1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
2 // RUN: %clang_cc1 -no-opaque-pointers -verify -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
4 // RUN: %clang_cc1 -no-opaque-pointers -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
5 
6 // 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 --implicit-check-not="{{__kmpc|__tgt}}"
7 // 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
8 // 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 --implicit-check-not="{{__kmpc|__tgt}}"
9 
10 // 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=CHECK1
11 // 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
12 // 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=CHECK1
13 
14 // 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 --implicit-check-not="{{__kmpc|__tgt}}"
15 // 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
16 // 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 --implicit-check-not="{{__kmpc|__tgt}}"
17 // 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=CHECK1
18 // 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
19 // 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=CHECK1
20 
21 // 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 --implicit-check-not="{{__kmpc|__tgt}}"
22 // 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
23 // 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 --implicit-check-not="{{__kmpc|__tgt}}"
24 
25 // 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=CHECK1
26 // 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
27 // 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=CHECK1
28 
29 // 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 --implicit-check-not="{{__kmpc|__tgt}}"
30 // 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
31 // 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 --implicit-check-not="{{__kmpc|__tgt}}"
32 
33 // expected-no-diagnostics
34 #ifndef HEADER
35 #define HEADER
36 
37 void fn1();
38 void fn2();
39 void fn3();
40 void fn4();
41 void fn5();
42 void fn6();
43 
44 int Arg;
45 
46 void gtid_test() {
47 #pragma omp target
48 #pragma omp teams
49 #pragma omp distribute parallel for
50   for(int i = 0 ; i < 100; i++) {}
51 
52 #pragma omp target
53 #pragma omp teams
54 #pragma omp distribute parallel for if (parallel: false)
55   for(int i = 0 ; i < 100; i++) {
56     gtid_test();
57   }
58 }
59 
60 
61 template <typename T>
62 int tmain(T Arg) {
63 #pragma omp target
64 #pragma omp teams
65 #pragma omp distribute parallel for if (true)
66   for(int i = 0 ; i < 100; i++) {
67     fn1();
68   }
69 #pragma omp target
70 #pragma omp teams
71 #pragma omp distribute parallel for if (false)
72   for(int i = 0 ; i < 100; i++) {
73     fn2();
74   }
75 #pragma omp target
76 #pragma omp teams
77 #pragma omp distribute parallel for if (parallel: Arg)
78   for(int i = 0 ; i < 100; i++) {
79     fn3();
80   }
81   return 0;
82 }
83 
84 int main() {
85 #pragma omp target
86 #pragma omp teams
87 #pragma omp distribute parallel for if (true)
88   for(int i = 0 ; i < 100; i++) {
89 
90 
91     fn4();
92   }
93 
94 #pragma omp target
95 #pragma omp teams
96 #pragma omp distribute parallel for if (false)
97   for(int i = 0 ; i < 100; i++) {
98 
99 
100     fn5();
101   }
102 
103 #pragma omp target
104 #pragma omp teams
105 #pragma omp distribute parallel for if (Arg)
106   for(int i = 0 ; i < 100; i++) {
107 
108 
109     fn6();
110   }
111 
112   return tmain(Arg);
113 }
114 
115 
116 
117 
118 
119 
120 // call void [[T_OUTLINE_FUN_3:@.+]](
121 
122 #endif
123 // CHECK1-LABEL: define {{[^@]+}}@_Z9gtid_testv
124 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
125 // CHECK1-NEXT:  entry:
126 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
127 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
128 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 100)
129 // 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_l47.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
130 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
131 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
132 // CHECK1:       omp_offload.failed:
133 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47() #[[ATTR2:[0-9]+]]
134 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
135 // CHECK1:       omp_offload.cont:
136 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
137 // 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_l52.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
138 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
139 // CHECK1-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
140 // CHECK1:       omp_offload.failed2:
141 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52() #[[ATTR2]]
142 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
143 // CHECK1:       omp_offload.cont3:
144 // CHECK1-NEXT:    ret void
145 //
146 //
147 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l47
148 // CHECK1-SAME: () #[[ATTR1:[0-9]+]] {
149 // CHECK1-NEXT:  entry:
150 // 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*, ...)*))
151 // CHECK1-NEXT:    ret void
152 //
153 //
154 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
155 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
156 // CHECK1-NEXT:  entry:
157 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
158 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
159 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
160 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
161 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
162 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
163 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
164 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
165 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
166 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
167 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
168 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
169 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
170 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
171 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
172 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
173 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
174 // 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)
175 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
176 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
177 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
178 // CHECK1:       cond.true:
179 // CHECK1-NEXT:    br label [[COND_END:%.*]]
180 // CHECK1:       cond.false:
181 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
182 // CHECK1-NEXT:    br label [[COND_END]]
183 // CHECK1:       cond.end:
184 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
185 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
186 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
187 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
188 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
189 // CHECK1:       omp.inner.for.cond:
190 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
191 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
192 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
193 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
194 // CHECK1:       omp.inner.for.body:
195 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
196 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
197 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
198 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
199 // 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]])
200 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
201 // CHECK1:       omp.inner.for.inc:
202 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
203 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
204 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
205 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
206 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
207 // CHECK1:       omp.inner.for.end:
208 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
209 // CHECK1:       omp.loop.exit:
210 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
211 // CHECK1-NEXT:    ret void
212 //
213 //
214 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
215 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
216 // CHECK1-NEXT:  entry:
217 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
218 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
219 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
220 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
221 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
222 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
223 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
224 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
225 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
226 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
227 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
228 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
