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 // add -fopenmp-targets
3 
4 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
5 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
6 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK2
7 
8 // RUN: %clang_cc1 -verify -fopenmp-simd -fopenmp-version=45 -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
9 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
10 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
11 // expected-no-diagnostics
12 #ifndef HEADER
13 #define HEADER
14 
15 typedef __INTPTR_TYPE__ intptr_t;
16 
17 
18 void foo();
19 
20 struct S {
21   intptr_t a, b, c;
SS22   S(intptr_t a) : a(a) {}
operator charS23   operator char() { return a; }
~SS24   ~S() {}
25 };
26 
27 template <typename T>
tmain()28 T tmain() {
29 #pragma omp target
30 #pragma omp teams distribute parallel for proc_bind(master)
31   for(int i = 0; i < 1000; i++) {}
32   return T();
33 }
34 
main()35 int main() {
36 #pragma omp target
37 #pragma omp teams distribute parallel for proc_bind(spread)
38   for(int i = 0; i < 1000; i++) {}
39 #pragma omp target
40 #pragma omp teams distribute parallel for proc_bind(close)
41   for(int i = 0; i < 1000; i++) {}
42   return tmain<int>();
43 }
44 
45 
46 
47 
48 
49 
50 
51 
52 #endif
53 // CHECK1-LABEL: define {{[^@]+}}@main
54 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
55 // CHECK1-NEXT:  entry:
56 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
57 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
58 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
59 // CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
60 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 1000)
61 // 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_l36.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
62 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
63 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
64 // CHECK1:       omp_offload.failed:
65 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36() #[[ATTR2:[0-9]+]]
66 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
67 // CHECK1:       omp_offload.cont:
68 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
69 // 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_l39.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
70 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
71 // CHECK1-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
72 // CHECK1:       omp_offload.failed2:
73 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l39() #[[ATTR2]]
74 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
75 // CHECK1:       omp_offload.cont3:
76 // CHECK1-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
77 // CHECK1-NEXT:    ret i32 [[CALL]]
78 //
79 //
80 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36
81 // CHECK1-SAME: () #[[ATTR1:[0-9]+]] {
82 // CHECK1-NEXT:  entry:
83 // 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*, ...)*))
84 // CHECK1-NEXT:    ret void
85 //
86 //
87 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
88 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
89 // CHECK1-NEXT:  entry:
90 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
91 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
92 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
93 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
94 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
95 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
96 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
97 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
98 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
99 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
100 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
101 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
102 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
103 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
104 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
105 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
106 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
107 // 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)
108 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
109 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
110 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
111 // CHECK1:       cond.true:
112 // CHECK1-NEXT:    br label [[COND_END:%.*]]
113 // CHECK1:       cond.false:
114 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
115 // CHECK1-NEXT:    br label [[COND_END]]
116 // CHECK1:       cond.end:
117 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
118 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
119 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
120 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
121 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
122 // CHECK1:       omp.inner.for.cond:
123 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
124 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
125 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
126 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
127 // CHECK1:       omp.inner.for.body:
128 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 4)
129 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
130 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
131 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
132 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
133 // 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]])
134 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
135 // CHECK1:       omp.inner.for.inc:
136 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
137 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
138 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
139 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
140 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
141 // CHECK1:       omp.inner.for.end:
142 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
143 // CHECK1:       omp.loop.exit:
144 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
145 // CHECK1-NEXT:    ret void
146 //
147 //
148 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
149 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
150 // CHECK1-NEXT:  entry:
151 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
152 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
153 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
154 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
155 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
156 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
157 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
158 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
159 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
160 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
161 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
162 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
163 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
164 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
165 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
166 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
167 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
168 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
169 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
170 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
171 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
172 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
173 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
174 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
175 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
176 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
177 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
178 // 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)
179 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
180 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
181 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
182 // CHECK1:       cond.true:
183 // CHECK1-NEXT:    br label [[COND_END:%.*]]
184 // CHECK1:       cond.false:
185 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
186 // CHECK1-NEXT:    br label [[COND_END]]
187 // CHECK1:       cond.end:
188 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
189 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
190 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
191 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
192 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
193 // CHECK1:       omp.inner.for.cond:
194 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
195 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
196 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
197 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
198 // CHECK1:       omp.inner.for.body:
199 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
200 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
201 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
202 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
203 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
204 // CHECK1:       omp.body.continue:
205 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
206 // CHECK1:       omp.inner.for.inc:
207 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
208 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
209 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
210 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
211 // CHECK1:       omp.inner.for.end:
212 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
213 // CHECK1:       omp.loop.exit:
214 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
215 // CHECK1-NEXT:    ret void
216 //
217 //
218 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l39
219 // CHECK1-SAME: () #[[ATTR1]] {
220 // CHECK1-NEXT:  entry:
221 // 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*, ...)*))
222 // CHECK1-NEXT:    ret void
223 //
224 //
225 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
226 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
227 // CHECK1-NEXT:  entry:
228 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
229 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
230 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
231 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
232 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
233 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
234 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
235 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
236 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
237 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
238 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
239 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
240 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
241 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
242 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
243 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
244 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
245 // 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)
246 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
247 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
248 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
249 // CHECK1:       cond.true:
250 // CHECK1-NEXT:    br label [[COND_END:%.*]]
251 // CHECK1:       cond.false:
252 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
253 // CHECK1-NEXT:    br label [[COND_END]]
254 // CHECK1:       cond.end:
255 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
256 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
257 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
258 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
259 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
260 // CHECK1:       omp.inner.for.cond:
261 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
262 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
263 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
264 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
265 // CHECK1:       omp.inner.for.body:
266 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 3)
267 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
268 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
269 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
270 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
271 // 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..3 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
272 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
273 // CHECK1:       omp.inner.for.inc:
274 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
275 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
276 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
277 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
278 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
279 // CHECK1:       omp.inner.for.end:
280 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
281 // CHECK1:       omp.loop.exit:
282 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
283 // CHECK1-NEXT:    ret void
284 //
285 //
286 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
287 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
288 // CHECK1-NEXT:  entry:
289 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
290 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
291 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
292 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
293 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
294 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
295 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
296 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
297 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
298 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
299 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
300 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
301 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
302 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
303 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
304 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
305 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
306 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
307 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
308 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
309 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
310 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
311 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
312 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
313 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
314 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
315 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
316 // 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)
317 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
318 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
319 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
320 // CHECK1:       cond.true:
321 // CHECK1-NEXT:    br label [[COND_END:%.*]]
322 // CHECK1:       cond.false:
323 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
324 // CHECK1-NEXT:    br label [[COND_END]]
325 // CHECK1:       cond.end:
326 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
327 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
328 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
329 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
330 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
331 // CHECK1:       omp.inner.for.cond:
332 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
333 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
334 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
335 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
336 // CHECK1:       omp.inner.for.body:
337 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
338 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
339 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
340 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
341 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
342 // CHECK1:       omp.body.continue:
343 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
344 // CHECK1:       omp.inner.for.inc:
345 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
346 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
347 // CHECK1-NEXT:    store i32 [[ADD3]], 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 [[TMP3]])
353 // CHECK1-NEXT:    ret void
354 //
355 //
356 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
357 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] comdat {
358 // CHECK1-NEXT:  entry:
359 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
360 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
361 // CHECK1-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
362 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
363 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
364 // CHECK1:       omp_offload.failed:
365 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29() #[[ATTR2]]
366 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
367 // CHECK1:       omp_offload.cont:
368 // CHECK1-NEXT:    ret i32 0
369 //
370 //
371 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29
372 // CHECK1-SAME: () #[[ATTR1]] {
373 // CHECK1-NEXT:  entry:
374 // 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*, ...)*))
375 // CHECK1-NEXT:    ret void
376 //
377 //
378 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
379 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
380 // CHECK1-NEXT:  entry:
381 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
382 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
383 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
384 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
385 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
386 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
387 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
388 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
389 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
390 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
391 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
392 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
393 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
394 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
395 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
396 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
397 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
398 // 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)
399 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
400 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
401 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
402 // CHECK1:       cond.true:
403 // CHECK1-NEXT:    br label [[COND_END:%.*]]
404 // CHECK1:       cond.false:
405 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
406 // CHECK1-NEXT:    br label [[COND_END]]
407 // CHECK1:       cond.end:
408 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
409 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
410 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
