align 4
9156 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
9157 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
9158 // CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
9159 // CHECK10-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9160 // CHECK10-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
9161 // CHECK10-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
9162 // CHECK10-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9163 // CHECK10-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
9164 // CHECK10-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
9165 // CHECK10-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
9166 // CHECK10-NEXT:    store i8* null, i8** [[TMP10]], align 8
9167 // CHECK10-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9168 // CHECK10-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9169 // CHECK10-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
9170 // CHECK10-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)
9171 // CHECK10-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
9172 // CHECK10-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
9173 // CHECK10:       omp_offload.failed5:
9174 // CHECK10-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99(i64 [[TMP5]]) #[[ATTR2]]
9175 // CHECK10-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
9176 // CHECK10:       omp_offload.cont6:
9177 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
9178 // CHECK10-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
9179 // CHECK10-NEXT:    ret i32 [[CALL]]
9180 //
9181 //
9182 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
9183 // CHECK10-SAME: () #[[ATTR1]] {
9184 // CHECK10-NEXT:  entry:
9185 // CHECK10-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*, ...)*))
9186 // CHECK10-NEXT:    ret void
9187 //
9188 //
9189 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..4
9190 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
9191 // CHECK10-NEXT:  entry:
9192 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9193 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9194 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9195 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9196 // CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9197 // CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9198 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9199 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9200 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
9201 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9202 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9203 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9204 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
9205 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9206 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9207 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9208 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9209 // CHECK10-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)
9210 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9211 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
9212 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9213 // CHECK10:       cond.true:
9214 // CHECK10-NEXT:    br label [[COND_END:%.*]]
9215 // CHECK10:       cond.false:
9216 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9217 // CHECK10-NEXT:    br label [[COND_END]]
9218 // CHECK10:       cond.end:
9219 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9220 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9221 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9222 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9223 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9224 // CHECK10:       omp.inner.for.cond:
9225 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
9226 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
9227 // CHECK10-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9228 // CHECK10-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9229 // CHECK10:       omp.inner.for.body:
9230 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
9231 // CHECK10-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
9232 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
9233 // CHECK10-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
9234 // CHECK10-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
9235 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9236 // CHECK10:       omp.inner.for.inc:
9237 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
9238 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
9239 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
9240 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
9241 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
9242 // CHECK10:       omp.inner.for.end:
9243 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9244 // CHECK10:       omp.loop.exit:
9245 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9246 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9247 // CHECK10-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
9248 // CHECK10-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9249 // CHECK10:       .omp.final.then:
9250 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
9251 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9252 // CHECK10:       .omp.final.done:
9253 // CHECK10-NEXT:    ret void
9254 //
9255 //
9256 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..5
9257 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
9258 // CHECK10-NEXT:  entry:
9259 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9260 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9261 // CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9262 // CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9263 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9264 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9265 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9266 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9267 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9268 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9269 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
9270 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9271 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9272 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9273 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9274 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9275 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
9276 // CHECK10-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9277 // CHECK10-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
9278 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9279 // CHECK10-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9280 // CHECK10-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9281 // CHECK10-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9282 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9283 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9284 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9285 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9286 // CHECK10-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)
9287 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9288 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
9289 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9290 // CHECK10:       cond.true:
9291 // CHECK10-NEXT:    br label [[COND_END:%.*]]
9292 // CHECK10:       cond.false:
9293 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9294 // CHECK10-NEXT:    br label [[COND_END]]
9295 // CHECK10:       cond.end:
9296 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9297 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9298 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9299 // CHECK10-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9300 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9301 // CHECK10:       omp.inner.for.cond:
9302 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
9303 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
9304 // CHECK10-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9305 // CHECK10-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9306 // CHECK10:       omp.inner.for.body:
9307 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
9308 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
9309 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9310 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
9311 // CHECK10-NEXT:    call void @_Z3fn4v(), !llvm.access.group !33
9312 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9313 // CHECK10:       omp.body.continue:
9314 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9315 // CHECK10:       omp.inner.for.inc:
9316 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
9317 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
9318 // CHECK10-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
9319 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
9320 // CHECK10:       omp.inner.for.end:
9321 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9322 // CHECK10:       omp.loop.exit:
9323 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9324 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9325 // CHECK10-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
9326 // CHECK10-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9327 // CHECK10:       .omp.final.then:
9328 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
9329 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9330 // CHECK10:       .omp.final.done:
9331 // CHECK10-NEXT:    ret void
9332 //
9333 //
9334 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90
9335 // CHECK10-SAME: () #[[ATTR1]] {
9336 // CHECK10-NEXT:  entry:
9337 // CHECK10-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*, ...)*))
9338 // CHECK10-NEXT:    ret void
9339 //
9340 //
9341 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..6
9342 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
9343 // CHECK10-NEXT:  entry:
9344 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9345 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9346 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9347 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9348 // CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9349 // CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9350 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9351 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9352 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
9353 // CHECK10-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
9354 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9355 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9356 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9357 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
9358 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9359 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9360 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9361 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9362 // CHECK10-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)
9363 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9364 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
9365 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9366 // CHECK10:       cond.true:
9367 // CHECK10-NEXT:    br label [[COND_END:%.*]]
9368 // CHECK10:       cond.false:
9369 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9370 // CHECK10-NEXT:    br label [[COND_END]]
9371 // CHECK10:       cond.end:
9372 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9373 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9374 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9375 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9376 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9377 // CHECK10:       omp.inner.for.cond:
9378 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
9379 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
9380 // CHECK10-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9381 // CHECK10-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9382 // CHECK10:       omp.inner.for.body:
9383 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !36
9384 // CHECK10-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
9385 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !36
9386 // CHECK10-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
9387 // CHECK10-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !36
9388 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !36
9389 // CHECK10-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !36
9390 // CHECK10-NEXT:    call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !36
9391 // CHECK10-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !36
9392 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9393 // CHECK10:       omp.inner.for.inc:
9394 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
9395 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !36
9396 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
9397 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !36
9398 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP37:![0-9]+]]
9399 // CHECK10:       omp.inner.for.end:
9400 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9401 // CHECK10:       omp.loop.exit:
9402 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9403 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9404 // CHECK10-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
9405 // CHECK10-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9406 // CHECK10:       .omp.final.then:
9407 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
9408 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9409 // CHECK10:       .omp.final.done:
9410 // CHECK10-NEXT:    ret void
9411 //
9412 //
9413 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..7
9414 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
9415 // CHECK10-NEXT:  entry:
9416 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9417 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9418 // CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9419 // CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9420 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9421 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9422 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9423 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9424 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9425 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9426 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
9427 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9428 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9429 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9430 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9431 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9432 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
9433 // CHECK10-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9434 // CHECK10-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
9435 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9436 // CHECK10-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9437 // CHECK10-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9438 // CHECK10-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9439 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9440 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9441 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9442 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9443 // CHECK10-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)
9444 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9445 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
9446 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9447 // CHECK10:       cond.true:
9448 // CHECK10-NEXT:    br label [[COND_END:%.*]]
9449 // CHECK10:       cond.false:
9450 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9451 // CHECK10-NEXT:    br label [[COND_END]]
9452 // CHECK10:       cond.end:
9453 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9454 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9455 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9456 // CHECK10-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9457 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9458 // CHECK10:       omp.inner.for.cond:
9459 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
9460 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !39
9461 // CHECK10-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9462 // CHECK10-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9463 // CHECK10:       omp.inner.for.body:
9464 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
9465 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
9466 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9467 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !39
9468 // CHECK10-NEXT:    call void @_Z3fn5v(), !llvm.access.group !39
9469 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9470 // CHECK10:       omp.body.continue:
9471 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9472 // CHECK10:       omp.inner.for.inc:
9473 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
9474 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
9475 // CHECK10-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
9476 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
9477 // CHECK10:       omp.inner.for.end:
9478 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9479 // CHECK10:       omp.loop.exit:
9480 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9481 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9482 // CHECK10-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
9483 // CHECK10-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9484 // CHECK10:       .omp.final.then:
9485 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
9486 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9487 // CHECK10:       .omp.final.done:
9488 // CHECK10-NEXT:    ret void
9489 //
9490 //
9491 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99
9492 // CHECK10-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
9493 // CHECK10-NEXT:  entry:
9494 // CHECK10-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
9495 // CHECK10-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
9496 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
9497 // CHECK10-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]])
9498 // CHECK10-NEXT:    ret void
9499 //
9500 //
9501 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..8
9502 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
9503 // CHECK10-NEXT:  entry:
9504 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9505 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9506 // CHECK10-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
9507 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9508 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9509 // CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9510 // CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9511 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9512 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9513 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
9514 // CHECK10-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
9515 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9516 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9517 // CHECK10-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
9518 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
9519 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9520 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
9521 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9522 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9523 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9524 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
9525 // CHECK10-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)
9526 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9527 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
9528 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9529 // CHECK10:       cond.true:
9530 // CHECK10-NEXT:    br label [[COND_END:%.*]]
9531 // CHECK10:       cond.false:
9532 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9533 // CHECK10-NEXT:    br label [[COND_END]]
9534 // CHECK10:       cond.end:
9535 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
9536 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9537 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9538 // CHECK10-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
9539 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9540 // CHECK10:       omp.inner.for.cond:
9541 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
9542 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !42
9543 // CHECK10-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
9544 // CHECK10-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9545 // CHECK10:       omp.inner.for.body:
9546 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !42
9547 // CHECK10-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
9548 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !42
9549 // CHECK10-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
9550 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !42
9551 // CHECK10-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
9552 // CHECK10-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
9553 // CHECK10:       omp_if.then:
9554 // CHECK10-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
9555 // CHECK10-NEXT:    br label [[OMP_IF_END:%.*]]
9556 // CHECK10:       omp_if.else:
9557 // CHECK10-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !42
9558 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !42
9559 // CHECK10-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !42
9560 // CHECK10-NEXT:    call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !42
9561 // CHECK10-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !42
9562 // CHECK10-NEXT:    br label [[OMP_IF_END]]
9563 // CHECK10:       omp_if.end:
9564 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9565 // CHECK10:       omp.inner.for.inc:
9566 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
9567 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !42
9568 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
9569 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !42
9570 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP43:![0-9]+]]
9571 // CHECK10:       omp.inner.for.end:
9572 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9573 // CHECK10:       omp.loop.exit:
9574 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
9575 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9576 // CHECK10-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
9577 // CHECK10-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9578 // CHECK10:       .omp.final.then:
9579 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
9580 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9581 // CHECK10:       .omp.final.done:
9582 // CHECK10-NEXT:    ret void
9583 //
9584 //
9585 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..9
9586 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
9587 // CHECK10-NEXT:  entry:
9588 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9589 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9590 // CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9591 // CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9592 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9593 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9594 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9595 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9596 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9597 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9598 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
9599 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9600 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9601 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9602 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9603 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9604 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
9605 // CHECK10-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9606 // CHECK10-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
9607 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9608 // CHECK10-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9609 // CHECK10-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9610 // CHECK10-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9611 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9612 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9613 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9614 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9615 // CHECK10-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)
9616 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9617 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
9618 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9619 // CHECK10:       cond.true:
9620 // CHECK10-NEXT:    br label [[COND_END:%.*]]
9621 // CHECK10:       cond.false:
9622 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9623 // CHECK10-NEXT:    br label [[COND_END]]
9624 // CHECK10:       cond.end:
9625 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9626 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9627 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9628 // CHECK10-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9629 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9630 // CHECK10:       omp.inner.for.cond:
9631 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
9632 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !45
9633 // CHECK10-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9634 // CHECK10-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9635 // CHECK10:       omp.inner.for.body:
9636 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
9637 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
9638 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9639 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !45
9640 // CHECK10-NEXT:    call void @_Z3fn6v(), !llvm.access.group !45
9641 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9642 // CHECK10:       omp.body.continue:
9643 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9644 // CHECK10:       omp.inner.for.inc:
9645 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
9646 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
9647 // CHECK10-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !45
9648 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP46:![0-9]+]]
9649 // CHECK10:       omp.inner.for.end:
9650 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9651 // CHECK10:       omp.loop.exit:
9652 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9653 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9654 // CHECK10-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
9655 // CHECK10-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9656 // CHECK10:       .omp.final.then:
9657 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
9658 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9659 // CHECK10:       .omp.final.done:
9660 // CHECK10-NEXT:    ret void
9661 //
9662 //
9663 // CHECK10-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
9664 // CHECK10-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
9665 // CHECK10-NEXT:  entry:
9666 // CHECK10-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
9667 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9668 // CHECK10-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
9669 // CHECK10-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
9670 // CHECK10-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
9671 // CHECK10-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
9672 // CHECK10-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
9673 // CHECK10-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
9674 // CHECK10-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
9675 // CHECK10-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
9676 // CHECK10-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)
9677 // CHECK10-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
9678 // CHECK10-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
9679 // CHECK10:       omp_offload.failed:
9680 // CHECK10-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59() #[[ATTR2]]
9681 // CHECK10-NEXT:    br label [[OMP_OFFLOAD_CONT]]
9682 // CHECK10:       omp_offload.cont:
9683 // CHECK10-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
9684 // CHECK10-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)
9685 // CHECK10-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
9686 // CHECK10-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
9687 // CHECK10:       omp_offload.failed2:
9688 // CHECK10-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65() #[[ATTR2]]
9689 // CHECK10-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
9690 // CHECK10:       omp_offload.cont3:
9691 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
9692 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
9693 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
9694 // CHECK10-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
9695 // CHECK10-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9696 // CHECK10-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
9697 // CHECK10-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
9698 // CHECK10-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9699 // CHECK10-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
9700 // CHECK10-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
9701 // CHECK10-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
9702 // CHECK10-NEXT:    store i8* null, i8** [[TMP10]], align 8
9703 // CHECK10-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
9704 // CHECK10-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
9705 // CHECK10-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
9706 // CHECK10-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)
9707 // CHECK10-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
9708 // CHECK10-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
9709 // CHECK10:       omp_offload.failed5:
9710 // CHECK10-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71(i64 [[TMP5]]) #[[ATTR2]]
9711 // CHECK10-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
9712 // CHECK10:       omp_offload.cont6:
9713 // CHECK10-NEXT:    ret i32 0
9714 //
9715 //
9716 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59
9717 // CHECK10-SAME: () #[[ATTR1]] {
9718 // CHECK10-NEXT:  entry:
9719 // CHECK10-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*, ...)*))
9720 // CHECK10-NEXT:    ret void
9721 //
9722 //
9723 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..10
9724 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
9725 // CHECK10-NEXT:  entry:
9726 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9727 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9728 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9729 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9730 // CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9731 // CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9732 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9733 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9734 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
9735 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9736 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9737 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9738 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
9739 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9740 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9741 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9742 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9743 // CHECK10-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)
9744 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9745 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
9746 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9747 // CHECK10:       cond.true:
9748 // CHECK10-NEXT:    br label [[COND_END:%.*]]
9749 // CHECK10:       cond.false:
9750 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9751 // CHECK10-NEXT:    br label [[COND_END]]
9752 // CHECK10:       cond.end:
9753 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9754 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9755 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9756 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9757 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9758 // CHECK10:       omp.inner.for.cond:
9759 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !48
9760 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !48
9761 // CHECK10-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9762 // CHECK10-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9763 // CHECK10:       omp.inner.for.body:
9764 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !48
9765 // CHECK10-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
9766 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !48
9767 // CHECK10-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
9768 // CHECK10-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
9769 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9770 // CHECK10:       omp.inner.for.inc:
9771 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !48
9772 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !48
9773 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
9774 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !48
9775 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP49:![0-9]+]]
9776 // CHECK10:       omp.inner.for.end:
9777 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9778 // CHECK10:       omp.loop.exit:
9779 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9780 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9781 // CHECK10-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
9782 // CHECK10-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9783 // CHECK10:       .omp.final.then:
9784 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
9785 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9786 // CHECK10:       .omp.final.done:
9787 // CHECK10-NEXT:    ret void
9788 //
9789 //
9790 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..11
9791 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
9792 // CHECK10-NEXT:  entry:
9793 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9794 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9795 // CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9796 // CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9797 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9798 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9799 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9800 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9801 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9802 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9803 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
9804 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9805 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9806 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9807 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9808 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9809 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
9810 // CHECK10-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9811 // CHECK10-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
9812 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9813 // CHECK10-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9814 // CHECK10-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9815 // CHECK10-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9816 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9817 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9818 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9819 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9820 // CHECK10-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)
9821 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9822 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
9823 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9824 // CHECK10:       cond.true:
9825 // CHECK10-NEXT:    br label [[COND_END:%.*]]
9826 // CHECK10:       cond.false:
9827 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9828 // CHECK10-NEXT:    br label [[COND_END]]
9829 // CHECK10:       cond.end:
9830 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9831 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9832 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9833 // CHECK10-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9834 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9835 // CHECK10:       omp.inner.for.cond:
9836 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
9837 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !51
9838 // CHECK10-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9839 // CHECK10-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9840 // CHECK10:       omp.inner.for.body:
9841 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
9842 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
9843 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
9844 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !51
9845 // CHECK10-NEXT:    call void @_Z3fn1v(), !llvm.access.group !51
9846 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
9847 // CHECK10:       omp.body.continue:
9848 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9849 // CHECK10:       omp.inner.for.inc:
9850 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
9851 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
9852 // CHECK10-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
9853 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP52:![0-9]+]]
9854 // CHECK10:       omp.inner.for.end:
9855 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9856 // CHECK10:       omp.loop.exit:
9857 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
9858 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9859 // CHECK10-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
9860 // CHECK10-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9861 // CHECK10:       .omp.final.then:
9862 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
9863 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9864 // CHECK10:       .omp.final.done:
9865 // CHECK10-NEXT:    ret void
9866 //
9867 //
9868 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65
9869 // CHECK10-SAME: () #[[ATTR1]] {
9870 // CHECK10-NEXT:  entry:
9871 // CHECK10-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*, ...)*))
9872 // CHECK10-NEXT:    ret void
9873 //
9874 //
9875 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..12
9876 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