229 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
230 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
231 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
232 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
233 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
234 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
235 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
236 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
237 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
238 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
239 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
240 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
241 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
242 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
243 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
244 // 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)
245 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
246 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
247 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
248 // CHECK1:       cond.true:
249 // CHECK1-NEXT:    br label [[COND_END:%.*]]
250 // CHECK1:       cond.false:
251 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
252 // CHECK1-NEXT:    br label [[COND_END]]
253 // CHECK1:       cond.end:
254 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
255 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
256 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
257 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
258 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
259 // CHECK1:       omp.inner.for.cond:
260 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
261 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
262 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
263 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
264 // CHECK1:       omp.inner.for.body:
265 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
266 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
267 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
268 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
269 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
270 // CHECK1:       omp.body.continue:
271 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
272 // CHECK1:       omp.inner.for.inc:
273 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
274 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
275 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
276 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
277 // CHECK1:       omp.inner.for.end:
278 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
279 // CHECK1:       omp.loop.exit:
280 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
281 // CHECK1-NEXT:    ret void
282 //
283 //
284 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z9gtid_testv_l52
285 // CHECK1-SAME: () #[[ATTR1]] {
286 // CHECK1-NEXT:  entry:
287 // 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*, ...)*))
288 // CHECK1-NEXT:    ret void
289 //
290 //
291 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
292 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
293 // CHECK1-NEXT:  entry:
294 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
295 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
296 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
297 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
298 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
299 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
300 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
301 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
302 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
303 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
304 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
305 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
306 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
307 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
308 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
309 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
310 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
311 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
312 // 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)
313 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
314 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
315 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
316 // CHECK1:       cond.true:
317 // CHECK1-NEXT:    br label [[COND_END:%.*]]
318 // CHECK1:       cond.false:
319 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
320 // CHECK1-NEXT:    br label [[COND_END]]
321 // CHECK1:       cond.end:
322 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
323 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
324 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
325 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
326 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
327 // CHECK1:       omp.inner.for.cond:
328 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
329 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
330 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
331 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
332 // CHECK1:       omp.inner.for.body:
333 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
334 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
335 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
336 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
337 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
338 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
339 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
340 // CHECK1-NEXT:    call void @.omp_outlined..3(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
341 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
342 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
343 // CHECK1:       omp.inner.for.inc:
344 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
345 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
346 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
347 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
348 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
349 // CHECK1:       omp.inner.for.end:
350 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
351 // CHECK1:       omp.loop.exit:
352 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
353 // CHECK1-NEXT:    ret void
354 //
355 //
356 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
357 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
358 // CHECK1-NEXT:  entry:
359 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
360 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
361 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
362 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
363 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
364 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
365 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
366 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
367 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
368 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
369 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
370 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
371 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
372 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
373 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
374 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
375 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
376 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
377 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
378 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
379 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
380 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
381 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
382 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
383 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
384 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
385 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
386 // 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)
387 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
388 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
389 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
390 // CHECK1:       cond.true:
391 // CHECK1-NEXT:    br label [[COND_END:%.*]]
392 // CHECK1:       cond.false:
393 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
394 // CHECK1-NEXT:    br label [[COND_END]]
395 // CHECK1:       cond.end:
396 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
397 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
398 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
399 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
400 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
401 // CHECK1:       omp.inner.for.cond:
402 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
403 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
404 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
405 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
406 // CHECK1:       omp.inner.for.body:
407 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
408 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
409 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
410 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
411 // CHECK1-NEXT:    call void @_Z9gtid_testv()
412 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
413 // CHECK1:       omp.body.continue:
414 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
415 // CHECK1:       omp.inner.for.inc:
416 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
417 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
418 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
419 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
420 // CHECK1:       omp.inner.for.end:
421 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
422 // CHECK1:       omp.loop.exit:
423 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
424 // CHECK1-NEXT:    ret void
425 //
426 //
427 // CHECK1-LABEL: define {{[^@]+}}@main
428 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
429 // CHECK1-NEXT:  entry:
430 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
431 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
432 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
433 // CHECK1-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
434 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
435 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
436 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
437 // CHECK1-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
438 // CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
439 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
440 // 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_l85.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
441 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
442 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
443 // CHECK1:       omp_offload.failed:
444 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85() #[[ATTR2]]
445 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
446 // CHECK1:       omp_offload.cont:
447 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
448 // 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_l94.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
449 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
450 // CHECK1-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
451 // CHECK1:       omp_offload.failed2:
452 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94() #[[ATTR2]]
453 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
454 // CHECK1:       omp_offload.cont3:
455 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* @Arg, align 4
456 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
457 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
458 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
459 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
460 // CHECK1-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
461 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
462 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
463 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
464 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
465 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
466 // CHECK1-NEXT:    store i8* null, i8** [[TMP10]], align 8
467 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
468 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
469 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
470 // 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_l103.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 0)
471 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
472 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
473 // CHECK1:       omp_offload.failed5:
474 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103(i64 [[TMP5]]) #[[ATTR2]]
475 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
476 // CHECK1:       omp_offload.cont6:
477 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* @Arg, align 4
478 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiEiT_(i32 noundef [[TMP15]])
479 // CHECK1-NEXT:    ret i32 [[CALL]]
480 //
481 //
482 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l85
483 // CHECK1-SAME: () #[[ATTR1]] {
484 // CHECK1-NEXT:  entry:
485 // 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*, ...)*))
486 // CHECK1-NEXT:    ret void
487 //
488 //
489 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
490 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
491 // CHECK1-NEXT:  entry:
492 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
493 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
494 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
495 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
496 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
497 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
498 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
499 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
500 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
501 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
502 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
503 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
504 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
505 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
506 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
507 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
508 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
509 // 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)
510 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
511 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
512 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
513 // CHECK1:       cond.true:
514 // CHECK1-NEXT:    br label [[COND_END:%.*]]
515 // CHECK1:       cond.false:
516 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
517 // CHECK1-NEXT:    br label [[COND_END]]
518 // CHECK1:       cond.end:
519 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
520 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
521 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
522 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
523 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
524 // CHECK1:       omp.inner.for.cond:
525 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
526 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
527 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
528 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
529 // CHECK1:       omp.inner.for.body:
530 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
531 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
532 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
533 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
534 // 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]])
535 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
536 // CHECK1:       omp.inner.for.inc:
537 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
538 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
539 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
540 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
541 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
542 // CHECK1:       omp.inner.for.end:
543 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
544 // CHECK1:       omp.loop.exit:
545 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
546 // CHECK1-NEXT:    ret void
547 //
548 //
549 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..5
550 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
551 // CHECK1-NEXT:  entry:
552 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
553 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
554 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
555 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
556 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
557 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
558 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
559 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
560 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
561 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
562 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
563 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
564 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
565 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
566 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
567 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
568 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
569 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
570 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
