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 
12 // expected-no-diagnostics
13 #ifndef HEADER
14 #define HEADER
15 
16 typedef __INTPTR_TYPE__ intptr_t;
17 
18 
19 void foo();
20 
21 struct S {
22   intptr_t a, b, c;
SS23   S(intptr_t a) : a(a) {}
operator charS24   operator char() { return a; }
~SS25   ~S() {}
26 };
27 
28 template <typename T>
tmain()29 T tmain() {
30 #pragma omp target 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 teams distribute parallel for proc_bind(spread)
37   for(int i = 0; i < 1000; i++) {}
38 #pragma omp target teams distribute parallel for proc_bind(close)
39   for(int i = 0; i < 1000; i++) {}
40   return tmain<int>();
41 }
42 
43 
44 
45 
46 
47 
48 
49 
50 #endif
51 // CHECK1-LABEL: define {{[^@]+}}@main
52 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
53 // CHECK1-NEXT:  entry:
54 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
55 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
56 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
57 // CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
58 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 1000)
59 // 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)
60 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
61 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
62 // CHECK1:       omp_offload.failed:
63 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36() #[[ATTR2:[0-9]+]]
64 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
65 // CHECK1:       omp_offload.cont:
66 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
67 // 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_l38.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
68 // CHECK1-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
69 // CHECK1-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
70 // CHECK1:       omp_offload.failed2:
71 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38() #[[ATTR2]]
72 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
73 // CHECK1:       omp_offload.cont3:
74 // CHECK1-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
75 // CHECK1-NEXT:    ret i32 [[CALL]]
76 //
77 //
78 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36
79 // CHECK1-SAME: () #[[ATTR1:[0-9]+]] {
80 // CHECK1-NEXT:  entry:
81 // 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*, ...)*))
82 // CHECK1-NEXT:    ret void
83 //
84 //
85 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
86 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
87 // CHECK1-NEXT:  entry:
88 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
89 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
90 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
91 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
92 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
93 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
94 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
95 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
96 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
97 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
98 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
99 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
100 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
101 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
102 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
103 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
104 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
105 // 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)
106 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
107 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
108 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
109 // CHECK1:       cond.true:
110 // CHECK1-NEXT:    br label [[COND_END:%.*]]
111 // CHECK1:       cond.false:
112 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
113 // CHECK1-NEXT:    br label [[COND_END]]
114 // CHECK1:       cond.end:
115 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
116 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
117 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
118 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
119 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
120 // CHECK1:       omp.inner.for.cond:
121 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
122 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
123 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
124 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
125 // CHECK1:       omp.inner.for.body:
126 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 4)
127 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
128 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
129 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
130 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
131 // 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]])
132 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
133 // CHECK1:       omp.inner.for.inc:
134 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
135 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
136 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
137 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
138 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
139 // CHECK1:       omp.inner.for.end:
140 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
141 // CHECK1:       omp.loop.exit:
142 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
143 // CHECK1-NEXT:    ret void
144 //
145 //
146 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
147 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
148 // CHECK1-NEXT:  entry:
149 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
150 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
151 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
152 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
153 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
154 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
155 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
156 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
157 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
158 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
159 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
160 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
161 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
162 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
163 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
164 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
165 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
166 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
167 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
168 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
169 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
170 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
171 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
172 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
173 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
174 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
175 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
176 // 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)
177 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
178 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
179 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
180 // CHECK1:       cond.true:
181 // CHECK1-NEXT:    br label [[COND_END:%.*]]
182 // CHECK1:       cond.false:
183 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
184 // CHECK1-NEXT:    br label [[COND_END]]
185 // CHECK1:       cond.end:
186 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
187 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
188 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
189 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
190 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
191 // CHECK1:       omp.inner.for.cond:
192 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
193 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
194 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
195 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
196 // CHECK1:       omp.inner.for.body:
197 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
198 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
199 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
200 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
201 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
202 // CHECK1:       omp.body.continue:
203 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
204 // CHECK1:       omp.inner.for.inc:
205 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
206 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
207 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
208 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
209 // CHECK1:       omp.inner.for.end:
210 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
211 // CHECK1:       omp.loop.exit:
212 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
213 // CHECK1-NEXT:    ret void
214 //
215 //
216 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38
217 // CHECK1-SAME: () #[[ATTR1]] {
218 // CHECK1-NEXT:  entry:
219 // 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*, ...)*))
220 // CHECK1-NEXT:    ret void
221 //
222 //
223 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
224 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
225 // CHECK1-NEXT:  entry:
226 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
227 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
228 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
229 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
230 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