9877 // CHECK10-NEXT:  entry:
9878 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9879 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9880 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9881 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9882 // CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
9883 // CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
9884 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9885 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9886 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
9887 // CHECK10-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
9888 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9889 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9890 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
9891 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
9892 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9893 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9894 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9895 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
9896 // CHECK10-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)
9897 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9898 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
9899 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9900 // CHECK10:       cond.true:
9901 // CHECK10-NEXT:    br label [[COND_END:%.*]]
9902 // CHECK10:       cond.false:
9903 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
9904 // CHECK10-NEXT:    br label [[COND_END]]
9905 // CHECK10:       cond.end:
9906 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
9907 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
9908 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
9909 // CHECK10-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
9910 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9911 // CHECK10:       omp.inner.for.cond:
9912 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
9913 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !54
9914 // CHECK10-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
9915 // CHECK10-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9916 // CHECK10:       omp.inner.for.body:
9917 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !54
9918 // CHECK10-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
9919 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !54
9920 // CHECK10-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
9921 // CHECK10-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !54
9922 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !54
9923 // CHECK10-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !54
9924 // CHECK10-NEXT:    call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !54
9925 // CHECK10-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !54
9926 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
9927 // CHECK10:       omp.inner.for.inc:
9928 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
9929 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !54
9930 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
9931 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
9932 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP55:![0-9]+]]
9933 // CHECK10:       omp.inner.for.end:
9934 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
9935 // CHECK10:       omp.loop.exit:
9936 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
9937 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
9938 // CHECK10-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
9939 // CHECK10-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
9940 // CHECK10:       .omp.final.then:
9941 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
9942 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
9943 // CHECK10:       .omp.final.done:
9944 // CHECK10-NEXT:    ret void
9945 //
9946 //
9947 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..13
9948 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
9949 // CHECK10-NEXT:  entry:
9950 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
9951 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
9952 // CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
9953 // CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
9954 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
9955 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
9956 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
9957 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
9958 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
9959 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
9960 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
9961 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
9962 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
9963 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9964 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9965 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
9966 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
9967 // CHECK10-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
9968 // CHECK10-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
9969 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
9970 // CHECK10-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
9971 // CHECK10-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
9972 // CHECK10-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
9973 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
9974 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
9975 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
9976 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
9977 // CHECK10-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)
9978 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9979 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
9980 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
9981 // CHECK10:       cond.true:
9982 // CHECK10-NEXT:    br label [[COND_END:%.*]]
9983 // CHECK10:       cond.false:
9984 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
9985 // CHECK10-NEXT:    br label [[COND_END]]
9986 // CHECK10:       cond.end:
9987 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
9988 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
9989 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
9990 // CHECK10-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
9991 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
9992 // CHECK10:       omp.inner.for.cond:
9993 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !57
9994 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !57
9995 // CHECK10-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
9996 // CHECK10-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
9997 // CHECK10:       omp.inner.for.body:
9998 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !57
9999 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
10000 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10001 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !57
10002 // CHECK10-NEXT:    call void @_Z3fn2v(), !llvm.access.group !57
10003 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10004 // CHECK10:       omp.body.continue:
10005 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10006 // CHECK10:       omp.inner.for.inc:
10007 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !57
10008 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
10009 // CHECK10-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !57
10010 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP58:![0-9]+]]
10011 // CHECK10:       omp.inner.for.end:
10012 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10013 // CHECK10:       omp.loop.exit:
10014 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
10015 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10016 // CHECK10-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
10017 // CHECK10-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10018 // CHECK10:       .omp.final.then:
10019 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
10020 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10021 // CHECK10:       .omp.final.done:
10022 // CHECK10-NEXT:    ret void
10023 //
10024 //
10025 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71
10026 // CHECK10-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
10027 // CHECK10-NEXT:  entry:
10028 // CHECK10-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
10029 // CHECK10-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
10030 // CHECK10-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
10031 // CHECK10-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]])
10032 // CHECK10-NEXT:    ret void
10033 //
10034 //
10035 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..14
10036 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
10037 // CHECK10-NEXT:  entry:
10038 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10039 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10040 // CHECK10-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
10041 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10042 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10043 // CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10044 // CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10045 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10046 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10047 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
10048 // CHECK10-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
10049 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10050 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10051 // CHECK10-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
10052 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
10053 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
10054 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
10055 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10056 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10057 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10058 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
10059 // CHECK10-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)
10060 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10061 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
10062 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10063 // CHECK10:       cond.true:
10064 // CHECK10-NEXT:    br label [[COND_END:%.*]]
10065 // CHECK10:       cond.false:
10066 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10067 // CHECK10-NEXT:    br label [[COND_END]]
10068 // CHECK10:       cond.end:
10069 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
10070 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
10071 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10072 // CHECK10-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
10073 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10074 // CHECK10:       omp.inner.for.cond:
10075 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !60
10076 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !60
10077 // CHECK10-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
10078 // CHECK10-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10079 // CHECK10:       omp.inner.for.body:
10080 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !60
10081 // CHECK10-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
10082 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !60
10083 // CHECK10-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
10084 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !60
10085 // CHECK10-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
10086 // CHECK10-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
10087 // CHECK10:       omp_if.then:
10088 // CHECK10-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
10089 // CHECK10-NEXT:    br label [[OMP_IF_END:%.*]]
10090 // CHECK10:       omp_if.else:
10091 // CHECK10-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !60
10092 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !60
10093 // CHECK10-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !60
10094 // CHECK10-NEXT:    call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !60
10095 // CHECK10-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !60
10096 // CHECK10-NEXT:    br label [[OMP_IF_END]]
10097 // CHECK10:       omp_if.end:
10098 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10099 // CHECK10:       omp.inner.for.inc:
10100 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !60
10101 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !60
10102 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
10103 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !60
10104 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP61:![0-9]+]]
10105 // CHECK10:       omp.inner.for.end:
10106 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10107 // CHECK10:       omp.loop.exit:
10108 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
10109 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10110 // CHECK10-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
10111 // CHECK10-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10112 // CHECK10:       .omp.final.then:
10113 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
10114 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10115 // CHECK10:       .omp.final.done:
10116 // CHECK10-NEXT:    ret void
10117 //
10118 //
10119 // CHECK10-LABEL: define {{[^@]+}}@.omp_outlined..15
10120 // CHECK10-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
10121 // CHECK10-NEXT:  entry:
10122 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10123 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10124 // CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
10125 // CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
10126 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10127 // CHECK10-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10128 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10129 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10130 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10131 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10132 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
10133 // CHECK10-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10134 // CHECK10-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10135 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10136 // CHECK10-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10137 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10138 // CHECK10-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
10139 // CHECK10-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10140 // CHECK10-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
10141 // CHECK10-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10142 // CHECK10-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
10143 // CHECK10-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
10144 // CHECK10-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
10145 // CHECK10-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10146 // CHECK10-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10147 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10148 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
10149 // CHECK10-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)
10150 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10151 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
10152 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10153 // CHECK10:       cond.true:
10154 // CHECK10-NEXT:    br label [[COND_END:%.*]]
10155 // CHECK10:       cond.false:
10156 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10157 // CHECK10-NEXT:    br label [[COND_END]]
10158 // CHECK10:       cond.end:
10159 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
10160 // CHECK10-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10161 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10162 // CHECK10-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
10163 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10164 // CHECK10:       omp.inner.for.cond:
10165 // CHECK10-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !63
10166 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !63
10167 // CHECK10-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10168 // CHECK10-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10169 // CHECK10:       omp.inner.for.body:
10170 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !63
10171 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
10172 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10173 // CHECK10-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !63
10174 // CHECK10-NEXT:    call void @_Z3fn3v(), !llvm.access.group !63
10175 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10176 // CHECK10:       omp.body.continue:
10177 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10178 // CHECK10:       omp.inner.for.inc:
10179 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !63
10180 // CHECK10-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
10181 // CHECK10-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !63
10182 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP64:![0-9]+]]
10183 // CHECK10:       omp.inner.for.end:
10184 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10185 // CHECK10:       omp.loop.exit:
10186 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
10187 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10188 // CHECK10-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
10189 // CHECK10-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10190 // CHECK10:       .omp.final.then:
10191 // CHECK10-NEXT:    store i32 100, i32* [[I]], align 4
10192 // CHECK10-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10193 // CHECK10:       .omp.final.done:
10194 // CHECK10-NEXT:    ret void
10195 //
10196 //
10197 // CHECK10-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
10198 // CHECK10-SAME: () #[[ATTR5:[0-9]+]] {
10199 // CHECK10-NEXT:  entry:
10200 // CHECK10-NEXT:    call void @__tgt_register_requires(i64 1)
10201 // CHECK10-NEXT:    ret void
10202 //
10203 //
10204 // CHECK11-LABEL: define {{[^@]+}}@_Z9gtid_testv
10205 // CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
10206 // CHECK11-NEXT:  entry:
10207 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10208 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
10209 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
10210 // 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)
10211 // CHECK11-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
10212 // CHECK11-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
10213 // CHECK11:       omp_offload.failed:
10214 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43() #[[ATTR2:[0-9]+]]
10215 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
10216 // CHECK11:       omp_offload.cont:
10217 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
10218 // 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)
10219 // CHECK11-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
10220 // CHECK11-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
10221 // CHECK11:       omp_offload.failed2:
10222 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48() #[[ATTR2]]
10223 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
10224 // CHECK11:       omp_offload.cont3:
10225 // CHECK11-NEXT:    ret void
10226 //
10227 //
10228 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43
10229 // CHECK11-SAME: () #[[ATTR1:[0-9]+]] {
10230 // CHECK11-NEXT:  entry:
10231 // 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*, ...)*))
10232 // CHECK11-NEXT:    ret void
10233 //
10234 //
10235 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
10236 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
10237 // CHECK11-NEXT:  entry:
10238 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10239 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10240 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10241 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10242 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10243 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10244 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10245 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10246 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
10247 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10248 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10249 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
10250 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
10251 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10252 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10253 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10254 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
10255 // 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)
10256 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10257 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
10258 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10259 // CHECK11:       cond.true:
10260 // CHECK11-NEXT:    br label [[COND_END:%.*]]
10261 // CHECK11:       cond.false:
10262 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10263 // CHECK11-NEXT:    br label [[COND_END]]
10264 // CHECK11:       cond.end:
10265 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
10266 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
10267 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10268 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
10269 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10270 // CHECK11:       omp.inner.for.cond:
10271 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
10272 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
10273 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
10274 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10275 // CHECK11:       omp.inner.for.body:
10276 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !15
10277 // CHECK11-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
10278 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
10279 // CHECK11-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
10280 // 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
10281 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10282 // CHECK11:       omp.inner.for.inc:
10283 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
10284 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !15
10285 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
10286 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
10287 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
10288 // CHECK11:       omp.inner.for.end:
10289 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10290 // CHECK11:       omp.loop.exit:
10291 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
10292 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10293 // CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
10294 // CHECK11-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10295 // CHECK11:       .omp.final.then:
10296 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
10297 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10298 // CHECK11:       .omp.final.done:
10299 // CHECK11-NEXT:    ret void
10300 //
10301 //
10302 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
10303 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
10304 // CHECK11-NEXT:  entry:
10305 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10306 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10307 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
10308 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
10309 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10310 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10311 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10312 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10313 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10314 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10315 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
10316 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10317 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10318 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10319 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10320 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10321 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
10322 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10323 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
10324 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10325 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
10326 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
10327 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
10328 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10329 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10330 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10331 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
10332 // 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)
10333 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10334 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
10335 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10336 // CHECK11:       cond.true:
10337 // CHECK11-NEXT:    br label [[COND_END:%.*]]
10338 // CHECK11:       cond.false:
10339 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10340 // CHECK11-NEXT:    br label [[COND_END]]
10341 // CHECK11:       cond.end:
10342 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
10343 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10344 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10345 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
10346 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10347 // CHECK11:       omp.inner.for.cond:
10348 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
10349 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
10350 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10351 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10352 // CHECK11:       omp.inner.for.body:
10353 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
10354 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
10355 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10356 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !19
10357 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10358 // CHECK11:       omp.body.continue:
10359 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10360 // CHECK11:       omp.inner.for.inc:
10361 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
10362 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
10363 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
10364 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
10365 // CHECK11:       omp.inner.for.end:
10366 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10367 // CHECK11:       omp.loop.exit:
10368 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
10369 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10370 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
10371 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10372 // CHECK11:       .omp.final.then:
10373 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
10374 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10375 // CHECK11:       .omp.final.done:
10376 // CHECK11-NEXT:    ret void
10377 //
10378 //
10379 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48
10380 // CHECK11-SAME: () #[[ATTR1]] {
10381 // CHECK11-NEXT:  entry:
10382 // 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*, ...)*))
10383 // CHECK11-NEXT:    ret void
10384 //
10385 //
10386 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..2
10387 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
10388 // CHECK11-NEXT:  entry:
10389 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10390 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10391 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10392 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10393 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10394 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10395 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10396 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10397 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
10398 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
10399 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10400 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10401 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
10402 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
10403 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10404 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10405 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10406 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
10407 // 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)
10408 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10409 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
10410 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10411 // CHECK11:       cond.true:
10412 // CHECK11-NEXT:    br label [[COND_END:%.*]]
10413 // CHECK11:       cond.false:
10414 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10415 // CHECK11-NEXT:    br label [[COND_END]]
10416 // CHECK11:       cond.end:
10417 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
10418 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
10419 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10420 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
10421 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10422 // CHECK11:       omp.inner.for.cond:
10423 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
10424 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
10425 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
10426 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10427 // CHECK11:       omp.inner.for.body:
10428 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
10429 // CHECK11-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
10430 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
10431 // CHECK11-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
10432 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !24
10433 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !24
10434 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !24
10435 // CHECK11-NEXT:    call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !24
10436 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !24
10437 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10438 // CHECK11:       omp.inner.for.inc:
10439 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
10440 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
10441 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
10442 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
10443 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
10444 // CHECK11:       omp.inner.for.end:
10445 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10446 // CHECK11:       omp.loop.exit:
10447 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
10448 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10449 // CHECK11-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
10450 // CHECK11-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10451 // CHECK11:       .omp.final.then:
10452 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
10453 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10454 // CHECK11:       .omp.final.done:
10455 // CHECK11-NEXT:    ret void
10456 //
10457 //
10458 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..3
10459 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
10460 // CHECK11-NEXT:  entry:
10461 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10462 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10463 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
10464 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
10465 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10466 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10467 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10468 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10469 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10470 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10471 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
10472 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10473 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10474 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10475 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10476 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10477 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
10478 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10479 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
10480 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10481 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
10482 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
10483 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
10484 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10485 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10486 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10487 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
10488 // 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)
10489 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10490 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
10491 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10492 // CHECK11:       cond.true:
10493 // CHECK11-NEXT:    br label [[COND_END:%.*]]
10494 // CHECK11:       cond.false:
10495 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10496 // CHECK11-NEXT:    br label [[COND_END]]
10497 // CHECK11:       cond.end:
10498 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
10499 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10500 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10501 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
10502 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10503 // CHECK11:       omp.inner.for.cond:
10504 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
10505 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
10506 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10507 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10508 // CHECK11:       omp.inner.for.body:
10509 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
10510 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
10511 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10512 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !27
10513 // CHECK11-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !27
10514 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10515 // CHECK11:       omp.body.continue:
10516 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10517 // CHECK11:       omp.inner.for.inc:
10518 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
10519 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
10520 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
10521 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
10522 // CHECK11:       omp.inner.for.end:
10523 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10524 // CHECK11:       omp.loop.exit:
10525 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
10526 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10527 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
10528 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10529 // CHECK11:       .omp.final.then:
10530 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
10531 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10532 // CHECK11:       .omp.final.done:
10533 // CHECK11-NEXT:    ret void
10534 //
10535 //
10536 // CHECK11-LABEL: define {{[^@]+}}@main
10537 // CHECK11-SAME: () #[[ATTR3:[0-9]+]] {
10538 // CHECK11-NEXT:  entry:
10539 // CHECK11-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
10540 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10541 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
10542 // CHECK11-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
10543 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
10544 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
10545 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
10546 // CHECK11-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
10547 // CHECK11-NEXT:    store i32 0, i32* [[RETVAL]], align 4
10548 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
10549 // 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)
10550 // CHECK11-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
10551 // CHECK11-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
10552 // CHECK11:       omp_offload.failed:
10553 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81() #[[ATTR2]]
10554 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
10555 // CHECK11:       omp_offload.cont:
10556 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
10557 // 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)
10558 // CHECK11-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
10559 // CHECK11-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
10560 // CHECK11:       omp_offload.failed2:
10561 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90() #[[ATTR2]]
10562 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
10563 // CHECK11:       omp_offload.cont3:
10564 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* @Arg, align 4
10565 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
10566 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
10567 // CHECK11-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
10568 // CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10569 // CHECK11-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
10570 // CHECK11-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
10571 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10572 // CHECK11-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
10573 // CHECK11-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
10574 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
10575 // CHECK11-NEXT:    store i8* null, i8** [[TMP10]], align 8
10576 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
10577 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
10578 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
10579 // 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)
10580 // CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
10581 // CHECK11-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
10582 // CHECK11:       omp_offload.failed5:
10583 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99(i64 [[TMP5]]) #[[ATTR2]]
10584 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
10585 // CHECK11:       omp_offload.cont6:
10586 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
10587 // CHECK11-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
10588 // CHECK11-NEXT:    ret i32 [[CALL]]
10589 //
10590 //
10591 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
10592 // CHECK11-SAME: () #[[ATTR1]] {
10593 // CHECK11-NEXT:  entry:
10594 // 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*, ...)*))
10595 // CHECK11-NEXT:    ret void
10596 //
10597 //
10598 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..4
10599 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
10600 // CHECK11-NEXT:  entry:
10601 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10602 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10603 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10604 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10605 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10606 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10607 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10608 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10609 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
10610 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10611 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10612 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
10613 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
10614 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10615 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10616 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10617 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
10618 // 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)
10619 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10620 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
10621 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10622 // CHECK11:       cond.true:
10623 // CHECK11-NEXT:    br label [[COND_END:%.*]]
10624 // CHECK11:       cond.false:
10625 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10626 // CHECK11-NEXT:    br label [[COND_END]]
10627 // CHECK11:       cond.end:
10628 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
10629 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
10630 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10631 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
10632 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10633 // CHECK11:       omp.inner.for.cond:
10634 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
10635 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
10636 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
10637 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10638 // CHECK11:       omp.inner.for.body:
10639 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
10640 // CHECK11-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
10641 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
10642 // CHECK11-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
10643 // 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
10644 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10645 // CHECK11:       omp.inner.for.inc:
10646 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
10647 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
10648 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
10649 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
10650 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
10651 // CHECK11:       omp.inner.for.end:
10652 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10653 // CHECK11:       omp.loop.exit:
10654 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
10655 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10656 // CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
10657 // CHECK11-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10658 // CHECK11:       .omp.final.then:
10659 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
10660 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10661 // CHECK11:       .omp.final.done:
10662 // CHECK11-NEXT:    ret void
10663 //
10664 //
10665 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..5
10666 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
10667 // CHECK11-NEXT:  entry:
10668 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10669 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10670 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
10671 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
10672 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10673 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10674 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10675 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10676 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10677 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10678 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
10679 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10680 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10681 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10682 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10683 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10684 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
10685 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10686 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
10687 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10688 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
10689 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
10690 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
10691 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10692 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10693 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10694 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
10695 // 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)
10696 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10697 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
10698 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10699 // CHECK11:       cond.true:
10700 // CHECK11-NEXT:    br label [[COND_END:%.*]]
10701 // CHECK11:       cond.false:
10702 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10703 // CHECK11-NEXT:    br label [[COND_END]]
10704 // CHECK11:       cond.end:
10705 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
10706 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10707 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10708 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
10709 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10710 // CHECK11:       omp.inner.for.cond:
10711 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
10712 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
10713 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10714 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10715 // CHECK11:       omp.inner.for.body:
10716 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
10717 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
10718 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10719 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
10720 // CHECK11-NEXT:    call void @_Z3fn4v(), !llvm.access.group !33
10721 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10722 // CHECK11:       omp.body.continue:
10723 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10724 // CHECK11:       omp.inner.for.inc:
10725 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
10726 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
10727 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
10728 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
10729 // CHECK11:       omp.inner.for.end:
10730 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10731 // CHECK11:       omp.loop.exit:
10732 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
10733 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10734 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
10735 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10736 // CHECK11:       .omp.final.then:
10737 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
10738 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10739 // CHECK11:       .omp.final.done:
10740 // CHECK11-NEXT:    ret void
10741 //
10742 //
10743 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90
10744 // CHECK11-SAME: () #[[ATTR1]] {
10745 // CHECK11-NEXT:  entry:
10746 // 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*, ...)*))
10747 // CHECK11-NEXT:    ret void
10748 //
10749 //
10750 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..6
10751 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
10752 // CHECK11-NEXT:  entry:
10753 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10754 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10755 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10756 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10757 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10758 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10759 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10760 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10761 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
10762 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
10763 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10764 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10765 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
10766 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
10767 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10768 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10769 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10770 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
10771 // 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)
10772 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10773 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
10774 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10775 // CHECK11:       cond.true:
10776 // CHECK11-NEXT:    br label [[COND_END:%.*]]
10777 // CHECK11:       cond.false:
10778 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10779 // CHECK11-NEXT:    br label [[COND_END]]
10780 // CHECK11:       cond.end:
10781 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
10782 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
10783 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10784 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
10785 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10786 // CHECK11:       omp.inner.for.cond:
10787 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10788 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10789 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
10790 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10791 // CHECK11:       omp.inner.for.body:
10792 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10793 // CHECK11-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
10794 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10795 // CHECK11-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
10796 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
10797 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10798 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
10799 // CHECK11-NEXT:    call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
10800 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
10801 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10802 // CHECK11:       omp.inner.for.inc:
10803 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10804 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
10805 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
10806 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
10807 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
10808 // CHECK11:       omp.inner.for.end:
10809 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10810 // CHECK11:       omp.loop.exit:
10811 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
10812 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10813 // CHECK11-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
10814 // CHECK11-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10815 // CHECK11:       .omp.final.then:
10816 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
10817 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10818 // CHECK11:       .omp.final.done:
10819 // CHECK11-NEXT:    ret void
10820 //
10821 //
10822 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..7
10823 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
10824 // CHECK11-NEXT:  entry:
10825 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10826 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10827 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
10828 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
10829 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10830 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10831 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
10832 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
10833 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10834 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10835 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
10836 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10837 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10838 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10839 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10840 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
10841 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
10842 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
10843 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
10844 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
10845 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
10846 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
10847 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
10848 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10849 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10850 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10851 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
10852 // 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)
10853 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10854 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
10855 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10856 // CHECK11:       cond.true:
10857 // CHECK11-NEXT:    br label [[COND_END:%.*]]
10858 // CHECK11:       cond.false:
10859 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10860 // CHECK11-NEXT:    br label [[COND_END]]
10861 // CHECK11:       cond.end:
10862 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
10863 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
10864 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
10865 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
10866 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10867 // CHECK11:       omp.inner.for.cond:
10868 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10869 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
10870 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
10871 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10872 // CHECK11:       omp.inner.for.body:
10873 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10874 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
10875 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
10876 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
10877 // CHECK11-NEXT:    call void @_Z3fn5v()
10878 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
10879 // CHECK11:       omp.body.continue:
10880 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10881 // CHECK11:       omp.inner.for.inc:
10882 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
10883 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
10884 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
10885 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP38:![0-9]+]]
10886 // CHECK11:       omp.inner.for.end:
10887 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
10888 // CHECK11:       omp.loop.exit:
10889 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
10890 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
10891 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
10892 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
10893 // CHECK11:       .omp.final.then:
10894 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
10895 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
10896 // CHECK11:       .omp.final.done:
10897 // CHECK11-NEXT:    ret void
10898 //
10899 //
10900 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99
10901 // CHECK11-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
10902 // CHECK11-NEXT:  entry:
10903 // CHECK11-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
10904 // CHECK11-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
10905 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
10906 // 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]])
10907 // CHECK11-NEXT:    ret void
10908 //
10909 //
10910 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..8
10911 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
10912 // CHECK11-NEXT:  entry:
10913 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
10914 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
10915 // CHECK11-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
10916 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
10917 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
10918 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
10919 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
10920 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
10921 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
10922 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
10923 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
10924 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
10925 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
10926 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__CASTED12:%.*]] = alloca i64, align 8
10927 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR18:%.*]] = alloca i32, align 4
10928 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
10929 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
10930 // CHECK11-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
10931 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
10932 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
10933 // CHECK11-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
10934 // CHECK11-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
10935 // CHECK11-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
10936 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
10937 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
10938 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
10939 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
10940 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
10941 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
10942 // 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)
10943 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10944 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
10945 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
10946 // CHECK11:       cond.true:
10947 // CHECK11-NEXT:    br label [[COND_END:%.*]]
10948 // CHECK11:       cond.false:
10949 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
10950 // CHECK11-NEXT:    br label [[COND_END]]
10951 // CHECK11:       cond.end:
10952 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
10953 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
10954 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
10955 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
10956 // CHECK11-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
10957 // CHECK11-NEXT:    [[TOBOOL1:%.*]] = trunc i8 [[TMP7]] to i1
10958 // CHECK11-NEXT:    br i1 [[TOBOOL1]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE7:%.*]]
10959 // CHECK11:       omp_if.then:
10960 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
10961 // CHECK11:       omp.inner.for.cond:
10962 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
10963 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !39
10964 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
10965 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
10966 // CHECK11:       omp.inner.for.body:
10967 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !39
10968 // CHECK11-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
10969 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !39
10970 // CHECK11-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
10971 // CHECK11-NEXT:    [[TMP14:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !39
10972 // CHECK11-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP14]] to i1
10973 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
10974 // CHECK11-NEXT:    [[FROMBOOL4:%.*]] = zext i1 [[TOBOOL3]] to i8
10975 // CHECK11-NEXT:    store i8 [[FROMBOOL4]], i8* [[CONV]], align 1, !llvm.access.group !39
10976 // CHECK11-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !39
10977 // CHECK11-NEXT:    [[TMP16:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !39
10978 // CHECK11-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP16]] to i1
10979 // CHECK11-NEXT:    br i1 [[TOBOOL5]], label [[OMP_IF_THEN6:%.*]], label [[OMP_IF_ELSE:%.*]]
10980 // CHECK11:       omp_if.then6:
10981 // 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
10982 // CHECK11-NEXT:    br label [[OMP_IF_END:%.*]]
10983 // CHECK11:       omp_if.else:
10984 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]]), !llvm.access.group !39
10985 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !39
10986 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !39
10987 // CHECK11-NEXT:    call void @.omp_outlined..9(i32* [[TMP17]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]]) #[[ATTR2]], !llvm.access.group !39
10988 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]]), !llvm.access.group !39
10989 // CHECK11-NEXT:    br label [[OMP_IF_END]]
10990 // CHECK11:       omp_if.end:
10991 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
10992 // CHECK11:       omp.inner.for.inc:
10993 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
10994 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !39
10995 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
10996 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
10997 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
10998 // CHECK11:       omp.inner.for.end:
10999 // CHECK11-NEXT:    br label [[OMP_IF_END23:%.*]]
11000 // CHECK11:       omp_if.else7:
11001 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
11002 // CHECK11:       omp.inner.for.cond8:
11003 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11004 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11005 // CHECK11-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
11006 // CHECK11-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END22:%.*]]
11007 // CHECK11:       omp.inner.for.body10:
11008 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11009 // CHECK11-NEXT:    [[TMP23:%.*]] = zext i32 [[TMP22]] to i64
11010 // CHECK11-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11011 // CHECK11-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP24]] to i64
11012 // CHECK11-NEXT:    [[TMP26:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
11013 // CHECK11-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP26]] to i1
11014 // CHECK11-NEXT:    [[CONV13:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED12]] to i8*
11015 // CHECK11-NEXT:    [[FROMBOOL14:%.*]] = zext i1 [[TOBOOL11]] to i8
11016 // CHECK11-NEXT:    store i8 [[FROMBOOL14]], i8* [[CONV13]], align 1
11017 // CHECK11-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED12]], align 8
11018 // CHECK11-NEXT:    [[TMP28:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
11019 // CHECK11-NEXT:    [[TOBOOL15:%.*]] = trunc i8 [[TMP28]] to i1
11020 // CHECK11-NEXT:    br i1 [[TOBOOL15]], label [[OMP_IF_THEN16:%.*]], label [[OMP_IF_ELSE17:%.*]]
11021 // CHECK11:       omp_if.then16:
11022 // 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]])
11023 // CHECK11-NEXT:    br label [[OMP_IF_END19:%.*]]
11024 // CHECK11:       omp_if.else17:
11025 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
11026 // CHECK11-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11027 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR18]], align 4
11028 // CHECK11-NEXT:    call void @.omp_outlined..10(i32* [[TMP29]], i32* [[DOTBOUND_ZERO_ADDR18]], i64 [[TMP23]], i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR2]]
11029 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
11030 // CHECK11-NEXT:    br label [[OMP_IF_END19]]
11031 // CHECK11:       omp_if.end19:
11032 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC20:%.*]]
11033 // CHECK11:       omp.inner.for.inc20:
11034 // CHECK11-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11035 // CHECK11-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
11036 // CHECK11-NEXT:    [[ADD21:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
11037 // CHECK11-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
11038 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP42:![0-9]+]]
11039 // CHECK11:       omp.inner.for.end22:
11040 // CHECK11-NEXT:    br label [[OMP_IF_END23]]
11041 // CHECK11:       omp_if.end23:
11042 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11043 // CHECK11:       omp.loop.exit:
11044 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
11045 // CHECK11-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11046 // CHECK11-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
11047 // CHECK11-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11048 // CHECK11:       .omp.final.then:
11049 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
11050 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11051 // CHECK11:       .omp.final.done:
11052 // CHECK11-NEXT:    ret void
11053 //
11054 //
11055 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..9
11056 // 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]] {
11057 // CHECK11-NEXT:  entry:
11058 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11059 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11060 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
11061 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
11062 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
11063 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11064 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11065 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11066 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11067 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11068 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11069 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
11070 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11071 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11072 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11073 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11074 // CHECK11-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
11075 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
11076 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11077 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
11078 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11079 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP0]] to i32
11080 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11081 // CHECK11-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
11082 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
11083 // CHECK11-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
11084 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11085 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11086 // CHECK11-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
11087 // CHECK11-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
11088 // CHECK11-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
11089 // CHECK11:       omp_if.then:
11090 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11091 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
11092 // 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)
11093 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11094 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
11095 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11096 // CHECK11:       cond.true:
11097 // CHECK11-NEXT:    br label [[COND_END:%.*]]
11098 // CHECK11:       cond.false:
11099 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11100 // CHECK11-NEXT:    br label [[COND_END]]
11101 // CHECK11:       cond.end:
11102 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
11103 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11104 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11105 // CHECK11-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
11106 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11107 // CHECK11:       omp.inner.for.cond:
11108 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
11109 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !43
11110 // CHECK11-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
11111 // CHECK11-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11112 // CHECK11:       omp.inner.for.body:
11113 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
11114 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
11115 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11116 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !43
11117 // CHECK11-NEXT:    call void @_Z3fn6v(), !llvm.access.group !43
11118 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11119 // CHECK11:       omp.body.continue:
11120 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11121 // CHECK11:       omp.inner.for.inc:
11122 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
11123 // CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
11124 // CHECK11-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
11125 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
11126 // CHECK11:       omp.inner.for.end:
11127 // CHECK11-NEXT:    br label [[OMP_IF_END:%.*]]
11128 // CHECK11:       omp_if.else:
11129 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11130 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
11131 // 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)
11132 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11133 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP14]], 99
11134 // CHECK11-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
11135 // CHECK11:       cond.true6:
11136 // CHECK11-NEXT:    br label [[COND_END8:%.*]]
11137 // CHECK11:       cond.false7:
11138 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11139 // CHECK11-NEXT:    br label [[COND_END8]]
11140 // CHECK11:       cond.end8:
11141 // CHECK11-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP15]], [[COND_FALSE7]] ]
11142 // CHECK11-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_UB]], align 4
11143 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11144 // CHECK11-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
11145 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND10:%.*]]
11146 // CHECK11:       omp.inner.for.cond10:
11147 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11148 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11149 // CHECK11-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
11150 // CHECK11-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY12:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
11151 // CHECK11:       omp.inner.for.body12:
11152 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11153 // CHECK11-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP19]], 1
11154 // CHECK11-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
11155 // CHECK11-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
11156 // CHECK11-NEXT:    call void @_Z3fn6v()
11157 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
11158 // CHECK11:       omp.body.continue15:
11159 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
11160 // CHECK11:       omp.inner.for.inc16:
11161 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11162 // CHECK11-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
11163 // CHECK11-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
11164 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND10]], !llvm.loop [[LOOP46:![0-9]+]]
11165 // CHECK11:       omp.inner.for.end18:
11166 // CHECK11-NEXT:    br label [[OMP_IF_END]]
11167 // CHECK11:       omp_if.end:
11168 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11169 // CHECK11:       omp.loop.exit:
11170 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11171 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
11172 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
11173 // CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11174 // CHECK11-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
11175 // CHECK11-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11176 // CHECK11:       .omp.final.then:
11177 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
11178 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11179 // CHECK11:       .omp.final.done:
11180 // CHECK11-NEXT:    ret void
11181 //
11182 //
11183 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..10
11184 // 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]] {
11185 // CHECK11-NEXT:  entry:
11186 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11187 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11188 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
11189 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
11190 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
11191 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11192 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11193 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11194 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11195 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11196 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11197 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
11198 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11199 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11200 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11201 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11202 // CHECK11-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
11203 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
11204 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11205 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
11206 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11207 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP0]] to i32
11208 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11209 // CHECK11-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
11210 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
11211 // CHECK11-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
11212 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11213 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11214 // CHECK11-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
11215 // CHECK11-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
11216 // CHECK11-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
11217 // CHECK11:       omp_if.then:
11218 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11219 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
11220 // 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)
11221 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11222 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
11223 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11224 // CHECK11:       cond.true:
11225 // CHECK11-NEXT:    br label [[COND_END:%.*]]
11226 // CHECK11:       cond.false:
11227 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11228 // CHECK11-NEXT:    br label [[COND_END]]
11229 // CHECK11:       cond.end:
11230 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
11231 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11232 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11233 // CHECK11-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
11234 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11235 // CHECK11:       omp.inner.for.cond:
11236 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
11237 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !47
11238 // CHECK11-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
11239 // CHECK11-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11240 // CHECK11:       omp.inner.for.body:
11241 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
11242 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
11243 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11244 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !47
11245 // CHECK11-NEXT:    call void @_Z3fn6v(), !llvm.access.group !47
11246 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11247 // CHECK11:       omp.body.continue:
11248 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11249 // CHECK11:       omp.inner.for.inc:
11250 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
11251 // CHECK11-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
11252 // CHECK11-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
11253 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
11254 // CHECK11:       omp.inner.for.end:
11255 // CHECK11-NEXT:    br label [[OMP_IF_END:%.*]]
11256 // CHECK11:       omp_if.else:
11257 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11258 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
11259 // 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)
11260 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11261 // CHECK11-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP14]], 99
11262 // CHECK11-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
11263 // CHECK11:       cond.true6:
11264 // CHECK11-NEXT:    br label [[COND_END8:%.*]]
11265 // CHECK11:       cond.false7:
11266 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11267 // CHECK11-NEXT:    br label [[COND_END8]]
11268 // CHECK11:       cond.end8:
11269 // CHECK11-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP15]], [[COND_FALSE7]] ]
11270 // CHECK11-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_UB]], align 4
11271 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11272 // CHECK11-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
11273 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND10:%.*]]
11274 // CHECK11:       omp.inner.for.cond10:
11275 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11276 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11277 // CHECK11-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
11278 // CHECK11-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY12:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
11279 // CHECK11:       omp.inner.for.body12:
11280 // CHECK11-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11281 // CHECK11-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP19]], 1
11282 // CHECK11-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
11283 // CHECK11-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
11284 // CHECK11-NEXT:    call void @_Z3fn6v()
11285 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
11286 // CHECK11:       omp.body.continue15:
11287 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
11288 // CHECK11:       omp.inner.for.inc16:
11289 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11290 // CHECK11-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
11291 // CHECK11-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
11292 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND10]], !llvm.loop [[LOOP50:![0-9]+]]
11293 // CHECK11:       omp.inner.for.end18:
11294 // CHECK11-NEXT:    br label [[OMP_IF_END]]
11295 // CHECK11:       omp_if.end:
11296 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11297 // CHECK11:       omp.loop.exit:
11298 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11299 // CHECK11-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
11300 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
11301 // CHECK11-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11302 // CHECK11-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
11303 // CHECK11-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11304 // CHECK11:       .omp.final.then:
11305 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
11306 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11307 // CHECK11:       .omp.final.done:
11308 // CHECK11-NEXT:    ret void
11309 //
11310 //
11311 // CHECK11-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
11312 // CHECK11-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
11313 // CHECK11-NEXT:  entry:
11314 // CHECK11-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
11315 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11316 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
11317 // CHECK11-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
11318 // CHECK11-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
11319 // CHECK11-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
11320 // CHECK11-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
11321 // CHECK11-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
11322 // CHECK11-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
11323 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
11324 // 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)
11325 // CHECK11-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
11326 // CHECK11-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
11327 // CHECK11:       omp_offload.failed:
11328 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59() #[[ATTR2]]
11329 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT]]
11330 // CHECK11:       omp_offload.cont:
11331 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
11332 // 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)
11333 // CHECK11-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
11334 // CHECK11-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
11335 // CHECK11:       omp_offload.failed2:
11336 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65() #[[ATTR2]]
11337 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
11338 // CHECK11:       omp_offload.cont3:
11339 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
11340 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
11341 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
11342 // CHECK11-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
11343 // CHECK11-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11344 // CHECK11-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
11345 // CHECK11-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
11346 // CHECK11-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11347 // CHECK11-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
11348 // CHECK11-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
11349 // CHECK11-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
11350 // CHECK11-NEXT:    store i8* null, i8** [[TMP10]], align 8
11351 // CHECK11-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
11352 // CHECK11-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
11353 // CHECK11-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
11354 // 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)
11355 // CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
11356 // CHECK11-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
11357 // CHECK11:       omp_offload.failed5:
11358 // CHECK11-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71(i64 [[TMP5]]) #[[ATTR2]]
11359 // CHECK11-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
11360 // CHECK11:       omp_offload.cont6:
11361 // CHECK11-NEXT:    ret i32 0
11362 //
11363 //
11364 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59
11365 // CHECK11-SAME: () #[[ATTR1]] {
11366 // CHECK11-NEXT:  entry:
11367 // 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*, ...)*))
11368 // CHECK11-NEXT:    ret void
11369 //
11370 //
11371 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..11
11372 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
11373 // CHECK11-NEXT:  entry:
11374 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11375 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11376 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11377 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11378 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
11379 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
11380 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11381 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11382 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
11383 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11384 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11385 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
11386 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
11387 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11388 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11389 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11390 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
11391 // 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)
11392 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11393 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
11394 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11395 // CHECK11:       cond.true:
11396 // CHECK11-NEXT:    br label [[COND_END:%.*]]
11397 // CHECK11:       cond.false:
11398 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11399 // CHECK11-NEXT:    br label [[COND_END]]
11400 // CHECK11:       cond.end:
11401 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
11402 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
11403 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11404 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
11405 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11406 // CHECK11:       omp.inner.for.cond:
11407 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
11408 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !51
11409 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
11410 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11411 // CHECK11:       omp.inner.for.body:
11412 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !51
11413 // CHECK11-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
11414 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !51
11415 // CHECK11-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
11416 // 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
11417 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11418 // CHECK11:       omp.inner.for.inc:
11419 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
11420 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !51
11421 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
11422 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
11423 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP52:![0-9]+]]
11424 // CHECK11:       omp.inner.for.end:
11425 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11426 // CHECK11:       omp.loop.exit:
11427 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
11428 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11429 // CHECK11-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
11430 // CHECK11-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11431 // CHECK11:       .omp.final.then:
11432 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
11433 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11434 // CHECK11:       .omp.final.done:
11435 // CHECK11-NEXT:    ret void
11436 //
11437 //
11438 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..12
11439 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
11440 // CHECK11-NEXT:  entry:
11441 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11442 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11443 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
11444 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
11445 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11446 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11447 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11448 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11449 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11450 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11451 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
11452 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11453 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11454 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11455 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11456 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11457 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
11458 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11459 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
11460 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11461 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
11462 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
11463 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
11464 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11465 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11466 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11467 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
11468 // 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)
11469 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11470 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
11471 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11472 // CHECK11:       cond.true:
11473 // CHECK11-NEXT:    br label [[COND_END:%.*]]
11474 // CHECK11:       cond.false:
11475 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11476 // CHECK11-NEXT:    br label [[COND_END]]
11477 // CHECK11:       cond.end:
11478 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
11479 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11480 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11481 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
11482 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11483 // CHECK11:       omp.inner.for.cond:
11484 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
11485 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !54
11486 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
11487 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11488 // CHECK11:       omp.inner.for.body:
11489 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
11490 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
11491 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11492 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !54
11493 // CHECK11-NEXT:    call void @_Z3fn1v(), !llvm.access.group !54
11494 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11495 // CHECK11:       omp.body.continue:
11496 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11497 // CHECK11:       omp.inner.for.inc:
11498 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
11499 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
11500 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
11501 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP55:![0-9]+]]
11502 // CHECK11:       omp.inner.for.end:
11503 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11504 // CHECK11:       omp.loop.exit:
11505 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
11506 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11507 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
11508 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11509 // CHECK11:       .omp.final.then:
11510 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
11511 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11512 // CHECK11:       .omp.final.done:
11513 // CHECK11-NEXT:    ret void
11514 //
11515 //
11516 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65
11517 // CHECK11-SAME: () #[[ATTR1]] {
11518 // CHECK11-NEXT:  entry:
11519 // 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*, ...)*))
11520 // CHECK11-NEXT:    ret void
11521 //
11522 //
11523 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..13
11524 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
11525 // CHECK11-NEXT:  entry:
11526 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11527 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11528 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11529 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11530 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
11531 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
11532 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11533 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11534 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
11535 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
11536 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11537 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11538 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
11539 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
11540 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11541 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11542 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11543 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
11544 // 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)
11545 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11546 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
11547 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11548 // CHECK11:       cond.true:
11549 // CHECK11-NEXT:    br label [[COND_END:%.*]]
11550 // CHECK11:       cond.false:
11551 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11552 // CHECK11-NEXT:    br label [[COND_END]]
11553 // CHECK11:       cond.end:
11554 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
11555 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
11556 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11557 // CHECK11-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
11558 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11559 // CHECK11:       omp.inner.for.cond:
11560 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11561 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11562 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
11563 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11564 // CHECK11:       omp.inner.for.body:
11565 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11566 // CHECK11-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
11567 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11568 // CHECK11-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
11569 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
11570 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11571 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
11572 // CHECK11-NEXT:    call void @.omp_outlined..14(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
11573 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
11574 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11575 // CHECK11:       omp.inner.for.inc:
11576 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11577 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
11578 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
11579 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
11580 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP57:![0-9]+]]
11581 // CHECK11:       omp.inner.for.end:
11582 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11583 // CHECK11:       omp.loop.exit:
11584 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
11585 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11586 // CHECK11-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
11587 // CHECK11-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11588 // CHECK11:       .omp.final.then:
11589 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
11590 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11591 // CHECK11:       .omp.final.done:
11592 // CHECK11-NEXT:    ret void
11593 //
11594 //
11595 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..14
11596 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
11597 // CHECK11-NEXT:  entry:
11598 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11599 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11600 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
11601 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
11602 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11603 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11604 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11605 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11606 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11607 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11608 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
11609 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11610 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11611 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11612 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11613 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11614 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
11615 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11616 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
11617 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11618 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
11619 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
11620 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
11621 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11622 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11623 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11624 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
11625 // 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)
11626 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11627 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
11628 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11629 // CHECK11:       cond.true:
11630 // CHECK11-NEXT:    br label [[COND_END:%.*]]
11631 // CHECK11:       cond.false:
11632 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11633 // CHECK11-NEXT:    br label [[COND_END]]
11634 // CHECK11:       cond.end:
11635 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
11636 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11637 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11638 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
11639 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11640 // CHECK11:       omp.inner.for.cond:
11641 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11642 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11643 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
11644 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11645 // CHECK11:       omp.inner.for.body:
11646 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11647 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
11648 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11649 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
11650 // CHECK11-NEXT:    call void @_Z3fn2v()
11651 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11652 // CHECK11:       omp.body.continue:
11653 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11654 // CHECK11:       omp.inner.for.inc:
11655 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
11656 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
11657 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
11658 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP58:![0-9]+]]
11659 // CHECK11:       omp.inner.for.end:
11660 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11661 // CHECK11:       omp.loop.exit:
11662 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
11663 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11664 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
11665 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11666 // CHECK11:       .omp.final.then:
11667 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
11668 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11669 // CHECK11:       .omp.final.done:
11670 // CHECK11-NEXT:    ret void
11671 //
11672 //
11673 // CHECK11-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71
11674 // CHECK11-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
11675 // CHECK11-NEXT:  entry:
11676 // CHECK11-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
11677 // CHECK11-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
11678 // CHECK11-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
11679 // 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]])
11680 // CHECK11-NEXT:    ret void
11681 //
11682 //
11683 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..15
11684 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
11685 // CHECK11-NEXT:  entry:
11686 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11687 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11688 // CHECK11-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
11689 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11690 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11691 // CHECK11-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
11692 // CHECK11-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
11693 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11694 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11695 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
11696 // CHECK11-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
11697 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11698 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11699 // CHECK11-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
11700 // CHECK11-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
11701 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
11702 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
11703 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11704 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11705 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11706 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
11707 // 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)
11708 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11709 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
11710 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11711 // CHECK11:       cond.true:
11712 // CHECK11-NEXT:    br label [[COND_END:%.*]]
11713 // CHECK11:       cond.false:
11714 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11715 // CHECK11-NEXT:    br label [[COND_END]]
11716 // CHECK11:       cond.end:
11717 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
11718 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
11719 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11720 // CHECK11-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
11721 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11722 // CHECK11:       omp.inner.for.cond:
11723 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
11724 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
11725 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
11726 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11727 // CHECK11:       omp.inner.for.body:
11728 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !59
11729 // CHECK11-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
11730 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
11731 // CHECK11-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
11732 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !59
11733 // CHECK11-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
11734 // CHECK11-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
11735 // CHECK11:       omp_if.then:
11736 // 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
11737 // CHECK11-NEXT:    br label [[OMP_IF_END:%.*]]
11738 // CHECK11:       omp_if.else:
11739 // CHECK11-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !59
11740 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !59
11741 // CHECK11-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !59
11742 // CHECK11-NEXT:    call void @.omp_outlined..16(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !59
11743 // CHECK11-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !59
11744 // CHECK11-NEXT:    br label [[OMP_IF_END]]
11745 // CHECK11:       omp_if.end:
11746 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11747 // CHECK11:       omp.inner.for.inc:
11748 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
11749 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !59
11750 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
11751 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
11752 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP60:![0-9]+]]
11753 // CHECK11:       omp.inner.for.end:
11754 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11755 // CHECK11:       omp.loop.exit:
11756 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
11757 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11758 // CHECK11-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
11759 // CHECK11-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11760 // CHECK11:       .omp.final.then:
11761 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
11762 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11763 // CHECK11:       .omp.final.done:
11764 // CHECK11-NEXT:    ret void
11765 //
11766 //
11767 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..16
11768 // CHECK11-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
11769 // CHECK11-NEXT:  entry:
11770 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11771 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11772 // CHECK11-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
11773 // CHECK11-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
11774 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11775 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11776 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11777 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11778 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11779 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11780 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
11781 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11782 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11783 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11784 // CHECK11-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11785 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11786 // CHECK11-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
11787 // CHECK11-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11788 // CHECK11-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
11789 // CHECK11-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11790 // CHECK11-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
11791 // CHECK11-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
11792 // CHECK11-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
11793 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11794 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11795 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11796 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
11797 // 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)
11798 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11799 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
11800 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11801 // CHECK11:       cond.true:
11802 // CHECK11-NEXT:    br label [[COND_END:%.*]]
11803 // CHECK11:       cond.false:
11804 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11805 // CHECK11-NEXT:    br label [[COND_END]]
11806 // CHECK11:       cond.end:
11807 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
11808 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11809 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11810 // CHECK11-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
11811 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11812 // CHECK11:       omp.inner.for.cond:
11813 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
11814 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !62
11815 // CHECK11-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
11816 // CHECK11-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11817 // CHECK11:       omp.inner.for.body:
11818 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
11819 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
11820 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
11821 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !62
11822 // CHECK11-NEXT:    call void @_Z3fn3v(), !llvm.access.group !62
11823 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
11824 // CHECK11:       omp.body.continue:
11825 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11826 // CHECK11:       omp.inner.for.inc:
11827 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
11828 // CHECK11-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
11829 // CHECK11-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
11830 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP63:![0-9]+]]
11831 // CHECK11:       omp.inner.for.end:
11832 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11833 // CHECK11:       omp.loop.exit:
11834 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
11835 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11836 // CHECK11-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
11837 // CHECK11-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11838 // CHECK11:       .omp.final.then:
11839 // CHECK11-NEXT:    store i32 100, i32* [[I]], align 4
11840 // CHECK11-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11841 // CHECK11:       .omp.final.done:
11842 // CHECK11-NEXT:    ret void
11843 //
11844 //
11845 // CHECK11-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
11846 // CHECK11-SAME: () #[[ATTR5:[0-9]+]] {
11847 // CHECK11-NEXT:  entry:
11848 // CHECK11-NEXT:    call void @__tgt_register_requires(i64 1)
11849 // CHECK11-NEXT:    ret void
11850 //
11851 //
11852 // CHECK12-LABEL: define {{[^@]+}}@_Z9gtid_testv
11853 // CHECK12-SAME: () #[[ATTR0:[0-9]+]] {
11854 // CHECK12-NEXT:  entry:
11855 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11856 // CHECK12-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
11857 // CHECK12-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
11858 // CHECK12-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)
11859 // CHECK12-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
11860 // CHECK12-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
11861 // CHECK12:       omp_offload.failed:
11862 // CHECK12-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43() #[[ATTR2:[0-9]+]]
11863 // CHECK12-NEXT:    br label [[OMP_OFFLOAD_CONT]]
11864 // CHECK12:       omp_offload.cont:
11865 // CHECK12-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
11866 // CHECK12-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)
11867 // CHECK12-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
11868 // CHECK12-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
11869 // CHECK12:       omp_offload.failed2:
11870 // CHECK12-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48() #[[ATTR2]]
11871 // CHECK12-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
11872 // CHECK12:       omp_offload.cont3:
11873 // CHECK12-NEXT:    ret void
11874 //
11875 //
11876 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l43
11877 // CHECK12-SAME: () #[[ATTR1:[0-9]+]] {
11878 // CHECK12-NEXT:  entry:
11879 // CHECK12-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*, ...)*))
11880 // CHECK12-NEXT:    ret void
11881 //
11882 //
11883 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined.
11884 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
11885 // CHECK12-NEXT:  entry:
11886 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11887 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11888 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11889 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11890 // CHECK12-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
11891 // CHECK12-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
11892 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11893 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11894 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
11895 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11896 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11897 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
11898 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
11899 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11900 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11901 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11902 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
11903 // CHECK12-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)
11904 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11905 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
11906 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11907 // CHECK12:       cond.true:
11908 // CHECK12-NEXT:    br label [[COND_END:%.*]]
11909 // CHECK12:       cond.false:
11910 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
11911 // CHECK12-NEXT:    br label [[COND_END]]
11912 // CHECK12:       cond.end:
11913 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
11914 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
11915 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
11916 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
11917 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11918 // CHECK12:       omp.inner.for.cond:
11919 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
11920 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
11921 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
11922 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
11923 // CHECK12:       omp.inner.for.body:
11924 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !15
11925 // CHECK12-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
11926 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !15
11927 // CHECK12-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
11928 // CHECK12-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
11929 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
11930 // CHECK12:       omp.inner.for.inc:
11931 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
11932 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !15
11933 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
11934 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !15
11935 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP16:![0-9]+]]
11936 // CHECK12:       omp.inner.for.end:
11937 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
11938 // CHECK12:       omp.loop.exit:
11939 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
11940 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
11941 // CHECK12-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
11942 // CHECK12-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
11943 // CHECK12:       .omp.final.then:
11944 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
11945 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
11946 // CHECK12:       .omp.final.done:
11947 // CHECK12-NEXT:    ret void
11948 //
11949 //
11950 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..1
11951 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
11952 // CHECK12-NEXT:  entry:
11953 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
11954 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
11955 // CHECK12-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
11956 // CHECK12-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
11957 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
11958 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
11959 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
11960 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
11961 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
11962 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
11963 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
11964 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
11965 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
11966 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11967 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11968 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
11969 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
11970 // CHECK12-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
11971 // CHECK12-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
11972 // CHECK12-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
11973 // CHECK12-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
11974 // CHECK12-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
11975 // CHECK12-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
11976 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
11977 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
11978 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
11979 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
11980 // CHECK12-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)
11981 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11982 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
11983 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
11984 // CHECK12:       cond.true:
11985 // CHECK12-NEXT:    br label [[COND_END:%.*]]
11986 // CHECK12:       cond.false:
11987 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
11988 // CHECK12-NEXT:    br label [[COND_END]]
11989 // CHECK12:       cond.end:
11990 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
11991 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
11992 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
11993 // CHECK12-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
11994 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
11995 // CHECK12:       omp.inner.for.cond:
11996 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
11997 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !19
11998 // CHECK12-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
11999 // CHECK12-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12000 // CHECK12:       omp.inner.for.body:
12001 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
12002 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