571 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
572 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
573 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
574 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
575 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
576 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
577 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
578 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
579 // 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)
580 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
581 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
582 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
583 // CHECK1:       cond.true:
584 // CHECK1-NEXT:    br label [[COND_END:%.*]]
585 // CHECK1:       cond.false:
586 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
587 // CHECK1-NEXT:    br label [[COND_END]]
588 // CHECK1:       cond.end:
589 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
590 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
591 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
592 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
593 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
594 // CHECK1:       omp.inner.for.cond:
595 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
596 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
597 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
598 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
599 // CHECK1:       omp.inner.for.body:
600 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
601 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
602 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
603 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
604 // CHECK1-NEXT:    call void @_Z3fn4v()
605 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
606 // CHECK1:       omp.body.continue:
607 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
608 // CHECK1:       omp.inner.for.inc:
609 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
610 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
611 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
612 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
613 // CHECK1:       omp.inner.for.end:
614 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
615 // CHECK1:       omp.loop.exit:
616 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
617 // CHECK1-NEXT:    ret void
618 //
619 //
620 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l94
621 // CHECK1-SAME: () #[[ATTR1]] {
622 // CHECK1-NEXT:  entry:
623 // 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*, ...)*))
624 // CHECK1-NEXT:    ret void
625 //
626 //
627 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
628 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
629 // CHECK1-NEXT:  entry:
630 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
631 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
632 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
633 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
634 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
635 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
636 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
637 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
638 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
639 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
640 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
641 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
642 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
643 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
644 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
645 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
646 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
647 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
648 // 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)
649 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
650 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
651 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
652 // CHECK1:       cond.true:
653 // CHECK1-NEXT:    br label [[COND_END:%.*]]
654 // CHECK1:       cond.false:
655 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
656 // CHECK1-NEXT:    br label [[COND_END]]
657 // CHECK1:       cond.end:
658 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
659 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
660 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
661 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
662 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
663 // CHECK1:       omp.inner.for.cond:
664 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
665 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
666 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
667 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
668 // CHECK1:       omp.inner.for.body:
669 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
670 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
671 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
672 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
673 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
674 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
675 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
676 // CHECK1-NEXT:    call void @.omp_outlined..7(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
677 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
678 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
679 // CHECK1:       omp.inner.for.inc:
680 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
681 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
682 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
683 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
684 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
685 // CHECK1:       omp.inner.for.end:
686 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
687 // CHECK1:       omp.loop.exit:
688 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
689 // CHECK1-NEXT:    ret void
690 //
691 //
692 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
693 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
694 // CHECK1-NEXT:  entry:
695 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
696 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
697 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
698 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
699 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
700 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
701 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
702 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
703 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
704 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
705 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
706 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
707 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
708 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
709 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
710 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
711 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
712 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
713 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
714 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
715 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
716 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
717 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
718 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
719 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
720 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
721 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
722 // 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)
723 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
724 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
725 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
726 // CHECK1:       cond.true:
727 // CHECK1-NEXT:    br label [[COND_END:%.*]]
728 // CHECK1:       cond.false:
729 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
730 // CHECK1-NEXT:    br label [[COND_END]]
731 // CHECK1:       cond.end:
732 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
733 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
734 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
735 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
736 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
737 // CHECK1:       omp.inner.for.cond:
738 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
739 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
740 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
741 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
742 // CHECK1:       omp.inner.for.body:
743 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