411 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
412 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
413 // CHECK1:       omp.inner.for.cond:
414 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
415 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
416 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
417 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
418 // CHECK1:       omp.inner.for.body:
419 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 2)
420 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
421 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
422 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
423 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
424 // 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]])
425 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
426 // CHECK1:       omp.inner.for.inc:
427 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
428 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
429 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
430 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
431 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
432 // CHECK1:       omp.inner.for.end:
433 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
434 // CHECK1:       omp.loop.exit:
435 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
436 // CHECK1-NEXT:    ret void
437 //
438 //
439 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..5
440 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
441 // CHECK1-NEXT:  entry:
442 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
443 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
444 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
445 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
446 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
447 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
448 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
449 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
450 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
451 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
452 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
453 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
454 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
455 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
456 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
457 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
458 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
459 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
460 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
461 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
462 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
463 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
464 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
465 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
466 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
467 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
468 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
469 // 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)
470 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
471 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
472 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
473 // CHECK1:       cond.true:
474 // CHECK1-NEXT:    br label [[COND_END:%.*]]
475 // CHECK1:       cond.false:
476 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
477 // CHECK1-NEXT:    br label [[COND_END]]
478 // CHECK1:       cond.end:
479 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
480 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
481 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
482 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
483 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
484 // CHECK1:       omp.inner.for.cond:
485 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
486 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
487 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
488 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
489 // CHECK1:       omp.inner.for.body:
490 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
491 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
492 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
493 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
494 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
495 // CHECK1:       omp.body.continue:
496 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
497 // CHECK1:       omp.inner.for.inc:
498 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
499 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
500 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
501 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
502 // CHECK1:       omp.inner.for.end:
503 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
504 // CHECK1:       omp.loop.exit:
505 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
506 // CHECK1-NEXT:    ret void
507 //
508 //
509 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
510 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
511 // CHECK1-NEXT:  entry:
512 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
513 // CHECK1-NEXT:    ret void
514 //
515 //
516 // CHECK2-LABEL: define {{[^@]+}}@main
517 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
518 // CHECK2-NEXT:  entry:
519 // CHECK2-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
520 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
521 // CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
522 // CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
523 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 1000)
524 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
525 // CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
526 // CHECK2-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
527 // CHECK2:       omp_offload.failed:
528 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36() #[[ATTR2:[0-9]+]]
529 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
530 // CHECK2:       omp_offload.cont:
531 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
532 // CHECK2-NEXT:    [[TMP2:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l39.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
533 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
534 // CHECK2-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
535 // CHECK2:       omp_offload.failed2:
536 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l39() #[[ATTR2]]
537 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
538 // CHECK2:       omp_offload.cont3:
539 // CHECK2-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
540 // CHECK2-NEXT:    ret i32 [[CALL]]
541 //
542 //
543 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36
544 // CHECK2-SAME: () #[[ATTR1:[0-9]+]] {
545 // CHECK2-NEXT:  entry:
546 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*))
547 // CHECK2-NEXT:    ret void
548 //
549 //
550 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
551 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
552 // CHECK2-NEXT:  entry:
553 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
554 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
555 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
556 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
557 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
558 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
559 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
560 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
561 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
562 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
563 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
564 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
565 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
566 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
567 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
568 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
569 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
570 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
571 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
572 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
573 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
574 // CHECK2:       cond.true:
575 // CHECK2-NEXT:    br label [[COND_END:%.*]]
576 // CHECK2:       cond.false:
577 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
578 // CHECK2-NEXT:    br label [[COND_END]]
579 // CHECK2:       cond.end:
580 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
581 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
582 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
583 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
584 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
585 // CHECK2:       omp.inner.for.cond:
586 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
587 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
588 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
589 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
590 // CHECK2:       omp.inner.for.body:
591 // CHECK2-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 4)
592 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
593 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
594 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
595 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
596 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
597 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
598 // CHECK2:       omp.inner.for.inc:
599 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
600 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
601 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
602 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
603 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
604 // CHECK2:       omp.inner.for.end:
605 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
606 // CHECK2:       omp.loop.exit:
607 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
608 // CHECK2-NEXT:    ret void
609 //
610 //
611 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
612 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
613 // CHECK2-NEXT:  entry:
614 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
615 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
616 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
617 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
618 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
619 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
620 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
621 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
622 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
623 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
624 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
625 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
626 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
627 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
628 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
629 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
630 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
631 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
632 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
633 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
634 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
635 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
636 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
637 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
638 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
639 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
640 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
641 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
642 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
643 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
644 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
645 // CHECK2:       cond.true:
646 // CHECK2-NEXT:    br label [[COND_END:%.*]]
647 // CHECK2:       cond.false:
648 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
649 // CHECK2-NEXT:    br label [[COND_END]]
650 // CHECK2:       cond.end:
651 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
652 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
653 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
654 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
655 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
656 // CHECK2:       omp.inner.for.cond:
657 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
658 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
659 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
660 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
661 // CHECK2:       omp.inner.for.body:
662 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
663 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
664 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
665 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
666 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
667 // CHECK2:       omp.body.continue:
668 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
669 // CHECK2:       omp.inner.for.inc:
670 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
671 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
672 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
673 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
674 // CHECK2:       omp.inner.for.end:
675 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
676 // CHECK2:       omp.loop.exit:
677 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
678 // CHECK2-NEXT:    ret void
679 //
680 //
681 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l39
682 // CHECK2-SAME: () #[[ATTR1]] {
683 // CHECK2-NEXT:  entry:
684 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*))
685 // CHECK2-NEXT:    ret void
686 //
687 //
688 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
689 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
690 // CHECK2-NEXT:  entry:
691 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
692 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
693 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
694 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
695 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
696 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
697 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
698 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
699 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
700 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
701 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
702 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
703 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
704 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
705 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
706 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
707 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
708 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
709 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
710 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
711 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
712 // CHECK2:       cond.true:
713 // CHECK2-NEXT:    br label [[COND_END:%.*]]
714 // CHECK2:       cond.false:
715 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
716 // CHECK2-NEXT:    br label [[COND_END]]
717 // CHECK2:       cond.end:
718 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
719 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
720 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
721 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
722 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
723 // CHECK2:       omp.inner.for.cond:
724 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
725 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
726 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
727 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
728 // CHECK2:       omp.inner.for.body:
729 // CHECK2-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 3)
730 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
731 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
732 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
733 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
734 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
735 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
736 // CHECK2:       omp.inner.for.inc:
737 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
738 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
739 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
740 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
741 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
742 // CHECK2:       omp.inner.for.end:
743 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
744 // CHECK2:       omp.loop.exit:
745 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
746 // CHECK2-NEXT:    ret void
747 //
748 //
749 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
750 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
751 // CHECK2-NEXT:  entry:
752 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
753 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
754 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
755 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
756 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
757 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
758 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
759 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
760 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
761 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
762 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
763 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
764 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
765 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
766 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
767 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
768 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
769 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
770 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
771 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
772 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
773 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
774 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
775 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
776 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
777 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
778 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
779 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
780 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
781 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
782 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
783 // CHECK2:       cond.true:
784 // CHECK2-NEXT:    br label [[COND_END:%.*]]
785 // CHECK2:       cond.false:
786 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
787 // CHECK2-NEXT:    br label [[COND_END]]