12003 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12004 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !19
12005 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
12006 // CHECK12:       omp.body.continue:
12007 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12008 // CHECK12:       omp.inner.for.inc:
12009 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
12010 // CHECK12-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
12011 // CHECK12-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !19
12012 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP20:![0-9]+]]
12013 // CHECK12:       omp.inner.for.end:
12014 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12015 // CHECK12:       omp.loop.exit:
12016 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
12017 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
12018 // CHECK12-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
12019 // CHECK12-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
12020 // CHECK12:       .omp.final.then:
12021 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
12022 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
12023 // CHECK12:       .omp.final.done:
12024 // CHECK12-NEXT:    ret void
12025 //
12026 //
12027 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l48
12028 // CHECK12-SAME: () #[[ATTR1]] {
12029 // CHECK12-NEXT:  entry:
12030 // CHECK12-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*, ...)*))
12031 // CHECK12-NEXT:    ret void
12032 //
12033 //
12034 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..2
12035 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
12036 // CHECK12-NEXT:  entry:
12037 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12038 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12039 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12040 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12041 // CHECK12-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
12042 // CHECK12-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
12043 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12044 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12045 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
12046 // CHECK12-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
12047 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12048 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12049 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
12050 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
12051 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12052 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12053 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12054 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
12055 // CHECK12-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)
12056 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12057 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
12058 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12059 // CHECK12:       cond.true:
12060 // CHECK12-NEXT:    br label [[COND_END:%.*]]
12061 // CHECK12:       cond.false:
12062 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12063 // CHECK12-NEXT:    br label [[COND_END]]
12064 // CHECK12:       cond.end:
12065 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
12066 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
12067 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
12068 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
12069 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12070 // CHECK12:       omp.inner.for.cond:
12071 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
12072 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
12073 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
12074 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12075 // CHECK12:       omp.inner.for.body:
12076 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !24
12077 // CHECK12-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
12078 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !24
12079 // CHECK12-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
12080 // CHECK12-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !24
12081 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !24
12082 // CHECK12-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !24
12083 // CHECK12-NEXT:    call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]], !llvm.access.group !24
12084 // CHECK12-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]]), !llvm.access.group !24
12085 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12086 // CHECK12:       omp.inner.for.inc:
12087 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
12088 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !24
12089 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
12090 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !24
12091 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
12092 // CHECK12:       omp.inner.for.end:
12093 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12094 // CHECK12:       omp.loop.exit:
12095 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
12096 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
12097 // CHECK12-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
12098 // CHECK12-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
12099 // CHECK12:       .omp.final.then:
12100 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
12101 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
12102 // CHECK12:       .omp.final.done:
12103 // CHECK12-NEXT:    ret void
12104 //
12105 //
12106 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..3
12107 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
12108 // CHECK12-NEXT:  entry:
12109 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12110 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12111 // CHECK12-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
12112 // CHECK12-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
12113 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12114 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12115 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
12116 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
12117 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12118 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12119 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
12120 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12121 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12122 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12123 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12124 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
12125 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
12126 // CHECK12-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12127 // CHECK12-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
12128 // CHECK12-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12129 // CHECK12-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
12130 // CHECK12-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
12131 // CHECK12-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
12132 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12133 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12134 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12135 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
12136 // CHECK12-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)
12137 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12138 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
12139 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12140 // CHECK12:       cond.true:
12141 // CHECK12-NEXT:    br label [[COND_END:%.*]]
12142 // CHECK12:       cond.false:
12143 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12144 // CHECK12-NEXT:    br label [[COND_END]]
12145 // CHECK12:       cond.end:
12146 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
12147 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
12148 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12149 // CHECK12-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
12150 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12151 // CHECK12:       omp.inner.for.cond:
12152 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
12153 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !27
12154 // CHECK12-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
12155 // CHECK12-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12156 // CHECK12:       omp.inner.for.body:
12157 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
12158 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
12159 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12160 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !27
12161 // CHECK12-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !27
12162 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
12163 // CHECK12:       omp.body.continue:
12164 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12165 // CHECK12:       omp.inner.for.inc:
12166 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
12167 // CHECK12-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
12168 // CHECK12-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !27
12169 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]]
12170 // CHECK12:       omp.inner.for.end:
12171 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12172 // CHECK12:       omp.loop.exit:
12173 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
12174 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
12175 // CHECK12-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
12176 // CHECK12-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
12177 // CHECK12:       .omp.final.then:
12178 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
12179 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
12180 // CHECK12:       .omp.final.done:
12181 // CHECK12-NEXT:    ret void
12182 //
12183 //
12184 // CHECK12-LABEL: define {{[^@]+}}@main
12185 // CHECK12-SAME: () #[[ATTR3:[0-9]+]] {
12186 // CHECK12-NEXT:  entry:
12187 // CHECK12-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
12188 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12189 // CHECK12-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
12190 // CHECK12-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
12191 // CHECK12-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
12192 // CHECK12-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
12193 // CHECK12-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
12194 // CHECK12-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
12195 // CHECK12-NEXT:    store i32 0, i32* [[RETVAL]], align 4
12196 // CHECK12-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
12197 // CHECK12-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)
12198 // CHECK12-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
12199 // CHECK12-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
12200 // CHECK12:       omp_offload.failed:
12201 // CHECK12-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81() #[[ATTR2]]
12202 // CHECK12-NEXT:    br label [[OMP_OFFLOAD_CONT]]
12203 // CHECK12:       omp_offload.cont:
12204 // CHECK12-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
12205 // CHECK12-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)
12206 // CHECK12-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
12207 // CHECK12-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
12208 // CHECK12:       omp_offload.failed2:
12209 // CHECK12-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90() #[[ATTR2]]
12210 // CHECK12-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
12211 // CHECK12:       omp_offload.cont3:
12212 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* @Arg, align 4
12213 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
12214 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
12215 // CHECK12-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
12216 // CHECK12-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12217 // CHECK12-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
12218 // CHECK12-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
12219 // CHECK12-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12220 // CHECK12-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
12221 // CHECK12-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
12222 // CHECK12-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
12223 // CHECK12-NEXT:    store i8* null, i8** [[TMP10]], align 8
12224 // CHECK12-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12225 // CHECK12-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12226 // CHECK12-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
12227 // CHECK12-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)
12228 // CHECK12-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
12229 // CHECK12-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
12230 // CHECK12:       omp_offload.failed5:
12231 // CHECK12-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99(i64 [[TMP5]]) #[[ATTR2]]
12232 // CHECK12-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
12233 // CHECK12:       omp_offload.cont6:
12234 // CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
12235 // CHECK12-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
12236 // CHECK12-NEXT:    ret i32 [[CALL]]
12237 //
12238 //
12239 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l81
12240 // CHECK12-SAME: () #[[ATTR1]] {
12241 // CHECK12-NEXT:  entry:
12242 // CHECK12-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*, ...)*))
12243 // CHECK12-NEXT:    ret void
12244 //
12245 //
12246 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..4
12247 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
12248 // CHECK12-NEXT:  entry:
12249 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12250 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12251 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12252 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12253 // CHECK12-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
12254 // CHECK12-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
12255 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12256 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12257 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
12258 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12259 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12260 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
12261 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
12262 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12263 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12264 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12265 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
12266 // CHECK12-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)
12267 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12268 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
12269 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12270 // CHECK12:       cond.true:
12271 // CHECK12-NEXT:    br label [[COND_END:%.*]]
12272 // CHECK12:       cond.false:
12273 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12274 // CHECK12-NEXT:    br label [[COND_END]]
12275 // CHECK12:       cond.end:
12276 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
12277 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
12278 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
12279 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
12280 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12281 // CHECK12:       omp.inner.for.cond:
12282 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
12283 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
12284 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
12285 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12286 // CHECK12:       omp.inner.for.body:
12287 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !30
12288 // CHECK12-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
12289 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !30
12290 // CHECK12-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
12291 // CHECK12-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
12292 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12293 // CHECK12:       omp.inner.for.inc:
12294 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
12295 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !30
12296 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
12297 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !30
12298 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP31:![0-9]+]]
12299 // CHECK12:       omp.inner.for.end:
12300 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12301 // CHECK12:       omp.loop.exit:
12302 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
12303 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
12304 // CHECK12-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
12305 // CHECK12-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
12306 // CHECK12:       .omp.final.then:
12307 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
12308 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
12309 // CHECK12:       .omp.final.done:
12310 // CHECK12-NEXT:    ret void
12311 //
12312 //
12313 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..5
12314 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
12315 // CHECK12-NEXT:  entry:
12316 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12317 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12318 // CHECK12-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
12319 // CHECK12-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
12320 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12321 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12322 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
12323 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
12324 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12325 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12326 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
12327 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12328 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12329 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12330 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12331 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
12332 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
12333 // CHECK12-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12334 // CHECK12-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
12335 // CHECK12-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12336 // CHECK12-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
12337 // CHECK12-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
12338 // CHECK12-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
12339 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12340 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12341 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12342 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
12343 // CHECK12-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)
12344 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12345 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
12346 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12347 // CHECK12:       cond.true:
12348 // CHECK12-NEXT:    br label [[COND_END:%.*]]
12349 // CHECK12:       cond.false:
12350 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12351 // CHECK12-NEXT:    br label [[COND_END]]
12352 // CHECK12:       cond.end:
12353 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
12354 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
12355 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12356 // CHECK12-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
12357 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12358 // CHECK12:       omp.inner.for.cond:
12359 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
12360 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !33
12361 // CHECK12-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
12362 // CHECK12-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12363 // CHECK12:       omp.inner.for.body:
12364 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
12365 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
12366 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12367 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !33
12368 // CHECK12-NEXT:    call void @_Z3fn4v(), !llvm.access.group !33
12369 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
12370 // CHECK12:       omp.body.continue:
12371 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12372 // CHECK12:       omp.inner.for.inc:
12373 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
12374 // CHECK12-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
12375 // CHECK12-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !33
12376 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP34:![0-9]+]]
12377 // CHECK12:       omp.inner.for.end:
12378 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12379 // CHECK12:       omp.loop.exit:
12380 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
12381 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
12382 // CHECK12-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
12383 // CHECK12-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
12384 // CHECK12:       .omp.final.then:
12385 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
12386 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
12387 // CHECK12:       .omp.final.done:
12388 // CHECK12-NEXT:    ret void
12389 //
12390 //
12391 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l90
12392 // CHECK12-SAME: () #[[ATTR1]] {
12393 // CHECK12-NEXT:  entry:
12394 // CHECK12-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*, ...)*))
12395 // CHECK12-NEXT:    ret void
12396 //
12397 //
12398 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..6
12399 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
12400 // CHECK12-NEXT:  entry:
12401 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12402 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12403 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12404 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12405 // CHECK12-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
12406 // CHECK12-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
12407 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12408 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12409 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
12410 // CHECK12-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
12411 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12412 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12413 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
12414 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
12415 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12416 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12417 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12418 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
12419 // CHECK12-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)
12420 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12421 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
12422 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12423 // CHECK12:       cond.true:
12424 // CHECK12-NEXT:    br label [[COND_END:%.*]]
12425 // CHECK12:       cond.false:
12426 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12427 // CHECK12-NEXT:    br label [[COND_END]]
12428 // CHECK12:       cond.end:
12429 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
12430 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
12431 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
12432 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
12433 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12434 // CHECK12:       omp.inner.for.cond:
12435 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12436 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12437 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
12438 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12439 // CHECK12:       omp.inner.for.body:
12440 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
12441 // CHECK12-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
12442 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12443 // CHECK12-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
12444 // CHECK12-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
12445 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12446 // CHECK12-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
12447 // CHECK12-NEXT:    call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
12448 // CHECK12-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
12449 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12450 // CHECK12:       omp.inner.for.inc:
12451 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12452 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
12453 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
12454 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
12455 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
12456 // CHECK12:       omp.inner.for.end:
12457 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12458 // CHECK12:       omp.loop.exit:
12459 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
12460 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
12461 // CHECK12-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
12462 // CHECK12-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
12463 // CHECK12:       .omp.final.then:
12464 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
12465 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
12466 // CHECK12:       .omp.final.done:
12467 // CHECK12-NEXT:    ret void
12468 //
12469 //
12470 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..7
12471 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
12472 // CHECK12-NEXT:  entry:
12473 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12474 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12475 // CHECK12-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
12476 // CHECK12-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
12477 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12478 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12479 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
12480 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
12481 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12482 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12483 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
12484 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12485 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12486 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12487 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12488 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
12489 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
12490 // CHECK12-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12491 // CHECK12-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
12492 // CHECK12-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12493 // CHECK12-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
12494 // CHECK12-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
12495 // CHECK12-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
12496 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12497 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12498 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12499 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
12500 // CHECK12-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)
12501 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12502 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
12503 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12504 // CHECK12:       cond.true:
12505 // CHECK12-NEXT:    br label [[COND_END:%.*]]
12506 // CHECK12:       cond.false:
12507 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12508 // CHECK12-NEXT:    br label [[COND_END]]
12509 // CHECK12:       cond.end:
12510 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
12511 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
12512 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12513 // CHECK12-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
12514 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12515 // CHECK12:       omp.inner.for.cond:
12516 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12517 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12518 // CHECK12-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
12519 // CHECK12-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12520 // CHECK12:       omp.inner.for.body:
12521 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12522 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
12523 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12524 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
12525 // CHECK12-NEXT:    call void @_Z3fn5v()
12526 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
12527 // CHECK12:       omp.body.continue:
12528 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12529 // CHECK12:       omp.inner.for.inc:
12530 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12531 // CHECK12-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
12532 // CHECK12-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
12533 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP38:![0-9]+]]
12534 // CHECK12:       omp.inner.for.end:
12535 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12536 // CHECK12:       omp.loop.exit:
12537 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
12538 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
12539 // CHECK12-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
12540 // CHECK12-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
12541 // CHECK12:       .omp.final.then:
12542 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
12543 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
12544 // CHECK12:       .omp.final.done:
12545 // CHECK12-NEXT:    ret void
12546 //
12547 //
12548 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l99
12549 // CHECK12-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
12550 // CHECK12-NEXT:  entry:
12551 // CHECK12-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
12552 // CHECK12-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
12553 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
12554 // CHECK12-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]])
12555 // CHECK12-NEXT:    ret void
12556 //
12557 //
12558 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..8
12559 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
12560 // CHECK12-NEXT:  entry:
12561 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12562 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12563 // CHECK12-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
12564 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
12565 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12566 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12567 // CHECK12-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
12568 // CHECK12-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
12569 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12570 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12571 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
12572 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__CASTED:%.*]] = alloca i64, align 8
12573 // CHECK12-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
12574 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__CASTED12:%.*]] = alloca i64, align 8
12575 // CHECK12-NEXT:    [[DOTBOUND_ZERO_ADDR18:%.*]] = alloca i32, align 4
12576 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12577 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12578 // CHECK12-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
12579 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
12580 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
12581 // CHECK12-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP1]], 0
12582 // CHECK12-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
12583 // CHECK12-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
12584 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
12585 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
12586 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12587 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12588 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12589 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
12590 // CHECK12-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)
12591 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12592 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
12593 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12594 // CHECK12:       cond.true:
12595 // CHECK12-NEXT:    br label [[COND_END:%.*]]
12596 // CHECK12:       cond.false:
12597 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12598 // CHECK12-NEXT:    br label [[COND_END]]
12599 // CHECK12:       cond.end:
12600 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
12601 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
12602 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
12603 // CHECK12-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
12604 // CHECK12-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
12605 // CHECK12-NEXT:    [[TOBOOL1:%.*]] = trunc i8 [[TMP7]] to i1
12606 // CHECK12-NEXT:    br i1 [[TOBOOL1]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE7:%.*]]
12607 // CHECK12:       omp_if.then:
12608 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12609 // CHECK12:       omp.inner.for.cond:
12610 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
12611 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !39
12612 // CHECK12-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
12613 // CHECK12-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12614 // CHECK12:       omp.inner.for.body:
12615 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !39
12616 // CHECK12-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
12617 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !39
12618 // CHECK12-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP12]] to i64
12619 // CHECK12-NEXT:    [[TMP14:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !39
12620 // CHECK12-NEXT:    [[TOBOOL3:%.*]] = trunc i8 [[TMP14]] to i1
12621 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED]] to i8*
12622 // CHECK12-NEXT:    [[FROMBOOL4:%.*]] = zext i1 [[TOBOOL3]] to i8
12623 // CHECK12-NEXT:    store i8 [[FROMBOOL4]], i8* [[CONV]], align 1, !llvm.access.group !39
12624 // CHECK12-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED]], align 8, !llvm.access.group !39
12625 // CHECK12-NEXT:    [[TMP16:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !39
12626 // CHECK12-NEXT:    [[TOBOOL5:%.*]] = trunc i8 [[TMP16]] to i1
12627 // CHECK12-NEXT:    br i1 [[TOBOOL5]], label [[OMP_IF_THEN6:%.*]], label [[OMP_IF_ELSE:%.*]]
12628 // CHECK12:       omp_if.then6:
12629 // CHECK12-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
12630 // CHECK12-NEXT:    br label [[OMP_IF_END:%.*]]
12631 // CHECK12:       omp_if.else:
12632 // CHECK12-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]]), !llvm.access.group !39
12633 // CHECK12-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !39
12634 // CHECK12-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !39
12635 // CHECK12-NEXT:    call void @.omp_outlined..9(i32* [[TMP17]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP11]], i64 [[TMP13]], i64 [[TMP15]]) #[[ATTR2]], !llvm.access.group !39
12636 // CHECK12-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]]), !llvm.access.group !39
12637 // CHECK12-NEXT:    br label [[OMP_IF_END]]
12638 // CHECK12:       omp_if.end:
12639 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12640 // CHECK12:       omp.inner.for.inc:
12641 // CHECK12-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
12642 // CHECK12-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !39
12643 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP18]], [[TMP19]]
12644 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !39
12645 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP40:![0-9]+]]
12646 // CHECK12:       omp.inner.for.end:
12647 // CHECK12-NEXT:    br label [[OMP_IF_END23:%.*]]
12648 // CHECK12:       omp_if.else7:
12649 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND8:%.*]]
12650 // CHECK12:       omp.inner.for.cond8:
12651 // CHECK12-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12652 // CHECK12-NEXT:    [[TMP21:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12653 // CHECK12-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP20]], [[TMP21]]
12654 // CHECK12-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY10:%.*]], label [[OMP_INNER_FOR_END22:%.*]]
12655 // CHECK12:       omp.inner.for.body10:
12656 // CHECK12-NEXT:    [[TMP22:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
12657 // CHECK12-NEXT:    [[TMP23:%.*]] = zext i32 [[TMP22]] to i64
12658 // CHECK12-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
12659 // CHECK12-NEXT:    [[TMP25:%.*]] = zext i32 [[TMP24]] to i64
12660 // CHECK12-NEXT:    [[TMP26:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
12661 // CHECK12-NEXT:    [[TOBOOL11:%.*]] = trunc i8 [[TMP26]] to i1
12662 // CHECK12-NEXT:    [[CONV13:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__CASTED12]] to i8*
12663 // CHECK12-NEXT:    [[FROMBOOL14:%.*]] = zext i1 [[TOBOOL11]] to i8
12664 // CHECK12-NEXT:    store i8 [[FROMBOOL14]], i8* [[CONV13]], align 1
12665 // CHECK12-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR__CASTED12]], align 8
12666 // CHECK12-NEXT:    [[TMP28:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
12667 // CHECK12-NEXT:    [[TOBOOL15:%.*]] = trunc i8 [[TMP28]] to i1
12668 // CHECK12-NEXT:    br i1 [[TOBOOL15]], label [[OMP_IF_THEN16:%.*]], label [[OMP_IF_ELSE17:%.*]]
12669 // CHECK12:       omp_if.then16:
12670 // CHECK12-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]])
12671 // CHECK12-NEXT:    br label [[OMP_IF_END19:%.*]]
12672 // CHECK12:       omp_if.else17:
12673 // CHECK12-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
12674 // CHECK12-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12675 // CHECK12-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR18]], align 4
12676 // CHECK12-NEXT:    call void @.omp_outlined..10(i32* [[TMP29]], i32* [[DOTBOUND_ZERO_ADDR18]], i64 [[TMP23]], i64 [[TMP25]], i64 [[TMP27]]) #[[ATTR2]]
12677 // CHECK12-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP3]])
12678 // CHECK12-NEXT:    br label [[OMP_IF_END19]]
12679 // CHECK12:       omp_if.end19:
12680 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC20:%.*]]
12681 // CHECK12:       omp.inner.for.inc20:
12682 // CHECK12-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12683 // CHECK12-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
12684 // CHECK12-NEXT:    [[ADD21:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