744 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
745 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
746 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
747 // CHECK1-NEXT:    call void @_Z3fn5v()
748 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
749 // CHECK1:       omp.body.continue:
750 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
751 // CHECK1:       omp.inner.for.inc:
752 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
753 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
754 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
755 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
756 // CHECK1:       omp.inner.for.end:
757 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
758 // CHECK1:       omp.loop.exit:
759 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
760 // CHECK1-NEXT:    ret void
761 //
762 //
763 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l103
764 // CHECK1-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
765 // CHECK1-NEXT:  entry:
766 // CHECK1-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
767 // CHECK1-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
768 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
769 // 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]])
770 // CHECK1-NEXT:    ret void
771 //
772 //
773 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..8
774 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
775 // CHECK1-NEXT:  entry:
776 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
777 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
778 // CHECK1-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
779 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
780 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
781 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
782 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
783 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
784 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
785 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
786 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
787 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
788 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
789 // CHECK1-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
790 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
791 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
792 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
793 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
794 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
795 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
796 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
797 // 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)
798 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
799 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
800 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
801 // CHECK1:       cond.true:
802 // CHECK1-NEXT:    br label [[COND_END:%.*]]
803 // CHECK1:       cond.false:
804 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
805 // CHECK1-NEXT:    br label [[COND_END]]
806 // CHECK1:       cond.end:
807 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
808 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
809 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
810 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
811 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
812 // CHECK1:       omp.inner.for.cond:
813 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
814 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
815 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
816 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
817 // CHECK1:       omp.inner.for.body:
818 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
819 // CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
820 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
821 // CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
822 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
823 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
824 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
825 // CHECK1:       omp_if.then:
826 // 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]])
827 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
828 // CHECK1:       omp_if.else:
829 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
830 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
831 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
832 // CHECK1-NEXT:    call void @.omp_outlined..9(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
833 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
834 // CHECK1-NEXT:    br label [[OMP_IF_END]]
835 // CHECK1:       omp_if.end:
836 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
837 // CHECK1:       omp.inner.for.inc:
838 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
839 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
840 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
841 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
842 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
843 // CHECK1:       omp.inner.for.end:
844 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
845 // CHECK1:       omp.loop.exit:
846 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
847 // CHECK1-NEXT:    ret void
848 //
849 //
850 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..9
851 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
852 // CHECK1-NEXT:  entry:
853 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
854 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
855 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
856 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
857 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
858 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
859 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
860 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
861 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
862 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
863 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
864 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
865 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
866 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
867 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
868 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
869 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
870 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
871 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
872 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
873 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
874 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
875 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
876 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
877 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
878 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
879 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
880 // 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)
881 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
882 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
883 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
884 // CHECK1:       cond.true:
885 // CHECK1-NEXT:    br label [[COND_END:%.*]]
886 // CHECK1:       cond.false:
887 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
888 // CHECK1-NEXT:    br label [[COND_END]]
889 // CHECK1:       cond.end:
890 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
891 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
892 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
893 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
894 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
895 // CHECK1:       omp.inner.for.cond:
896 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
897 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
898 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
899 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
900 // CHECK1:       omp.inner.for.body:
901 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
902 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
903 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
904 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
905 // CHECK1-NEXT:    call void @_Z3fn6v()
906 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
907 // CHECK1:       omp.body.continue:
908 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
909 // CHECK1:       omp.inner.for.inc:
910 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
911 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
912 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
913 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
914 // CHECK1:       omp.inner.for.end:
915 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
916 // CHECK1:       omp.loop.exit:
917 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