788 // CHECK2:       cond.end:
789 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
790 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
791 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
792 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
793 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
794 // CHECK2:       omp.inner.for.cond:
795 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
796 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
797 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
798 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
799 // CHECK2:       omp.inner.for.body:
800 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
801 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
802 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
803 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
804 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
805 // CHECK2:       omp.body.continue:
806 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
807 // CHECK2:       omp.inner.for.inc:
808 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
809 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
810 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
811 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
812 // CHECK2:       omp.inner.for.end:
813 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
814 // CHECK2:       omp.loop.exit:
815 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
816 // CHECK2-NEXT:    ret void
817 //
818 //
819 // CHECK2-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
820 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] comdat {
821 // CHECK2-NEXT:  entry:
822 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
823 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
824 // CHECK2-NEXT:    [[TMP0:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
825 // CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
826 // CHECK2-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
827 // CHECK2:       omp_offload.failed:
828 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29() #[[ATTR2]]
829 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
830 // CHECK2:       omp_offload.cont:
831 // CHECK2-NEXT:    ret i32 0
832 //
833 //
834 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l29
835 // CHECK2-SAME: () #[[ATTR1]] {
836 // CHECK2-NEXT:  entry:
837 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 0, void (i32*, i32*, ...)* bitcast (void (i32*, i32*)* @.omp_outlined..4 to void (i32*, i32*, ...)*))
838 // CHECK2-NEXT:    ret void
839 //
840 //
841 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
842 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
843 // CHECK2-NEXT:  entry:
844 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
845 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
846 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
847 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
848 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
849 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
850 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
851 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
852 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
853 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
854 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
855 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
856 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
857 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
858 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
859 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
860 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
861 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
862 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
863 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
864 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
865 // CHECK2:       cond.true:
866 // CHECK2-NEXT:    br label [[COND_END:%.*]]
867 // CHECK2:       cond.false:
868 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
869 // CHECK2-NEXT:    br label [[COND_END]]
870 // CHECK2:       cond.end:
871 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
872 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
873 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
874 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
875 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
876 // CHECK2:       omp.inner.for.cond:
877 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
878 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
879 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
880 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
881 // CHECK2:       omp.inner.for.body:
882 // CHECK2-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 2)
883 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
884 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
885 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
886 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
887 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64)* @.omp_outlined..5 to void (i32*, i32*, ...)*), i64 [[TMP8]], i64 [[TMP10]])
888 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
889 // CHECK2:       omp.inner.for.inc:
890 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
891 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
892 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
893 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
894 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
895 // CHECK2:       omp.inner.for.end:
896 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
897 // CHECK2:       omp.loop.exit:
898 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
899 // CHECK2-NEXT:    ret void
900 //
901 //
902 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..5
903 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
904 // CHECK2-NEXT:  entry:
905 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
906 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
907 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
908 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
909 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
910 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
911 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
912 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
913 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
914 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
915 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
916 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
917 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
918 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
919 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
920 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
921 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
922 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
923 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
924 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
925 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
926 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
927 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
928 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
929 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
930 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
931 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
932 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
933 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
934 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
935 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
936 // CHECK2:       cond.true:
937 // CHECK2-NEXT:    br label [[COND_END:%.*]]
938 // CHECK2:       cond.false:
939 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
940 // CHECK2-NEXT:    br label [[COND_END]]
941 // CHECK2:       cond.end:
942 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
943 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
944 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
945 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
946 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
947 // CHECK2:       omp.inner.for.cond:
948 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
949 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
950 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
951 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
952 // CHECK2:       omp.inner.for.body:
953 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
954 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
955 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
956 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
957 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
958 // CHECK2:       omp.body.continue:
959 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
960 // CHECK2:       omp.inner.for.inc:
961 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
962 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
963 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
964 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
965 // CHECK2:       omp.inner.for.end:
966 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
967 // CHECK2:       omp.loop.exit:
968 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
969 // CHECK2-NEXT:    ret void
970 //
971 //
972 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
973 // CHECK2-SAME: () #[[ATTR4:[0-9]+]] {
974 // CHECK2-NEXT:  entry:
975 // CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
976 // CHECK2-NEXT:    ret void
977 //
978 //