12685 // CHECK12-NEXT:    store i32 [[ADD21]], i32* [[DOTOMP_IV]], align 4
12686 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND8]], !llvm.loop [[LOOP42:![0-9]+]]
12687 // CHECK12:       omp.inner.for.end22:
12688 // CHECK12-NEXT:    br label [[OMP_IF_END23]]
12689 // CHECK12:       omp_if.end23:
12690 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12691 // CHECK12:       omp.loop.exit:
12692 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
12693 // CHECK12-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
12694 // CHECK12-NEXT:    [[TMP33:%.*]] = icmp ne i32 [[TMP32]], 0
12695 // CHECK12-NEXT:    br i1 [[TMP33]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
12696 // CHECK12:       .omp.final.then:
12697 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
12698 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
12699 // CHECK12:       .omp.final.done:
12700 // CHECK12-NEXT:    ret void
12701 //
12702 //
12703 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..9
12704 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
12705 // CHECK12-NEXT:  entry:
12706 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12707 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12708 // CHECK12-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
12709 // CHECK12-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
12710 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
12711 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12712 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12713 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
12714 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
12715 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12716 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12717 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
12718 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12719 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12720 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12721 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12722 // CHECK12-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
12723 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
12724 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
12725 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
12726 // CHECK12-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12727 // CHECK12-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP0]] to i32
12728 // CHECK12-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12729 // CHECK12-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
12730 // CHECK12-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
12731 // CHECK12-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
12732 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12733 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12734 // CHECK12-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
12735 // CHECK12-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
12736 // CHECK12-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
12737 // CHECK12:       omp_if.then:
12738 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12739 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
12740 // CHECK12-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)
12741 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12742 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
12743 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12744 // CHECK12:       cond.true:
12745 // CHECK12-NEXT:    br label [[COND_END:%.*]]
12746 // CHECK12:       cond.false:
12747 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12748 // CHECK12-NEXT:    br label [[COND_END]]
12749 // CHECK12:       cond.end:
12750 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
12751 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
12752 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12753 // CHECK12-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
12754 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12755 // CHECK12:       omp.inner.for.cond:
12756 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
12757 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !43
12758 // CHECK12-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
12759 // CHECK12-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12760 // CHECK12:       omp.inner.for.body:
12761 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
12762 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
12763 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12764 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !43
12765 // CHECK12-NEXT:    call void @_Z3fn6v(), !llvm.access.group !43
12766 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
12767 // CHECK12:       omp.body.continue:
12768 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12769 // CHECK12:       omp.inner.for.inc:
12770 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
12771 // CHECK12-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
12772 // CHECK12-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !43
12773 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP44:![0-9]+]]
12774 // CHECK12:       omp.inner.for.end:
12775 // CHECK12-NEXT:    br label [[OMP_IF_END:%.*]]
12776 // CHECK12:       omp_if.else:
12777 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12778 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
12779 // CHECK12-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)
12780 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12781 // CHECK12-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP14]], 99
12782 // CHECK12-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
12783 // CHECK12:       cond.true6:
12784 // CHECK12-NEXT:    br label [[COND_END8:%.*]]
12785 // CHECK12:       cond.false7:
12786 // CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12787 // CHECK12-NEXT:    br label [[COND_END8]]
12788 // CHECK12:       cond.end8:
12789 // CHECK12-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP15]], [[COND_FALSE7]] ]
12790 // CHECK12-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_UB]], align 4
12791 // CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12792 // CHECK12-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
12793 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND10:%.*]]
12794 // CHECK12:       omp.inner.for.cond10:
12795 // CHECK12-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12796 // CHECK12-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12797 // CHECK12-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
12798 // CHECK12-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY12:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
12799 // CHECK12:       omp.inner.for.body12:
12800 // CHECK12-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12801 // CHECK12-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP19]], 1
12802 // CHECK12-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
12803 // CHECK12-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
12804 // CHECK12-NEXT:    call void @_Z3fn6v()
12805 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
12806 // CHECK12:       omp.body.continue15:
12807 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
12808 // CHECK12:       omp.inner.for.inc16:
12809 // CHECK12-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12810 // CHECK12-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
12811 // CHECK12-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
12812 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND10]], !llvm.loop [[LOOP46:![0-9]+]]
12813 // CHECK12:       omp.inner.for.end18:
12814 // CHECK12-NEXT:    br label [[OMP_IF_END]]
12815 // CHECK12:       omp_if.end:
12816 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12817 // CHECK12:       omp.loop.exit:
12818 // CHECK12-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12819 // CHECK12-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
12820 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
12821 // CHECK12-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
12822 // CHECK12-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
12823 // CHECK12-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
12824 // CHECK12:       .omp.final.then:
12825 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
12826 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
12827 // CHECK12:       .omp.final.done:
12828 // CHECK12-NEXT:    ret void
12829 //
12830 //
12831 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..10
12832 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[DOTCAPTURE_EXPR_:%.*]]) #[[ATTR1]] {
12833 // CHECK12-NEXT:  entry:
12834 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
12835 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
12836 // CHECK12-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
12837 // CHECK12-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
12838 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR__ADDR:%.*]] = alloca i64, align 8
12839 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
12840 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12841 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
12842 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
12843 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
12844 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
12845 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
12846 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
12847 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
12848 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12849 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12850 // CHECK12-NEXT:    store i64 [[DOTCAPTURE_EXPR_]], i64* [[DOTCAPTURE_EXPR__ADDR]], align 8
12851 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i64* [[DOTCAPTURE_EXPR__ADDR]] to i8*
12852 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
12853 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
12854 // CHECK12-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
12855 // CHECK12-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP0]] to i32
12856 // CHECK12-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
12857 // CHECK12-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
12858 // CHECK12-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_LB]], align 4
12859 // CHECK12-NEXT:    store i32 [[CONV2]], i32* [[DOTOMP_UB]], align 4
12860 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
12861 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
12862 // CHECK12-NEXT:    [[TMP2:%.*]] = load i8, i8* [[CONV]], align 1
12863 // CHECK12-NEXT:    [[TOBOOL:%.*]] = trunc i8 [[TMP2]] to i1
12864 // CHECK12-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
12865 // CHECK12:       omp_if.then:
12866 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12867 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[TMP3]], align 4
12868 // CHECK12-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)
12869 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12870 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP5]], 99
12871 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
12872 // CHECK12:       cond.true:
12873 // CHECK12-NEXT:    br label [[COND_END:%.*]]
12874 // CHECK12:       cond.false:
12875 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12876 // CHECK12-NEXT:    br label [[COND_END]]
12877 // CHECK12:       cond.end:
12878 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP6]], [[COND_FALSE]] ]
12879 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
12880 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12881 // CHECK12-NEXT:    store i32 [[TMP7]], i32* [[DOTOMP_IV]], align 4
12882 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
12883 // CHECK12:       omp.inner.for.cond:
12884 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
12885 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !47
12886 // CHECK12-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP8]], [[TMP9]]
12887 // CHECK12-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
12888 // CHECK12:       omp.inner.for.body:
12889 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
12890 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP10]], 1
12891 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
12892 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !47
12893 // CHECK12-NEXT:    call void @_Z3fn6v(), !llvm.access.group !47
12894 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
12895 // CHECK12:       omp.body.continue:
12896 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
12897 // CHECK12:       omp.inner.for.inc:
12898 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
12899 // CHECK12-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP11]], 1
12900 // CHECK12-NEXT:    store i32 [[ADD4]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !47
12901 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP48:![0-9]+]]
12902 // CHECK12:       omp.inner.for.end:
12903 // CHECK12-NEXT:    br label [[OMP_IF_END:%.*]]
12904 // CHECK12:       omp_if.else:
12905 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12906 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
12907 // CHECK12-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)
12908 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12909 // CHECK12-NEXT:    [[CMP5:%.*]] = icmp sgt i32 [[TMP14]], 99
12910 // CHECK12-NEXT:    br i1 [[CMP5]], label [[COND_TRUE6:%.*]], label [[COND_FALSE7:%.*]]
12911 // CHECK12:       cond.true6:
12912 // CHECK12-NEXT:    br label [[COND_END8:%.*]]
12913 // CHECK12:       cond.false7:
12914 // CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12915 // CHECK12-NEXT:    br label [[COND_END8]]
12916 // CHECK12:       cond.end8:
12917 // CHECK12-NEXT:    [[COND9:%.*]] = phi i32 [ 99, [[COND_TRUE6]] ], [ [[TMP15]], [[COND_FALSE7]] ]
12918 // CHECK12-NEXT:    store i32 [[COND9]], i32* [[DOTOMP_UB]], align 4
12919 // CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
12920 // CHECK12-NEXT:    store i32 [[TMP16]], i32* [[DOTOMP_IV]], align 4
12921 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND10:%.*]]
12922 // CHECK12:       omp.inner.for.cond10:
12923 // CHECK12-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12924 // CHECK12-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
12925 // CHECK12-NEXT:    [[CMP11:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
12926 // CHECK12-NEXT:    br i1 [[CMP11]], label [[OMP_INNER_FOR_BODY12:%.*]], label [[OMP_INNER_FOR_END18:%.*]]
12927 // CHECK12:       omp.inner.for.body12:
12928 // CHECK12-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12929 // CHECK12-NEXT:    [[MUL13:%.*]] = mul nsw i32 [[TMP19]], 1
12930 // CHECK12-NEXT:    [[ADD14:%.*]] = add nsw i32 0, [[MUL13]]
12931 // CHECK12-NEXT:    store i32 [[ADD14]], i32* [[I]], align 4
12932 // CHECK12-NEXT:    call void @_Z3fn6v()
12933 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE15:%.*]]
12934 // CHECK12:       omp.body.continue15:
12935 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC16:%.*]]
12936 // CHECK12:       omp.inner.for.inc16:
12937 // CHECK12-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
12938 // CHECK12-NEXT:    [[ADD17:%.*]] = add nsw i32 [[TMP20]], 1
12939 // CHECK12-NEXT:    store i32 [[ADD17]], i32* [[DOTOMP_IV]], align 4
12940 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND10]], !llvm.loop [[LOOP50:![0-9]+]]
12941 // CHECK12:       omp.inner.for.end18:
12942 // CHECK12-NEXT:    br label [[OMP_IF_END]]
12943 // CHECK12:       omp_if.end:
12944 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
12945 // CHECK12:       omp.loop.exit:
12946 // CHECK12-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
12947 // CHECK12-NEXT:    [[TMP22:%.*]] = load i32, i32* [[TMP21]], align 4
12948 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP22]])
12949 // CHECK12-NEXT:    [[TMP23:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
12950 // CHECK12-NEXT:    [[TMP24:%.*]] = icmp ne i32 [[TMP23]], 0
12951 // CHECK12-NEXT:    br i1 [[TMP24]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
12952 // CHECK12:       .omp.final.then:
12953 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
12954 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
12955 // CHECK12:       .omp.final.done:
12956 // CHECK12-NEXT:    ret void
12957 //
12958 //
12959 // CHECK12-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
12960 // CHECK12-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
12961 // CHECK12-NEXT:  entry:
12962 // CHECK12-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
12963 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
12964 // CHECK12-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
12965 // CHECK12-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
12966 // CHECK12-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
12967 // CHECK12-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
12968 // CHECK12-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
12969 // CHECK12-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
12970 // CHECK12-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
12971 // CHECK12-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
12972 // CHECK12-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)
12973 // CHECK12-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
12974 // CHECK12-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
12975 // CHECK12:       omp_offload.failed:
12976 // CHECK12-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59() #[[ATTR2]]
12977 // CHECK12-NEXT:    br label [[OMP_OFFLOAD_CONT]]
12978 // CHECK12:       omp_offload.cont:
12979 // CHECK12-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
12980 // CHECK12-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)
12981 // CHECK12-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
12982 // CHECK12-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
12983 // CHECK12:       omp_offload.failed2:
12984 // CHECK12-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65() #[[ATTR2]]
12985 // CHECK12-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
12986 // CHECK12:       omp_offload.cont3:
12987 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
12988 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
12989 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
12990 // CHECK12-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
12991 // CHECK12-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
12992 // CHECK12-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
12993 // CHECK12-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
12994 // CHECK12-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
12995 // CHECK12-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
12996 // CHECK12-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
12997 // CHECK12-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
12998 // CHECK12-NEXT:    store i8* null, i8** [[TMP10]], align 8
12999 // CHECK12-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
13000 // CHECK12-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
13001 // CHECK12-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
13002 // CHECK12-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)
13003 // CHECK12-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
13004 // CHECK12-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
13005 // CHECK12:       omp_offload.failed5:
13006 // CHECK12-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71(i64 [[TMP5]]) #[[ATTR2]]
13007 // CHECK12-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
13008 // CHECK12:       omp_offload.cont6:
13009 // CHECK12-NEXT:    ret i32 0
13010 //
13011 //
13012 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l59
13013 // CHECK12-SAME: () #[[ATTR1]] {
13014 // CHECK12-NEXT:  entry:
13015 // CHECK12-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*, ...)*))
13016 // CHECK12-NEXT:    ret void
13017 //
13018 //
13019 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..11
13020 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
13021 // CHECK12-NEXT:  entry:
13022 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13023 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13024 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13025 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13026 // CHECK12-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
13027 // CHECK12-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
13028 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13029 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13030 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
13031 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13032 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13033 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
13034 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
13035 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13036 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13037 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
13038 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
13039 // CHECK12-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)
13040 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13041 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
13042 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13043 // CHECK12:       cond.true:
13044 // CHECK12-NEXT:    br label [[COND_END:%.*]]
13045 // CHECK12:       cond.false:
13046 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13047 // CHECK12-NEXT:    br label [[COND_END]]
13048 // CHECK12:       cond.end:
13049 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
13050 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
13051 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13052 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
13053 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13054 // CHECK12:       omp.inner.for.cond:
13055 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
13056 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !51
13057 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
13058 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13059 // CHECK12:       omp.inner.for.body:
13060 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !51
13061 // CHECK12-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
13062 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !51
13063 // CHECK12-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
13064 // CHECK12-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
13065 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13066 // CHECK12:       omp.inner.for.inc:
13067 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
13068 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !51
13069 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
13070 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !51
13071 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP52:![0-9]+]]
13072 // CHECK12:       omp.inner.for.end:
13073 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
13074 // CHECK12:       omp.loop.exit:
13075 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
13076 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
13077 // CHECK12-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
13078 // CHECK12-NEXT:    br i1 [[TMP14]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
13079 // CHECK12:       .omp.final.then:
13080 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
13081 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
13082 // CHECK12:       .omp.final.done:
13083 // CHECK12-NEXT:    ret void
13084 //
13085 //
13086 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..12
13087 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
13088 // CHECK12-NEXT:  entry:
13089 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13090 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13091 // CHECK12-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
13092 // CHECK12-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
13093 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13094 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13095 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
13096 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
13097 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13098 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13099 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
13100 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13101 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13102 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
13103 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
13104 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
13105 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
13106 // CHECK12-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
13107 // CHECK12-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
13108 // CHECK12-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
13109 // CHECK12-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
13110 // CHECK12-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
13111 // CHECK12-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
13112 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13113 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13114 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
13115 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
13116 // CHECK12-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)
13117 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13118 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
13119 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13120 // CHECK12:       cond.true:
13121 // CHECK12-NEXT:    br label [[COND_END:%.*]]
13122 // CHECK12:       cond.false:
13123 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13124 // CHECK12-NEXT:    br label [[COND_END]]
13125 // CHECK12:       cond.end:
13126 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
13127 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
13128 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13129 // CHECK12-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
13130 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13131 // CHECK12:       omp.inner.for.cond:
13132 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
13133 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !54
13134 // CHECK12-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
13135 // CHECK12-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13136 // CHECK12:       omp.inner.for.body:
13137 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
13138 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
13139 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13140 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !54
13141 // CHECK12-NEXT:    call void @_Z3fn1v(), !llvm.access.group !54
13142 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13143 // CHECK12:       omp.body.continue:
13144 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13145 // CHECK12:       omp.inner.for.inc:
13146 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
13147 // CHECK12-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
13148 // CHECK12-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !54
13149 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP55:![0-9]+]]
13150 // CHECK12:       omp.inner.for.end:
13151 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
13152 // CHECK12:       omp.loop.exit:
13153 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
13154 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
13155 // CHECK12-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
13156 // CHECK12-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
13157 // CHECK12:       .omp.final.then:
13158 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
13159 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
13160 // CHECK12:       .omp.final.done:
13161 // CHECK12-NEXT:    ret void
13162 //
13163 //
13164 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l65
13165 // CHECK12-SAME: () #[[ATTR1]] {
13166 // CHECK12-NEXT:  entry:
13167 // CHECK12-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*, ...)*))
13168 // CHECK12-NEXT:    ret void
13169 //
13170 //
13171 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..13
13172 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
13173 // CHECK12-NEXT:  entry:
13174 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13175 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13176 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13177 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13178 // CHECK12-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
13179 // CHECK12-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
13180 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13181 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13182 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
13183 // CHECK12-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
13184 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13185 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13186 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
13187 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
13188 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13189 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13190 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
13191 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
13192 // CHECK12-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)
13193 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13194 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
13195 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13196 // CHECK12:       cond.true:
13197 // CHECK12-NEXT:    br label [[COND_END:%.*]]
13198 // CHECK12:       cond.false:
13199 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13200 // CHECK12-NEXT:    br label [[COND_END]]
13201 // CHECK12:       cond.end:
13202 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
13203 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
13204 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13205 // CHECK12-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
13206 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13207 // CHECK12:       omp.inner.for.cond:
13208 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13209 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13210 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
13211 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13212 // CHECK12:       omp.inner.for.body:
13213 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13214 // CHECK12-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
13215 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13216 // CHECK12-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
13217 // CHECK12-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
13218 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
13219 // CHECK12-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
13220 // CHECK12-NEXT:    call void @.omp_outlined..14(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
13221 // CHECK12-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
13222 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13223 // CHECK12:       omp.inner.for.inc:
13224 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13225 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
13226 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
13227 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
13228 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP57:![0-9]+]]
13229 // CHECK12:       omp.inner.for.end:
13230 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
13231 // CHECK12:       omp.loop.exit:
13232 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
13233 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
13234 // CHECK12-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
13235 // CHECK12-NEXT:    br i1 [[TMP15]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
13236 // CHECK12:       .omp.final.then:
13237 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
13238 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
13239 // CHECK12:       .omp.final.done:
13240 // CHECK12-NEXT:    ret void
13241 //
13242 //
13243 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..14
13244 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
13245 // CHECK12-NEXT:  entry:
13246 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13247 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13248 // CHECK12-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
13249 // CHECK12-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
13250 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13251 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13252 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
13253 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
13254 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13255 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13256 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
13257 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13258 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13259 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
13260 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
13261 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
13262 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
13263 // CHECK12-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
13264 // CHECK12-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
13265 // CHECK12-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
13266 // CHECK12-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
13267 // CHECK12-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
13268 // CHECK12-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
13269 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13270 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13271 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
13272 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
13273 // CHECK12-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)
13274 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13275 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
13276 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13277 // CHECK12:       cond.true:
13278 // CHECK12-NEXT:    br label [[COND_END:%.*]]
13279 // CHECK12:       cond.false:
13280 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13281 // CHECK12-NEXT:    br label [[COND_END]]
13282 // CHECK12:       cond.end:
13283 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
13284 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
13285 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13286 // CHECK12-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
13287 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13288 // CHECK12:       omp.inner.for.cond:
13289 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13290 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13291 // CHECK12-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
13292 // CHECK12-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13293 // CHECK12:       omp.inner.for.body:
13294 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13295 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
13296 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13297 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
13298 // CHECK12-NEXT:    call void @_Z3fn2v()
13299 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13300 // CHECK12:       omp.body.continue:
13301 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13302 // CHECK12:       omp.inner.for.inc:
13303 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
13304 // CHECK12-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
13305 // CHECK12-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
13306 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP58:![0-9]+]]
13307 // CHECK12:       omp.inner.for.end:
13308 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
13309 // CHECK12:       omp.loop.exit:
13310 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
13311 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
13312 // CHECK12-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
13313 // CHECK12-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
13314 // CHECK12:       .omp.final.then:
13315 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
13316 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
13317 // CHECK12:       .omp.final.done:
13318 // CHECK12-NEXT:    ret void
13319 //
13320 //
13321 // CHECK12-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l71
13322 // CHECK12-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
13323 // CHECK12-NEXT:  entry:
13324 // CHECK12-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
13325 // CHECK12-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
13326 // CHECK12-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
13327 // CHECK12-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]])
13328 // CHECK12-NEXT:    ret void
13329 //
13330 //
13331 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..15
13332 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
13333 // CHECK12-NEXT:  entry:
13334 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13335 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13336 // CHECK12-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
13337 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13338 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13339 // CHECK12-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
13340 // CHECK12-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
13341 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13342 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13343 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
13344 // CHECK12-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
13345 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13346 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13347 // CHECK12-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
13348 // CHECK12-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
13349 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
13350 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
13351 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13352 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13353 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
13354 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
13355 // CHECK12-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)
13356 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13357 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
13358 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13359 // CHECK12:       cond.true:
13360 // CHECK12-NEXT:    br label [[COND_END:%.*]]
13361 // CHECK12:       cond.false:
13362 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
13363 // CHECK12-NEXT:    br label [[COND_END]]
13364 // CHECK12:       cond.end:
13365 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
13366 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