918 // CHECK1-NEXT:    ret void
919 //
920 //
921 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiEiT_
922 // CHECK1-SAME: (i32 noundef [[ARG:%.*]]) #[[ATTR0]] comdat {
923 // CHECK1-NEXT:  entry:
924 // CHECK1-NEXT:    [[ARG_ADDR:%.*]] = alloca i32, align 4
925 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
926 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
927 // CHECK1-NEXT:    [[ARG_CASTED:%.*]] = alloca i64, align 8
928 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [1 x i8*], align 8
929 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [1 x i8*], align 8
930 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [1 x i8*], align 8
931 // CHECK1-NEXT:    [[_TMP4:%.*]] = alloca i32, align 4
932 // CHECK1-NEXT:    store i32 [[ARG]], i32* [[ARG_ADDR]], align 4
933 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
934 // 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__l63.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
935 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
936 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
937 // CHECK1:       omp_offload.failed:
938 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63() #[[ATTR2]]
939 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
940 // CHECK1:       omp_offload.cont:
941 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
942 // 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__l69.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
943 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
944 // CHECK1-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
945 // CHECK1:       omp_offload.failed2:
946 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69() #[[ATTR2]]
947 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
948 // CHECK1:       omp_offload.cont3:
949 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[ARG_ADDR]], align 4
950 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_CASTED]] to i32*
951 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[CONV]], align 4
952 // CHECK1-NEXT:    [[TMP5:%.*]] = load i64, i64* [[ARG_CASTED]], align 8
953 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
954 // CHECK1-NEXT:    [[TMP7:%.*]] = bitcast i8** [[TMP6]] to i64*
955 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP7]], align 8
956 // CHECK1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
957 // CHECK1-NEXT:    [[TMP9:%.*]] = bitcast i8** [[TMP8]] to i64*
958 // CHECK1-NEXT:    store i64 [[TMP5]], i64* [[TMP9]], align 8
959 // CHECK1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
960 // CHECK1-NEXT:    store i8* null, i8** [[TMP10]], align 8
961 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
962 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
963 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 100)
964 // 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__l75.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 0)
965 // CHECK1-NEXT:    [[TMP14:%.*]] = icmp ne i32 [[TMP13]], 0
966 // CHECK1-NEXT:    br i1 [[TMP14]], label [[OMP_OFFLOAD_FAILED5:%.*]], label [[OMP_OFFLOAD_CONT6:%.*]]
967 // CHECK1:       omp_offload.failed5:
968 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75(i64 [[TMP5]]) #[[ATTR2]]
969 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT6]]
970 // CHECK1:       omp_offload.cont6:
971 // CHECK1-NEXT:    ret i32 0
972 //
973 //
974 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l63
975 // CHECK1-SAME: () #[[ATTR1]] {
976 // CHECK1-NEXT:  entry:
977 // 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*, ...)*))
978 // CHECK1-NEXT:    ret void
979 //
980 //
981 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..10
982 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
983 // CHECK1-NEXT:  entry:
984 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
985 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
986 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
987 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
988 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
989 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
990 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
991 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
992 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
993 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
994 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
995 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
996 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
997 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
998 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
999 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1000 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1001 // 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)
1002 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1003 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1004 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1005 // CHECK1:       cond.true:
1006 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1007 // CHECK1:       cond.false:
1008 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1009 // CHECK1-NEXT:    br label [[COND_END]]
1010 // CHECK1:       cond.end:
1011 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1012 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1013 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1014 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1015 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1016 // CHECK1:       omp.inner.for.cond:
1017 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1018 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1019 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1020 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1021 // CHECK1:       omp.inner.for.body:
1022 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1023 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1024 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1025 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1026 // 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]])
1027 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1028 // CHECK1:       omp.inner.for.inc:
1029 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1030 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1031 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
1032 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1033 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1034 // CHECK1:       omp.inner.for.end:
1035 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1036 // CHECK1:       omp.loop.exit:
1037 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1038 // CHECK1-NEXT:    ret void
1039 //
1040 //
1041 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..11
1042 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1043 // CHECK1-NEXT:  entry:
1044 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1045 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1046 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1047 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1048 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1049 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1050 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1051 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1052 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1053 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1054 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1055 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1056 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1057 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1058 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1059 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1060 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
1061 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1062 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1063 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1064 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1065 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1066 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1067 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1068 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1069 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1070 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1071 // 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)
1072 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1073 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1074 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1075 // CHECK1:       cond.true:
1076 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1077 // CHECK1:       cond.false:
1078 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1079 // CHECK1-NEXT:    br label [[COND_END]]
1080 // CHECK1:       cond.end:
1081 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1082 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1083 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1084 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1085 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1086 // CHECK1:       omp.inner.for.cond:
1087 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1088 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1089 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1090 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1091 // CHECK1:       omp.inner.for.body:
1092 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1093 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1094 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1095 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
1096 // CHECK1-NEXT:    call void @_Z3fn1v()
1097 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1098 // CHECK1:       omp.body.continue:
1099 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1100 // CHECK1:       omp.inner.for.inc:
1101 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1102 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1103 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1104 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1105 // CHECK1:       omp.inner.for.end:
1106 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1107 // CHECK1:       omp.loop.exit:
1108 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1109 // CHECK1-NEXT:    ret void
1110 //
1111 //
1112 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l69
1113 // CHECK1-SAME: () #[[ATTR1]] {
1114 // CHECK1-NEXT:  entry:
1115 // 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*, ...)*))
1116 // CHECK1-NEXT:    ret void
1117 //
1118 //
1119 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..12
1120 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
1121 // CHECK1-NEXT:  entry:
1122 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1123 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1124 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1125 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1126 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1127 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1128 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1129 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1130 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1131 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1132 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1133 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1134 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1135 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1136 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1137 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1138 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1139 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
1140 // 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)
1141 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1142 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 99
1143 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1144 // CHECK1:       cond.true:
1145 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1146 // CHECK1:       cond.false:
1147 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1148 // CHECK1-NEXT:    br label [[COND_END]]
1149 // CHECK1:       cond.end:
1150 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
1151 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1152 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1153 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
1154 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1155 // CHECK1:       omp.inner.for.cond:
1156 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1157 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1158 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
1159 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1160 // CHECK1:       omp.inner.for.body:
1161 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1162 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
1163 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1164 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
1165 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
1166 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1167 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1168 // CHECK1-NEXT:    call void @.omp_outlined..13(i32* [[TMP11]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP8]], i64 [[TMP10]]) #[[ATTR2]]
1169 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]])
1170 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1171 // CHECK1:       omp.inner.for.inc:
1172 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1173 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1174 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP12]], [[TMP13]]
1175 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1176 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1177 // CHECK1:       omp.inner.for.end:
1178 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1179 // CHECK1:       omp.loop.exit:
1180 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
1181 // CHECK1-NEXT:    ret void
1182 //
1183 //
1184 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..13
1185 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1186 // CHECK1-NEXT:  entry:
1187 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1188 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1189 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1190 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1191 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1192 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1193 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1194 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1195 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1196 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1197 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1198 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1199 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1200 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1201 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1202 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1203 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
1204 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1205 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1206 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1207 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1208 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1209 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1210 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1211 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1212 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1213 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1214 // 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)
1215 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1216 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1217 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1218 // CHECK1:       cond.true:
1219 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1220 // CHECK1:       cond.false:
1221 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1222 // CHECK1-NEXT:    br label [[COND_END]]
1223 // CHECK1:       cond.end:
1224 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1225 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1226 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1227 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1228 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1229 // CHECK1:       omp.inner.for.cond:
1230 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1231 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1232 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1233 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1234 // CHECK1:       omp.inner.for.body:
1235 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1236 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1237 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1238 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
1239 // CHECK1-NEXT:    call void @_Z3fn2v()
1240 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1241 // CHECK1:       omp.body.continue:
1242 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1243 // CHECK1:       omp.inner.for.inc:
1244 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1245 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1246 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1247 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1248 // CHECK1:       omp.inner.for.end:
1249 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1250 // CHECK1:       omp.loop.exit:
1251 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1252 // CHECK1-NEXT:    ret void
1253 //
1254 //
1255 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiEiT__l75
1256 // CHECK1-SAME: (i64 noundef [[ARG:%.*]]) #[[ATTR1]] {
1257 // CHECK1-NEXT:  entry:
1258 // CHECK1-NEXT:    [[ARG_ADDR:%.*]] = alloca i64, align 8
1259 // CHECK1-NEXT:    store i64 [[ARG]], i64* [[ARG_ADDR]], align 8
1260 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[ARG_ADDR]] to i32*
1261 // 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]])
1262 // CHECK1-NEXT:    ret void
1263 //
1264 //
1265 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..14
1266 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[ARG:%.*]]) #[[ATTR1]] {
1267 // CHECK1-NEXT:  entry:
1268 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1269 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1270 // CHECK1-NEXT:    [[ARG_ADDR:%.*]] = alloca i32*, align 8
1271 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1272 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1273 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1274 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1275 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1276 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1277 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1278 // CHECK1-NEXT:    [[DOTBOUND_ZERO_ADDR:%.*]] = alloca i32, align 4
1279 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1280 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1281 // CHECK1-NEXT:    store i32* [[ARG]], i32** [[ARG_ADDR]], align 8
1282 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[ARG_ADDR]], align 8
1283 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
1284 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_COMB_UB]], align 4
1285 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1286 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1287 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1288 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[TMP1]], align 4
1289 // 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)
1290 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1291 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP3]], 99
1292 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1293 // CHECK1:       cond.true:
1294 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1295 // CHECK1:       cond.false:
1296 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1297 // CHECK1-NEXT:    br label [[COND_END]]
1298 // CHECK1:       cond.end:
1299 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP4]], [[COND_FALSE]] ]
1300 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
1301 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1302 // CHECK1-NEXT:    store i32 [[TMP5]], i32* [[DOTOMP_IV]], align 4
1303 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1304 // CHECK1:       omp.inner.for.cond:
1305 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1306 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1307 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP6]], [[TMP7]]
1308 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1309 // CHECK1:       omp.inner.for.body:
1310 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
1311 // CHECK1-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP8]] to i64
1312 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
1313 // CHECK1-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
1314 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP0]], align 4
1315 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP12]], 0
1316 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_IF_THEN:%.*]], label [[OMP_IF_ELSE:%.*]]
1317 // CHECK1:       omp_if.then:
1318 // 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]])
1319 // CHECK1-NEXT:    br label [[OMP_IF_END:%.*]]
1320 // CHECK1:       omp_if.else:
1321 // CHECK1-NEXT:    call void @__kmpc_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
1322 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1323 // CHECK1-NEXT:    store i32 0, i32* [[DOTBOUND_ZERO_ADDR]], align 4
1324 // CHECK1-NEXT:    call void @.omp_outlined..15(i32* [[TMP13]], i32* [[DOTBOUND_ZERO_ADDR]], i64 [[TMP9]], i64 [[TMP11]]) #[[ATTR2]]
1325 // CHECK1-NEXT:    call void @__kmpc_end_serialized_parallel(%struct.ident_t* @[[GLOB3]], i32 [[TMP2]])
1326 // CHECK1-NEXT:    br label [[OMP_IF_END]]
1327 // CHECK1:       omp_if.end:
1328 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1329 // CHECK1:       omp.inner.for.inc:
1330 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1331 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1332 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP14]], [[TMP15]]
1333 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1334 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1335 // CHECK1:       omp.inner.for.end:
1336 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1337 // CHECK1:       omp.loop.exit:
1338 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP2]])
1339 // CHECK1-NEXT:    ret void
1340 //
1341 //
1342 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..15
1343 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
1344 // CHECK1-NEXT:  entry:
1345 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1346 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1347 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1348 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1349 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1350 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1351 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1352 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1353 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1354 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1355 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1356 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1357 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1358 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1359 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1360 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1361 // CHECK1-NEXT:    store i32 99, i32* [[DOTOMP_UB]], align 4
1362 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1363 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
1364 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1365 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
1366 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1367 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1368 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1369 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1370 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1371 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
1372 // 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)
1373 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1374 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 99
1375 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1376 // CHECK1:       cond.true:
1377 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1378 // CHECK1:       cond.false:
1379 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1380 // CHECK1-NEXT:    br label [[COND_END]]
1381 // CHECK1:       cond.end:
1382 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 99, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
1383 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1384 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1385 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
1386 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1387 // CHECK1:       omp.inner.for.cond:
1388 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1389 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1390 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
1391 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1392 // CHECK1:       omp.inner.for.body:
1393 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1394 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
1395 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1396 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
1397 // CHECK1-NEXT:    call void @_Z3fn3v()
1398 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1399 // CHECK1:       omp.body.continue:
1400 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1401 // CHECK1:       omp.inner.for.inc:
1402 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1403 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
1404 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1405 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1406 // CHECK1:       omp.inner.for.end:
1407 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1408 // CHECK1:       omp.loop.exit:
1409 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1410 // CHECK1-NEXT:    ret void
1411 //
1412 //
1413 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1414 // CHECK1-SAME: () #[[ATTR5:[0-9]+]] {
1415 // CHECK1-NEXT:  entry:
1416 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
1417 // CHECK1-NEXT:    ret void
1418 //
1419