13367 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
13368 // CHECK12-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
13369 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13370 // CHECK12:       omp.inner.for.cond:
13371 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
13372 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
13373 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
13374 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13375 // CHECK12:       omp.inner.for.body:
13376 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4, !llvm.access.group !59
13377 // CHECK12-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
13378 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4, !llvm.access.group !59
13379 // CHECK12-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
13380 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4, !llvm.access.group !59
13381 // CHECK12-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
13382 // CHECK12-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
13383 // CHECK12:       omp_if.then:
13384 // CHECK12-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
13385 // CHECK12-NEXT:    br label [[OMP_IF_END:%.*]]
13386 // CHECK12:       omp_if.else:
13387 // CHECK12-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !59
13388 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !llvm.access.group !59
13389 // CHECK12-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4, !llvm.access.group !59
13390 // CHECK12-NEXT:    call void @.omp_outlined..16(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]], !llvm.access.group !59
13391 // CHECK12-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]]), !llvm.access.group !59
13392 // CHECK12-NEXT:    br label [[OMP_IF_END]]
13393 // CHECK12:       omp_if.end:
13394 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13395 // CHECK12:       omp.inner.for.inc:
13396 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
13397 // CHECK12-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !llvm.access.group !59
13398 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
13399 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !59
13400 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP60:![0-9]+]]
13401 // CHECK12:       omp.inner.for.end:
13402 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
13403 // CHECK12:       omp.loop.exit:
13404 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
13405 // CHECK12-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
13406 // CHECK12-NEXT:    [[TMP17:%.*]] = icmp ne i32 [[TMP16]], 0
13407 // CHECK12-NEXT:    br i1 [[TMP17]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
13408 // CHECK12:       .omp.final.then:
13409 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
13410 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
13411 // CHECK12:       .omp.final.done:
13412 // CHECK12-NEXT:    ret void
13413 //
13414 //
13415 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..16
13416 // CHECK12-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
13417 // CHECK12-NEXT:  entry:
13418 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
13419 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
13420 // CHECK12-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
13421 // CHECK12-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
13422 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13423 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13424 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
13425 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
13426 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
13427 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
13428 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
13429 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
13430 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
13431 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
13432 // CHECK12-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
13433 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
13434 // CHECK12-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
13435 // CHECK12-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
13436 // CHECK12-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
13437 // CHECK12-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
13438 // CHECK12-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
13439 // CHECK12-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
13440 // CHECK12-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
13441 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
13442 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
13443 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
13444 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
13445 // CHECK12-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)
13446 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13447 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
13448 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
13449 // CHECK12:       cond.true:
13450 // CHECK12-NEXT:    br label [[COND_END:%.*]]
13451 // CHECK12:       cond.false:
13452 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
13453 // CHECK12-NEXT:    br label [[COND_END]]
13454 // CHECK12:       cond.end:
13455 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
13456 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
13457 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13458 // CHECK12-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
13459 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13460 // CHECK12:       omp.inner.for.cond:
13461 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
13462 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !62
13463 // CHECK12-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
13464 // CHECK12-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13465 // CHECK12:       omp.inner.for.body:
13466 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
13467 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
13468 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13469 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !62
13470 // CHECK12-NEXT:    call void @_Z3fn3v(), !llvm.access.group !62
13471 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13472 // CHECK12:       omp.body.continue:
13473 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13474 // CHECK12:       omp.inner.for.inc:
13475 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
13476 // CHECK12-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
13477 // CHECK12-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !62
13478 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP63:![0-9]+]]
13479 // CHECK12:       omp.inner.for.end:
13480 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
13481 // CHECK12:       omp.loop.exit:
13482 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
13483 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
13484 // CHECK12-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
13485 // CHECK12-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
13486 // CHECK12:       .omp.final.then:
13487 // CHECK12-NEXT:    store i32 100, i32* [[I]], align 4
13488 // CHECK12-NEXT:    br label [[DOTOMP_FINAL_DONE]]
13489 // CHECK12:       .omp.final.done:
13490 // CHECK12-NEXT:    ret void
13491 //
13492 //
13493 // CHECK12-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
13494 // CHECK12-SAME: () #[[ATTR5:[0-9]+]] {
13495 // CHECK12-NEXT:  entry:
13496 // CHECK12-NEXT:    call void @__tgt_register_requires(i64 1)
13497 // CHECK12-NEXT:    ret void
13498 //
13499 //
13500 // CHECK13-LABEL: define {{[^@]+}}@_Z9gtid_testv
13501 // CHECK13-SAME: () #[[ATTR0:[0-9]+]] {
13502 // CHECK13-NEXT:  entry:
13503 // CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13504 // CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
13505 // CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
13506 // CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13507 // CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
13508 // CHECK13-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
13509 // CHECK13-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
13510 // CHECK13-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
13511 // CHECK13-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
13512 // CHECK13-NEXT:    [[I6:%.*]] = alloca i32, align 4
13513 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
13514 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
13515 // CHECK13-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13516 // CHECK13-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
13517 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13518 // CHECK13:       omp.inner.for.cond:
13519 // CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
13520 // CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6
13521 // CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
13522 // CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13523 // CHECK13:       omp.inner.for.body:
13524 // CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
13525 // CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
13526 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13527 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !6
13528 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13529 // CHECK13:       omp.body.continue:
13530 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13531 // CHECK13:       omp.inner.for.inc:
13532 // CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
13533 // CHECK13-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
13534 // CHECK13-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
13535 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
13536 // CHECK13:       omp.inner.for.end:
13537 // CHECK13-NEXT:    store i32 100, i32* [[I]], align 4
13538 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
13539 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
13540 // CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
13541 // CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
13542 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
13543 // CHECK13:       omp.inner.for.cond7:
13544 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
13545 // CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !10
13546 // CHECK13-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
13547 // CHECK13-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
13548 // CHECK13:       omp.inner.for.body9:
13549 // CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
13550 // CHECK13-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
13551 // CHECK13-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
13552 // CHECK13-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !10
13553 // CHECK13-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !10
13554 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
13555 // CHECK13:       omp.body.continue12:
13556 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
13557 // CHECK13:       omp.inner.for.inc13:
13558 // CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
13559 // CHECK13-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
13560 // CHECK13-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
13561 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP11:![0-9]+]]
13562 // CHECK13:       omp.inner.for.end15:
13563 // CHECK13-NEXT:    store i32 100, i32* [[I6]], align 4
13564 // CHECK13-NEXT:    ret void
13565 //
13566 //
13567 // CHECK13-LABEL: define {{[^@]+}}@main
13568 // CHECK13-SAME: () #[[ATTR1:[0-9]+]] {
13569 // CHECK13-NEXT:  entry:
13570 // CHECK13-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
13571 // CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13572 // CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
13573 // CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
13574 // CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13575 // CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
13576 // CHECK13-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
13577 // CHECK13-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
13578 // CHECK13-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
13579 // CHECK13-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
13580 // CHECK13-NEXT:    [[I6:%.*]] = alloca i32, align 4
13581 // CHECK13-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
13582 // CHECK13-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
13583 // CHECK13-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
13584 // CHECK13-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
13585 // CHECK13-NEXT:    [[I20:%.*]] = alloca i32, align 4
13586 // CHECK13-NEXT:    store i32 0, i32* [[RETVAL]], align 4
13587 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
13588 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
13589 // CHECK13-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13590 // CHECK13-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
13591 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13592 // CHECK13:       omp.inner.for.cond:
13593 // CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
13594 // CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
13595 // CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
13596 // CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13597 // CHECK13:       omp.inner.for.body:
13598 // CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
13599 // CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
13600 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13601 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
13602 // CHECK13-NEXT:    call void @_Z3fn4v(), !llvm.access.group !13
13603 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13604 // CHECK13:       omp.body.continue:
13605 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13606 // CHECK13:       omp.inner.for.inc:
13607 // CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
13608 // CHECK13-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
13609 // CHECK13-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
13610 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
13611 // CHECK13:       omp.inner.for.end:
13612 // CHECK13-NEXT:    store i32 100, i32* [[I]], align 4
13613 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
13614 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
13615 // CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
13616 // CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
13617 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
13618 // CHECK13:       omp.inner.for.cond7:
13619 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
13620 // CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !16
13621 // CHECK13-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
13622 // CHECK13-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
13623 // CHECK13:       omp.inner.for.body9:
13624 // CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
13625 // CHECK13-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
13626 // CHECK13-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
13627 // CHECK13-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !16
13628 // CHECK13-NEXT:    call void @_Z3fn5v(), !llvm.access.group !16
13629 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
13630 // CHECK13:       omp.body.continue12:
13631 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
13632 // CHECK13:       omp.inner.for.inc13:
13633 // CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
13634 // CHECK13-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
13635 // CHECK13-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
13636 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP17:![0-9]+]]
13637 // CHECK13:       omp.inner.for.end15:
13638 // CHECK13-NEXT:    store i32 100, i32* [[I6]], align 4
13639 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
13640 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
13641 // CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
13642 // CHECK13-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
13643 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
13644 // CHECK13:       omp.inner.for.cond21:
13645 // CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !19
13646 // CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !19
13647 // CHECK13-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
13648 // CHECK13-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
13649 // CHECK13:       omp.inner.for.body23:
13650 // CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !19
13651 // CHECK13-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
13652 // CHECK13-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
13653 // CHECK13-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !19
13654 // CHECK13-NEXT:    call void @_Z3fn6v(), !llvm.access.group !19
13655 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
13656 // CHECK13:       omp.body.continue26:
13657 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
13658 // CHECK13:       omp.inner.for.inc27:
13659 // CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !19
13660 // CHECK13-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
13661 // CHECK13-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !19
13662 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP20:![0-9]+]]
13663 // CHECK13:       omp.inner.for.end29:
13664 // CHECK13-NEXT:    store i32 100, i32* [[I20]], align 4
13665 // CHECK13-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
13666 // CHECK13-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
13667 // CHECK13-NEXT:    ret i32 [[CALL]]
13668 //
13669 //
13670 // CHECK13-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
13671 // CHECK13-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
13672 // CHECK13-NEXT:  entry:
13673 // CHECK13-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
13674 // CHECK13-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13675 // CHECK13-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
13676 // CHECK13-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
13677 // CHECK13-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13678 // CHECK13-NEXT:    [[I:%.*]] = alloca i32, align 4
13679 // CHECK13-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
13680 // CHECK13-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
13681 // CHECK13-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
13682 // CHECK13-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
13683 // CHECK13-NEXT:    [[I6:%.*]] = alloca i32, align 4
13684 // CHECK13-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
13685 // CHECK13-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
13686 // CHECK13-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
13687 // CHECK13-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
13688 // CHECK13-NEXT:    [[I20:%.*]] = alloca i32, align 4
13689 // CHECK13-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
13690 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
13691 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
13692 // CHECK13-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13693 // CHECK13-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
13694 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13695 // CHECK13:       omp.inner.for.cond:
13696 // CHECK13-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
13697 // CHECK13-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
13698 // CHECK13-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
13699 // CHECK13-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13700 // CHECK13:       omp.inner.for.body:
13701 // CHECK13-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
13702 // CHECK13-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
13703 // CHECK13-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13704 // CHECK13-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
13705 // CHECK13-NEXT:    call void @_Z3fn1v(), !llvm.access.group !22
13706 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13707 // CHECK13:       omp.body.continue:
13708 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13709 // CHECK13:       omp.inner.for.inc:
13710 // CHECK13-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
13711 // CHECK13-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
13712 // CHECK13-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
13713 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
13714 // CHECK13:       omp.inner.for.end:
13715 // CHECK13-NEXT:    store i32 100, i32* [[I]], align 4
13716 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
13717 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
13718 // CHECK13-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
13719 // CHECK13-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
13720 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
13721 // CHECK13:       omp.inner.for.cond7:
13722 // CHECK13-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !25
13723 // CHECK13-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !25
13724 // CHECK13-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
13725 // CHECK13-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
13726 // CHECK13:       omp.inner.for.body9:
13727 // CHECK13-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !25
13728 // CHECK13-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
13729 // CHECK13-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
13730 // CHECK13-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !25
13731 // CHECK13-NEXT:    call void @_Z3fn2v(), !llvm.access.group !25
13732 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
13733 // CHECK13:       omp.body.continue12:
13734 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
13735 // CHECK13:       omp.inner.for.inc13:
13736 // CHECK13-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !25
13737 // CHECK13-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
13738 // CHECK13-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !25
13739 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP26:![0-9]+]]
13740 // CHECK13:       omp.inner.for.end15:
13741 // CHECK13-NEXT:    store i32 100, i32* [[I6]], align 4
13742 // CHECK13-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
13743 // CHECK13-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
13744 // CHECK13-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
13745 // CHECK13-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
13746 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
13747 // CHECK13:       omp.inner.for.cond21:
13748 // CHECK13-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !28
13749 // CHECK13-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !28
13750 // CHECK13-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
13751 // CHECK13-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
13752 // CHECK13:       omp.inner.for.body23:
13753 // CHECK13-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !28
13754 // CHECK13-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
13755 // CHECK13-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
13756 // CHECK13-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !28
13757 // CHECK13-NEXT:    call void @_Z3fn3v(), !llvm.access.group !28
13758 // CHECK13-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
13759 // CHECK13:       omp.body.continue26:
13760 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
13761 // CHECK13:       omp.inner.for.inc27:
13762 // CHECK13-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !28
13763 // CHECK13-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
13764 // CHECK13-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !28
13765 // CHECK13-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP29:![0-9]+]]
13766 // CHECK13:       omp.inner.for.end29:
13767 // CHECK13-NEXT:    store i32 100, i32* [[I20]], align 4
13768 // CHECK13-NEXT:    ret i32 0
13769 //
13770 //
13771 // CHECK14-LABEL: define {{[^@]+}}@_Z9gtid_testv
13772 // CHECK14-SAME: () #[[ATTR0:[0-9]+]] {
13773 // CHECK14-NEXT:  entry:
13774 // CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13775 // CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
13776 // CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
13777 // CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13778 // CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
13779 // CHECK14-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
13780 // CHECK14-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
13781 // CHECK14-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
13782 // CHECK14-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
13783 // CHECK14-NEXT:    [[I6:%.*]] = alloca i32, align 4
13784 // CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
13785 // CHECK14-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
13786 // CHECK14-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13787 // CHECK14-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
13788 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13789 // CHECK14:       omp.inner.for.cond:
13790 // CHECK14-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
13791 // CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6
13792 // CHECK14-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
13793 // CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13794 // CHECK14:       omp.inner.for.body:
13795 // CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
13796 // CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
13797 // CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13798 // CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !6
13799 // CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13800 // CHECK14:       omp.body.continue:
13801 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13802 // CHECK14:       omp.inner.for.inc:
13803 // CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
13804 // CHECK14-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
13805 // CHECK14-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
13806 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
13807 // CHECK14:       omp.inner.for.end:
13808 // CHECK14-NEXT:    store i32 100, i32* [[I]], align 4
13809 // CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
13810 // CHECK14-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
13811 // CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
13812 // CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
13813 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
13814 // CHECK14:       omp.inner.for.cond7:
13815 // CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
13816 // CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !10
13817 // CHECK14-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
13818 // CHECK14-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
13819 // CHECK14:       omp.inner.for.body9:
13820 // CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
13821 // CHECK14-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
13822 // CHECK14-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
13823 // CHECK14-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !10
13824 // CHECK14-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !10
13825 // CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
13826 // CHECK14:       omp.body.continue12:
13827 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
13828 // CHECK14:       omp.inner.for.inc13:
13829 // CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
13830 // CHECK14-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
13831 // CHECK14-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
13832 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP11:![0-9]+]]
13833 // CHECK14:       omp.inner.for.end15:
13834 // CHECK14-NEXT:    store i32 100, i32* [[I6]], align 4
13835 // CHECK14-NEXT:    ret void
13836 //
13837 //
13838 // CHECK14-LABEL: define {{[^@]+}}@main
13839 // CHECK14-SAME: () #[[ATTR1:[0-9]+]] {
13840 // CHECK14-NEXT:  entry:
13841 // CHECK14-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
13842 // CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13843 // CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
13844 // CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
13845 // CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13846 // CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
13847 // CHECK14-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
13848 // CHECK14-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
13849 // CHECK14-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
13850 // CHECK14-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
13851 // CHECK14-NEXT:    [[I6:%.*]] = alloca i32, align 4
13852 // CHECK14-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
13853 // CHECK14-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
13854 // CHECK14-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
13855 // CHECK14-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
13856 // CHECK14-NEXT:    [[I20:%.*]] = alloca i32, align 4
13857 // CHECK14-NEXT:    store i32 0, i32* [[RETVAL]], align 4
13858 // CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
13859 // CHECK14-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
13860 // CHECK14-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13861 // CHECK14-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
13862 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13863 // CHECK14:       omp.inner.for.cond:
13864 // CHECK14-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
13865 // CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
13866 // CHECK14-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
13867 // CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13868 // CHECK14:       omp.inner.for.body:
13869 // CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
13870 // CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
13871 // CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13872 // CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
13873 // CHECK14-NEXT:    call void @_Z3fn4v(), !llvm.access.group !13
13874 // CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13875 // CHECK14:       omp.body.continue:
13876 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13877 // CHECK14:       omp.inner.for.inc:
13878 // CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
13879 // CHECK14-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
13880 // CHECK14-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
13881 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
13882 // CHECK14:       omp.inner.for.end:
13883 // CHECK14-NEXT:    store i32 100, i32* [[I]], align 4
13884 // CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
13885 // CHECK14-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
13886 // CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
13887 // CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
13888 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
13889 // CHECK14:       omp.inner.for.cond7:
13890 // CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
13891 // CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !16
13892 // CHECK14-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
13893 // CHECK14-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
13894 // CHECK14:       omp.inner.for.body9:
13895 // CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
13896 // CHECK14-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
13897 // CHECK14-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
13898 // CHECK14-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !16
13899 // CHECK14-NEXT:    call void @_Z3fn5v(), !llvm.access.group !16
13900 // CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
13901 // CHECK14:       omp.body.continue12:
13902 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
13903 // CHECK14:       omp.inner.for.inc13:
13904 // CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
13905 // CHECK14-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
13906 // CHECK14-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !16
13907 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP17:![0-9]+]]
13908 // CHECK14:       omp.inner.for.end15:
13909 // CHECK14-NEXT:    store i32 100, i32* [[I6]], align 4
13910 // CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
13911 // CHECK14-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
13912 // CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
13913 // CHECK14-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
13914 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
13915 // CHECK14:       omp.inner.for.cond21:
13916 // CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !19
13917 // CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !19
13918 // CHECK14-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
13919 // CHECK14-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
13920 // CHECK14:       omp.inner.for.body23:
13921 // CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !19
13922 // CHECK14-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
13923 // CHECK14-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
13924 // CHECK14-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !19
13925 // CHECK14-NEXT:    call void @_Z3fn6v(), !llvm.access.group !19
13926 // CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
13927 // CHECK14:       omp.body.continue26:
13928 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
13929 // CHECK14:       omp.inner.for.inc27:
13930 // CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !19
13931 // CHECK14-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
13932 // CHECK14-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !19
13933 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP20:![0-9]+]]
13934 // CHECK14:       omp.inner.for.end29:
13935 // CHECK14-NEXT:    store i32 100, i32* [[I20]], align 4
13936 // CHECK14-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
13937 // CHECK14-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
13938 // CHECK14-NEXT:    ret i32 [[CALL]]
13939 //
13940 //
13941 // CHECK14-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
13942 // CHECK14-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
13943 // CHECK14-NEXT:  entry:
13944 // CHECK14-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
13945 // CHECK14-NEXT:    [[TMP:%.*]] = alloca i32, align 4
13946 // CHECK14-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
13947 // CHECK14-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
13948 // CHECK14-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
13949 // CHECK14-NEXT:    [[I:%.*]] = alloca i32, align 4
13950 // CHECK14-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
13951 // CHECK14-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
13952 // CHECK14-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
13953 // CHECK14-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
13954 // CHECK14-NEXT:    [[I6:%.*]] = alloca i32, align 4
13955 // CHECK14-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
13956 // CHECK14-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
13957 // CHECK14-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
13958 // CHECK14-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
13959 // CHECK14-NEXT:    [[I20:%.*]] = alloca i32, align 4
13960 // CHECK14-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
13961 // CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
13962 // CHECK14-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
13963 // CHECK14-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
13964 // CHECK14-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
13965 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
13966 // CHECK14:       omp.inner.for.cond:
13967 // CHECK14-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
13968 // CHECK14-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
13969 // CHECK14-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
13970 // CHECK14-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
13971 // CHECK14:       omp.inner.for.body:
13972 // CHECK14-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
13973 // CHECK14-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
13974 // CHECK14-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
13975 // CHECK14-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
13976 // CHECK14-NEXT:    call void @_Z3fn1v(), !llvm.access.group !22
13977 // CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
13978 // CHECK14:       omp.body.continue:
13979 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
13980 // CHECK14:       omp.inner.for.inc:
13981 // CHECK14-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
13982 // CHECK14-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
13983 // CHECK14-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
13984 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
13985 // CHECK14:       omp.inner.for.end:
13986 // CHECK14-NEXT:    store i32 100, i32* [[I]], align 4
13987 // CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
13988 // CHECK14-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
13989 // CHECK14-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
13990 // CHECK14-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
13991 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
13992 // CHECK14:       omp.inner.for.cond7:
13993 // CHECK14-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !25
13994 // CHECK14-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !25
13995 // CHECK14-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
13996 // CHECK14-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
13997 // CHECK14:       omp.inner.for.body9:
13998 // CHECK14-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !25
13999 // CHECK14-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
14000 // CHECK14-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
14001 // CHECK14-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !25
14002 // CHECK14-NEXT:    call void @_Z3fn2v(), !llvm.access.group !25
14003 // CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
14004 // CHECK14:       omp.body.continue12:
14005 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
14006 // CHECK14:       omp.inner.for.inc13:
14007 // CHECK14-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !25
14008 // CHECK14-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
14009 // CHECK14-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !25
14010 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP26:![0-9]+]]
14011 // CHECK14:       omp.inner.for.end15:
14012 // CHECK14-NEXT:    store i32 100, i32* [[I6]], align 4
14013 // CHECK14-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
14014 // CHECK14-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
14015 // CHECK14-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
14016 // CHECK14-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
14017 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
14018 // CHECK14:       omp.inner.for.cond21:
14019 // CHECK14-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !28
14020 // CHECK14-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !28
14021 // CHECK14-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
14022 // CHECK14-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
14023 // CHECK14:       omp.inner.for.body23:
14024 // CHECK14-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !28
14025 // CHECK14-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
14026 // CHECK14-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
14027 // CHECK14-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !28
14028 // CHECK14-NEXT:    call void @_Z3fn3v(), !llvm.access.group !28
14029 // CHECK14-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
14030 // CHECK14:       omp.body.continue26:
14031 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
14032 // CHECK14:       omp.inner.for.inc27:
14033 // CHECK14-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !28
14034 // CHECK14-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
14035 // CHECK14-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !28
14036 // CHECK14-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP29:![0-9]+]]
14037 // CHECK14:       omp.inner.for.end29:
14038 // CHECK14-NEXT:    store i32 100, i32* [[I20]], align 4
14039 // CHECK14-NEXT:    ret i32 0
14040 //
14041 //
14042 // CHECK15-LABEL: define {{[^@]+}}@_Z9gtid_testv
14043 // CHECK15-SAME: () #[[ATTR0:[0-9]+]] {
14044 // CHECK15-NEXT:  entry:
14045 // CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14046 // CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
14047 // CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
14048 // CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
14049 // CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
14050 // CHECK15-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
14051 // CHECK15-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
14052 // CHECK15-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
14053 // CHECK15-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
14054 // CHECK15-NEXT:    [[I6:%.*]] = alloca i32, align 4
14055 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
14056 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
14057 // CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14058 // CHECK15-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
14059 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
14060 // CHECK15:       omp.inner.for.cond:
14061 // CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
14062 // CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6
14063 // CHECK15-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
14064 // CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14065 // CHECK15:       omp.inner.for.body:
14066 // CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
14067 // CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
14068 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14069 // CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !6
14070 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
14071 // CHECK15:       omp.body.continue:
14072 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
14073 // CHECK15:       omp.inner.for.inc:
14074 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
14075 // CHECK15-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
14076 // CHECK15-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
14077 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
14078 // CHECK15:       omp.inner.for.end:
14079 // CHECK15-NEXT:    store i32 100, i32* [[I]], align 4
14080 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
14081 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
14082 // CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
14083 // CHECK15-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
14084 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
14085 // CHECK15:       omp.inner.for.cond7:
14086 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
14087 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !10
14088 // CHECK15-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
14089 // CHECK15-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
14090 // CHECK15:       omp.inner.for.body9:
14091 // CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
14092 // CHECK15-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
14093 // CHECK15-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
14094 // CHECK15-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !10
14095 // CHECK15-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !10
14096 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
14097 // CHECK15:       omp.body.continue12:
14098 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
14099 // CHECK15:       omp.inner.for.inc13:
14100 // CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
14101 // CHECK15-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
14102 // CHECK15-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
14103 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP11:![0-9]+]]
14104 // CHECK15:       omp.inner.for.end15:
14105 // CHECK15-NEXT:    store i32 100, i32* [[I6]], align 4
14106 // CHECK15-NEXT:    ret void
14107 //
14108 //
14109 // CHECK15-LABEL: define {{[^@]+}}@main
14110 // CHECK15-SAME: () #[[ATTR1:[0-9]+]] {
14111 // CHECK15-NEXT:  entry:
14112 // CHECK15-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
14113 // CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14114 // CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
14115 // CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
14116 // CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
14117 // CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
14118 // CHECK15-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
14119 // CHECK15-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
14120 // CHECK15-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
14121 // CHECK15-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
14122 // CHECK15-NEXT:    [[I6:%.*]] = alloca i32, align 4
14123 // CHECK15-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
14124 // CHECK15-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
14125 // CHECK15-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
14126 // CHECK15-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
14127 // CHECK15-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
14128 // CHECK15-NEXT:    [[I20:%.*]] = alloca i32, align 4
14129 // CHECK15-NEXT:    store i32 0, i32* [[RETVAL]], align 4
14130 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
14131 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
14132 // CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14133 // CHECK15-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
14134 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
14135 // CHECK15:       omp.inner.for.cond:
14136 // CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
14137 // CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
14138 // CHECK15-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
14139 // CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14140 // CHECK15:       omp.inner.for.body:
14141 // CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
14142 // CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
14143 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14144 // CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
14145 // CHECK15-NEXT:    call void @_Z3fn4v(), !llvm.access.group !13
14146 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
14147 // CHECK15:       omp.body.continue:
14148 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
14149 // CHECK15:       omp.inner.for.inc:
14150 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
14151 // CHECK15-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
14152 // CHECK15-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
14153 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
14154 // CHECK15:       omp.inner.for.end:
14155 // CHECK15-NEXT:    store i32 100, i32* [[I]], align 4
14156 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
14157 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
14158 // CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
14159 // CHECK15-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
14160 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
14161 // CHECK15:       omp.inner.for.cond7:
14162 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
14163 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4
14164 // CHECK15-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
14165 // CHECK15-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
14166 // CHECK15:       omp.inner.for.body9:
14167 // CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
14168 // CHECK15-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
14169 // CHECK15-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
14170 // CHECK15-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4
14171 // CHECK15-NEXT:    call void @_Z3fn5v()
14172 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
14173 // CHECK15:       omp.body.continue12:
14174 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
14175 // CHECK15:       omp.inner.for.inc13:
14176 // CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
14177 // CHECK15-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
14178 // CHECK15-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4
14179 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP16:![0-9]+]]
14180 // CHECK15:       omp.inner.for.end15:
14181 // CHECK15-NEXT:    store i32 100, i32* [[I6]], align 4
14182 // CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* @Arg, align 4
14183 // CHECK15-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP10]], 0
14184 // CHECK15-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
14185 // CHECK15-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
14186 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
14187 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
14188 // CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
14189 // CHECK15-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV19]], align 4
14190 // CHECK15-NEXT:    [[TMP12:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
14191 // CHECK15-NEXT:    [[TOBOOL21:%.*]] = trunc i8 [[TMP12]] to i1
14192 // CHECK15-NEXT:    br i1 [[TOBOOL21]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
14193 // CHECK15:       omp_if.then:
14194 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND22:%.*]]
14195 // CHECK15:       omp.inner.for.cond22:
14196 // CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !18
14197 // CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !18
14198 // CHECK15-NEXT:    [[CMP23:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
14199 // CHECK15-NEXT:    br i1 [[CMP23]], label [[OMP_INNER_FOR_BODY24:%.*]], label [[OMP_INNER_FOR_END30:%.*]]
14200 // CHECK15:       omp.inner.for.body24:
14201 // CHECK15-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !18
14202 // CHECK15-NEXT:    [[MUL25:%.*]] = mul nsw i32 [[TMP15]], 1
14203 // CHECK15-NEXT:    [[ADD26:%.*]] = add nsw i32 0, [[MUL25]]
14204 // CHECK15-NEXT:    store i32 [[ADD26]], i32* [[I20]], align 4, !llvm.access.group !18
14205 // CHECK15-NEXT:    call void @_Z3fn6v(), !llvm.access.group !18
14206 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE27:%.*]]
14207 // CHECK15:       omp.body.continue27:
14208 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC28:%.*]]
14209 // CHECK15:       omp.inner.for.inc28:
14210 // CHECK15-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !18
14211 // CHECK15-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP16]], 1
14212 // CHECK15-NEXT:    store i32 [[ADD29]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !18
14213 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND22]], !llvm.loop [[LOOP19:![0-9]+]]
14214 // CHECK15:       omp.inner.for.end30:
14215 // CHECK15-NEXT:    br label [[OMP_IF_END:%.*]]
14216 // CHECK15:       omp_if.else:
14217 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND31:%.*]]
14218 // CHECK15:       omp.inner.for.cond31:
14219 // CHECK15-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
14220 // CHECK15-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4
14221 // CHECK15-NEXT:    [[CMP32:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
14222 // CHECK15-NEXT:    br i1 [[CMP32]], label [[OMP_INNER_FOR_BODY33:%.*]], label [[OMP_INNER_FOR_END39:%.*]]
14223 // CHECK15:       omp.inner.for.body33:
14224 // CHECK15-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
14225 // CHECK15-NEXT:    [[MUL34:%.*]] = mul nsw i32 [[TMP19]], 1
14226 // CHECK15-NEXT:    [[ADD35:%.*]] = add nsw i32 0, [[MUL34]]
14227 // CHECK15-NEXT:    store i32 [[ADD35]], i32* [[I20]], align 4
14228 // CHECK15-NEXT:    call void @_Z3fn6v()
14229 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE36:%.*]]
14230 // CHECK15:       omp.body.continue36:
14231 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC37:%.*]]
14232 // CHECK15:       omp.inner.for.inc37:
14233 // CHECK15-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
14234 // CHECK15-NEXT:    [[ADD38:%.*]] = add nsw i32 [[TMP20]], 1
14235 // CHECK15-NEXT:    store i32 [[ADD38]], i32* [[DOTOMP_IV19]], align 4
14236 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND31]], !llvm.loop [[LOOP21:![0-9]+]]
14237 // CHECK15:       omp.inner.for.end39:
14238 // CHECK15-NEXT:    br label [[OMP_IF_END]]
14239 // CHECK15:       omp_if.end:
14240 // CHECK15-NEXT:    store i32 100, i32* [[I20]], align 4
14241 // CHECK15-NEXT:    [[TMP21:%.*]] = load i32, i32* @Arg, align 4
14242 // CHECK15-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP21]])
14243 // CHECK15-NEXT:    ret i32 [[CALL]]
14244 //
14245 //
14246 // CHECK15-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
14247 // CHECK15-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
14248 // CHECK15-NEXT:  entry:
14249 // CHECK15-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
14250 // CHECK15-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14251 // CHECK15-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
14252 // CHECK15-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
14253 // CHECK15-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
14254 // CHECK15-NEXT:    [[I:%.*]] = alloca i32, align 4
14255 // CHECK15-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
14256 // CHECK15-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
14257 // CHECK15-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
14258 // CHECK15-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
14259 // CHECK15-NEXT:    [[I6:%.*]] = alloca i32, align 4
14260 // CHECK15-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
14261 // CHECK15-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
14262 // CHECK15-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
14263 // CHECK15-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
14264 // CHECK15-NEXT:    [[I20:%.*]] = alloca i32, align 4
14265 // CHECK15-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
14266 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
14267 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
14268 // CHECK15-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14269 // CHECK15-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
14270 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
14271 // CHECK15:       omp.inner.for.cond:
14272 // CHECK15-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
14273 // CHECK15-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
14274 // CHECK15-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
14275 // CHECK15-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14276 // CHECK15:       omp.inner.for.body:
14277 // CHECK15-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
14278 // CHECK15-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
14279 // CHECK15-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14280 // CHECK15-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
14281 // CHECK15-NEXT:    call void @_Z3fn1v(), !llvm.access.group !22
14282 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
14283 // CHECK15:       omp.body.continue:
14284 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
14285 // CHECK15:       omp.inner.for.inc:
14286 // CHECK15-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
14287 // CHECK15-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
14288 // CHECK15-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
14289 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
14290 // CHECK15:       omp.inner.for.end:
14291 // CHECK15-NEXT:    store i32 100, i32* [[I]], align 4
14292 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
14293 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
14294 // CHECK15-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
14295 // CHECK15-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
14296 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
14297 // CHECK15:       omp.inner.for.cond7:
14298 // CHECK15-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
14299 // CHECK15-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4
14300 // CHECK15-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
14301 // CHECK15-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
14302 // CHECK15:       omp.inner.for.body9:
14303 // CHECK15-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
14304 // CHECK15-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
14305 // CHECK15-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
14306 // CHECK15-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4
14307 // CHECK15-NEXT:    call void @_Z3fn2v()
14308 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
14309 // CHECK15:       omp.body.continue12:
14310 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
14311 // CHECK15:       omp.inner.for.inc13:
14312 // CHECK15-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
14313 // CHECK15-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
14314 // CHECK15-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4
14315 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP25:![0-9]+]]
14316 // CHECK15:       omp.inner.for.end15:
14317 // CHECK15-NEXT:    store i32 100, i32* [[I6]], align 4
14318 // CHECK15-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
14319 // CHECK15-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
14320 // CHECK15-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
14321 // CHECK15-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
14322 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
14323 // CHECK15:       omp.inner.for.cond21:
14324 // CHECK15-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !26
14325 // CHECK15-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !26
14326 // CHECK15-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
14327 // CHECK15-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
14328 // CHECK15:       omp.inner.for.body23:
14329 // CHECK15-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !26
14330 // CHECK15-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
14331 // CHECK15-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
14332 // CHECK15-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !26
14333 // CHECK15-NEXT:    call void @_Z3fn3v(), !llvm.access.group !26
14334 // CHECK15-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
14335 // CHECK15:       omp.body.continue26:
14336 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
14337 // CHECK15:       omp.inner.for.inc27:
14338 // CHECK15-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !26
14339 // CHECK15-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
14340 // CHECK15-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !26
14341 // CHECK15-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP27:![0-9]+]]
14342 // CHECK15:       omp.inner.for.end29:
14343 // CHECK15-NEXT:    store i32 100, i32* [[I20]], align 4
14344 // CHECK15-NEXT:    ret i32 0
14345 //
14346 //
14347 // CHECK16-LABEL: define {{[^@]+}}@_Z9gtid_testv
14348 // CHECK16-SAME: () #[[ATTR0:[0-9]+]] {
14349 // CHECK16-NEXT:  entry:
14350 // CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14351 // CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
14352 // CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
14353 // CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
14354 // CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
14355 // CHECK16-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
14356 // CHECK16-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
14357 // CHECK16-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
14358 // CHECK16-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
14359 // CHECK16-NEXT:    [[I6:%.*]] = alloca i32, align 4
14360 // CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
14361 // CHECK16-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
14362 // CHECK16-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14363 // CHECK16-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
14364 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
14365 // CHECK16:       omp.inner.for.cond:
14366 // CHECK16-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
14367 // CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !6
14368 // CHECK16-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
14369 // CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14370 // CHECK16:       omp.inner.for.body:
14371 // CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
14372 // CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
14373 // CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14374 // CHECK16-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !6
14375 // CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
14376 // CHECK16:       omp.body.continue:
14377 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
14378 // CHECK16:       omp.inner.for.inc:
14379 // CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
14380 // CHECK16-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
14381 // CHECK16-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !6
14382 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
14383 // CHECK16:       omp.inner.for.end:
14384 // CHECK16-NEXT:    store i32 100, i32* [[I]], align 4
14385 // CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
14386 // CHECK16-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
14387 // CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
14388 // CHECK16-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
14389 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
14390 // CHECK16:       omp.inner.for.cond7:
14391 // CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
14392 // CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4, !llvm.access.group !10
14393 // CHECK16-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
14394 // CHECK16-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
14395 // CHECK16:       omp.inner.for.body9:
14396 // CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
14397 // CHECK16-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
14398 // CHECK16-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
14399 // CHECK16-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4, !llvm.access.group !10
14400 // CHECK16-NEXT:    call void @_Z9gtid_testv(), !llvm.access.group !10
14401 // CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
14402 // CHECK16:       omp.body.continue12:
14403 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
14404 // CHECK16:       omp.inner.for.inc13:
14405 // CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
14406 // CHECK16-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
14407 // CHECK16-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4, !llvm.access.group !10
14408 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP11:![0-9]+]]
14409 // CHECK16:       omp.inner.for.end15:
14410 // CHECK16-NEXT:    store i32 100, i32* [[I6]], align 4
14411 // CHECK16-NEXT:    ret void
14412 //
14413 //
14414 // CHECK16-LABEL: define {{[^@]+}}@main
14415 // CHECK16-SAME: () #[[ATTR1:[0-9]+]] {
14416 // CHECK16-NEXT:  entry:
14417 // CHECK16-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
14418 // CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14419 // CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
14420 // CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
14421 // CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
14422 // CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
14423 // CHECK16-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
14424 // CHECK16-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
14425 // CHECK16-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
14426 // CHECK16-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
14427 // CHECK16-NEXT:    [[I6:%.*]] = alloca i32, align 4
14428 // CHECK16-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
14429 // CHECK16-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
14430 // CHECK16-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
14431 // CHECK16-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
14432 // CHECK16-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
14433 // CHECK16-NEXT:    [[I20:%.*]] = alloca i32, align 4
14434 // CHECK16-NEXT:    store i32 0, i32* [[RETVAL]], align 4
14435 // CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
14436 // CHECK16-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
14437 // CHECK16-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14438 // CHECK16-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
14439 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
14440 // CHECK16:       omp.inner.for.cond:
14441 // CHECK16-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
14442 // CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !13
14443 // CHECK16-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
14444 // CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14445 // CHECK16:       omp.inner.for.body:
14446 // CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
14447 // CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
14448 // CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14449 // CHECK16-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !13
14450 // CHECK16-NEXT:    call void @_Z3fn4v(), !llvm.access.group !13
14451 // CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
14452 // CHECK16:       omp.body.continue:
14453 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
14454 // CHECK16:       omp.inner.for.inc:
14455 // CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
14456 // CHECK16-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
14457 // CHECK16-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !13
14458 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP14:![0-9]+]]
14459 // CHECK16:       omp.inner.for.end:
14460 // CHECK16-NEXT:    store i32 100, i32* [[I]], align 4
14461 // CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
14462 // CHECK16-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
14463 // CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
14464 // CHECK16-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
14465 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
14466 // CHECK16:       omp.inner.for.cond7:
14467 // CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
14468 // CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4
14469 // CHECK16-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
14470 // CHECK16-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
14471 // CHECK16:       omp.inner.for.body9:
14472 // CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
14473 // CHECK16-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
14474 // CHECK16-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
14475 // CHECK16-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4
14476 // CHECK16-NEXT:    call void @_Z3fn5v()
14477 // CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
14478 // CHECK16:       omp.body.continue12:
14479 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
14480 // CHECK16:       omp.inner.for.inc13:
14481 // CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
14482 // CHECK16-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
14483 // CHECK16-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4
14484 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP16:![0-9]+]]
14485 // CHECK16:       omp.inner.for.end15:
14486 // CHECK16-NEXT:    store i32 100, i32* [[I6]], align 4
14487 // CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* @Arg, align 4
14488 // CHECK16-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP10]], 0
14489 // CHECK16-NEXT:    [[FROMBOOL:%.*]] = zext i1 [[TOBOOL]] to i8
14490 // CHECK16-NEXT:    store i8 [[FROMBOOL]], i8* [[DOTCAPTURE_EXPR_]], align 1
14491 // CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
14492 // CHECK16-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
14493 // CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
14494 // CHECK16-NEXT:    store i32 [[TMP11]], i32* [[DOTOMP_IV19]], align 4
14495 // CHECK16-NEXT:    [[TMP12:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
14496 // CHECK16-NEXT:    [[TOBOOL21:%.*]] = trunc i8 [[TMP12]] to i1
14497 // CHECK16-NEXT:    br i1 [[TOBOOL21]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
14498 // CHECK16:       omp_if.then:
14499 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND22:%.*]]
14500 // CHECK16:       omp.inner.for.cond22:
14501 // CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !18
14502 // CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !18
14503 // CHECK16-NEXT:    [[CMP23:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
14504 // CHECK16-NEXT:    br i1 [[CMP23]], label [[OMP_INNER_FOR_BODY24:%.*]], label [[OMP_INNER_FOR_END30:%.*]]
14505 // CHECK16:       omp.inner.for.body24:
14506 // CHECK16-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !18
14507 // CHECK16-NEXT:    [[MUL25:%.*]] = mul nsw i32 [[TMP15]], 1
14508 // CHECK16-NEXT:    [[ADD26:%.*]] = add nsw i32 0, [[MUL25]]
14509 // CHECK16-NEXT:    store i32 [[ADD26]], i32* [[I20]], align 4, !llvm.access.group !18
14510 // CHECK16-NEXT:    call void @_Z3fn6v(), !llvm.access.group !18
14511 // CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE27:%.*]]
14512 // CHECK16:       omp.body.continue27:
14513 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC28:%.*]]
14514 // CHECK16:       omp.inner.for.inc28:
14515 // CHECK16-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !18
14516 // CHECK16-NEXT:    [[ADD29:%.*]] = add nsw i32 [[TMP16]], 1
14517 // CHECK16-NEXT:    store i32 [[ADD29]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !18
14518 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND22]], !llvm.loop [[LOOP19:![0-9]+]]
14519 // CHECK16:       omp.inner.for.end30:
14520 // CHECK16-NEXT:    br label [[OMP_IF_END:%.*]]
14521 // CHECK16:       omp_if.else:
14522 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND31:%.*]]
14523 // CHECK16:       omp.inner.for.cond31:
14524 // CHECK16-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
14525 // CHECK16-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4
14526 // CHECK16-NEXT:    [[CMP32:%.*]] = icmp sle i32 [[TMP17]], [[TMP18]]
14527 // CHECK16-NEXT:    br i1 [[CMP32]], label [[OMP_INNER_FOR_BODY33:%.*]], label [[OMP_INNER_FOR_END39:%.*]]
14528 // CHECK16:       omp.inner.for.body33:
14529 // CHECK16-NEXT:    [[TMP19:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
14530 // CHECK16-NEXT:    [[MUL34:%.*]] = mul nsw i32 [[TMP19]], 1
14531 // CHECK16-NEXT:    [[ADD35:%.*]] = add nsw i32 0, [[MUL34]]
14532 // CHECK16-NEXT:    store i32 [[ADD35]], i32* [[I20]], align 4
14533 // CHECK16-NEXT:    call void @_Z3fn6v()
14534 // CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE36:%.*]]
14535 // CHECK16:       omp.body.continue36:
14536 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC37:%.*]]
14537 // CHECK16:       omp.inner.for.inc37:
14538 // CHECK16-NEXT:    [[TMP20:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4
14539 // CHECK16-NEXT:    [[ADD38:%.*]] = add nsw i32 [[TMP20]], 1
14540 // CHECK16-NEXT:    store i32 [[ADD38]], i32* [[DOTOMP_IV19]], align 4
14541 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND31]], !llvm.loop [[LOOP21:![0-9]+]]
14542 // CHECK16:       omp.inner.for.end39:
14543 // CHECK16-NEXT:    br label [[OMP_IF_END]]
14544 // CHECK16:       omp_if.end:
14545 // CHECK16-NEXT:    store i32 100, i32* [[I20]], align 4
14546 // CHECK16-NEXT:    [[TMP21:%.*]] = load i32, i32* @Arg, align 4
14547 // CHECK16-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP21]])
14548 // CHECK16-NEXT:    ret i32 [[CALL]]
14549 //
14550 //
14551 // CHECK16-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
14552 // CHECK16-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
14553 // CHECK16-NEXT:  entry:
14554 // CHECK16-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
14555 // CHECK16-NEXT:    [[TMP:%.*]] = alloca i32, align 4
14556 // CHECK16-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
14557 // CHECK16-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
14558 // CHECK16-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
14559 // CHECK16-NEXT:    [[I:%.*]] = alloca i32, align 4
14560 // CHECK16-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
14561 // CHECK16-NEXT:    [[DOTOMP_LB3:%.*]] = alloca i32, align 4
14562 // CHECK16-NEXT:    [[DOTOMP_UB4:%.*]] = alloca i32, align 4
14563 // CHECK16-NEXT:    [[DOTOMP_IV5:%.*]] = alloca i32, align 4
14564 // CHECK16-NEXT:    [[I6:%.*]] = alloca i32, align 4
14565 // CHECK16-NEXT:    [[_TMP16:%.*]] = alloca i32, align 4
14566 // CHECK16-NEXT:    [[DOTOMP_LB17:%.*]] = alloca i32, align 4
14567 // CHECK16-NEXT:    [[DOTOMP_UB18:%.*]] = alloca i32, align 4
14568 // CHECK16-NEXT:    [[DOTOMP_IV19:%.*]] = alloca i32, align 4
14569 // CHECK16-NEXT:    [[I20:%.*]] = alloca i32, align 4
14570 // CHECK16-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
14571 // CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
14572 // CHECK16-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
14573 // CHECK16-NEXT:    [[TMP0:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
14574 // CHECK16-NEXT:    store i32 [[TMP0]], i32* [[DOTOMP_IV]], align 4
14575 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
14576 // CHECK16:       omp.inner.for.cond:
14577 // CHECK16-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
14578 // CHECK16-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !22
14579 // CHECK16-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP1]], [[TMP2]]
14580 // CHECK16-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
14581 // CHECK16:       omp.inner.for.body:
14582 // CHECK16-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
14583 // CHECK16-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP3]], 1
14584 // CHECK16-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
14585 // CHECK16-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !llvm.access.group !22
14586 // CHECK16-NEXT:    call void @_Z3fn1v(), !llvm.access.group !22
14587 // CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
14588 // CHECK16:       omp.body.continue:
14589 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
14590 // CHECK16:       omp.inner.for.inc:
14591 // CHECK16-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
14592 // CHECK16-NEXT:    [[ADD1:%.*]] = add nsw i32 [[TMP4]], 1
14593 // CHECK16-NEXT:    store i32 [[ADD1]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !22
14594 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP23:![0-9]+]]
14595 // CHECK16:       omp.inner.for.end:
14596 // CHECK16-NEXT:    store i32 100, i32* [[I]], align 4
14597 // CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB3]], align 4
14598 // CHECK16-NEXT:    store i32 99, i32* [[DOTOMP_UB4]], align 4
14599 // CHECK16-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_LB3]], align 4
14600 // CHECK16-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV5]], align 4
14601 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND7:%.*]]
14602 // CHECK16:       omp.inner.for.cond7:
14603 // CHECK16-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
14604 // CHECK16-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB4]], align 4
14605 // CHECK16-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
14606 // CHECK16-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY9:%.*]], label [[OMP_INNER_FOR_END15:%.*]]
14607 // CHECK16:       omp.inner.for.body9:
14608 // CHECK16-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
14609 // CHECK16-NEXT:    [[MUL10:%.*]] = mul nsw i32 [[TMP8]], 1
14610 // CHECK16-NEXT:    [[ADD11:%.*]] = add nsw i32 0, [[MUL10]]
14611 // CHECK16-NEXT:    store i32 [[ADD11]], i32* [[I6]], align 4
14612 // CHECK16-NEXT:    call void @_Z3fn2v()
14613 // CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE12:%.*]]
14614 // CHECK16:       omp.body.continue12:
14615 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC13:%.*]]
14616 // CHECK16:       omp.inner.for.inc13:
14617 // CHECK16-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV5]], align 4
14618 // CHECK16-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP9]], 1
14619 // CHECK16-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV5]], align 4
14620 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND7]], !llvm.loop [[LOOP25:![0-9]+]]
14621 // CHECK16:       omp.inner.for.end15:
14622 // CHECK16-NEXT:    store i32 100, i32* [[I6]], align 4
14623 // CHECK16-NEXT:    store i32 0, i32* [[DOTOMP_LB17]], align 4
14624 // CHECK16-NEXT:    store i32 99, i32* [[DOTOMP_UB18]], align 4
14625 // CHECK16-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_LB17]], align 4
14626 // CHECK16-NEXT:    store i32 [[TMP10]], i32* [[DOTOMP_IV19]], align 4
14627 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND21:%.*]]
14628 // CHECK16:       omp.inner.for.cond21:
14629 // CHECK16-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !26
14630 // CHECK16-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB18]], align 4, !llvm.access.group !26
14631 // CHECK16-NEXT:    [[CMP22:%.*]] = icmp sle i32 [[TMP11]], [[TMP12]]
14632 // CHECK16-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY23:%.*]], label [[OMP_INNER_FOR_END29:%.*]]
14633 // CHECK16:       omp.inner.for.body23:
14634 // CHECK16-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !26
14635 // CHECK16-NEXT:    [[MUL24:%.*]] = mul nsw i32 [[TMP13]], 1
14636 // CHECK16-NEXT:    [[ADD25:%.*]] = add nsw i32 0, [[MUL24]]
14637 // CHECK16-NEXT:    store i32 [[ADD25]], i32* [[I20]], align 4, !llvm.access.group !26
14638 // CHECK16-NEXT:    call void @_Z3fn3v(), !llvm.access.group !26
14639 // CHECK16-NEXT:    br label [[OMP_BODY_CONTINUE26:%.*]]
14640 // CHECK16:       omp.body.continue26:
14641 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_INC27:%.*]]
14642 // CHECK16:       omp.inner.for.inc27:
14643 // CHECK16-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !26
14644 // CHECK16-NEXT:    [[ADD28:%.*]] = add nsw i32 [[TMP14]], 1
14645 // CHECK16-NEXT:    store i32 [[ADD28]], i32* [[DOTOMP_IV19]], align 4, !llvm.access.group !26
14646 // CHECK16-NEXT:    br label [[OMP_INNER_FOR_COND21]], !llvm.loop [[LOOP27:![0-9]+]]
14647 // CHECK16:       omp.inner.for.end29:
14648 // CHECK16-NEXT:    store i32 100, i32* [[I20]], align 4
14649 // CHECK16-NEXT:    ret i32 0
14650 //
14651