231 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
232 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
233 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
234 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
235 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
236 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
237 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
238 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
239 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
240 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
241 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
242 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
243 // 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)
244 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
245 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
246 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
247 // CHECK1:       cond.true:
248 // CHECK1-NEXT:    br label [[COND_END:%.*]]
249 // CHECK1:       cond.false:
250 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
251 // CHECK1-NEXT:    br label [[COND_END]]
252 // CHECK1:       cond.end:
253 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
254 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
255 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
256 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
257 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
258 // CHECK1:       omp.inner.for.cond:
259 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
260 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
261 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
262 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
263 // CHECK1:       omp.inner.for.body:
264 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 3)
265 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
266 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
267 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
268 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
269 // 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]])
270 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
271 // CHECK1:       omp.inner.for.inc:
272 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
273 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
274 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
275 // CHECK1-NEXT:    store i32 [[ADD]], 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 [[TMP1]])
281 // CHECK1-NEXT:    ret void
282 //
283 //
284 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
285 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
286 // CHECK1-NEXT:  entry:
287 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
288 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
289 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
290 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
291 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
292 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
293 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
294 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
295 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
296 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
297 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
298 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
299 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
300 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
301 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
302 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
303 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
304 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
305 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
306 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
307 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
308 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
309 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
310 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
311 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
312 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
313 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
314 // 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)
315 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
316 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
317 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
318 // CHECK1:       cond.true:
319 // CHECK1-NEXT:    br label [[COND_END:%.*]]
320 // CHECK1:       cond.false:
321 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
322 // CHECK1-NEXT:    br label [[COND_END]]
323 // CHECK1:       cond.end:
324 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
325 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
326 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
327 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
328 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
329 // CHECK1:       omp.inner.for.cond:
330 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
331 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
332 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
333 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
334 // CHECK1:       omp.inner.for.body:
335 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
336 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
337 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
338 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
339 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
340 // CHECK1:       omp.body.continue:
341 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
342 // CHECK1:       omp.inner.for.inc:
343 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
344 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
345 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
346 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
347 // CHECK1:       omp.inner.for.end:
348 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
349 // CHECK1:       omp.loop.exit:
350 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
351 // CHECK1-NEXT:    ret void
352 //
353 //
354 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
355 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] comdat {
356 // CHECK1-NEXT:  entry:
357 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
358 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
359 // 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_l30.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
360 // CHECK1-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
361 // CHECK1-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
362 // CHECK1:       omp_offload.failed:
363 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l30() #[[ATTR2]]
364 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
365 // CHECK1:       omp_offload.cont:
366 // CHECK1-NEXT:    ret i32 0
367 //
368 //
369 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l30
370 // CHECK1-SAME: () #[[ATTR1]] {
371 // CHECK1-NEXT:  entry:
372 // 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*, ...)*))
373 // CHECK1-NEXT:    ret void
374 //
375 //
376 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
377 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
378 // CHECK1-NEXT:  entry:
379 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
380 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
381 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
382 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
383 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
384 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
385 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
386 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
387 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
388 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
389 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
390 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
391 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
392 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
393 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
394 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
395 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
396 // 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)
397 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
398 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
399 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
400 // CHECK1:       cond.true:
401 // CHECK1-NEXT:    br label [[COND_END:%.*]]
402 // CHECK1:       cond.false:
403 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
404 // CHECK1-NEXT:    br label [[COND_END]]
405 // CHECK1:       cond.end:
406 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
407 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
408 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
409 // CHECK1-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
410 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
411 // CHECK1:       omp.inner.for.cond:
412 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
413 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
414 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
415 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
416 // CHECK1:       omp.inner.for.body:
417 // CHECK1-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 2)
418 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
419 // CHECK1-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
420 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
421 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
422 // 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]])
423 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
424 // CHECK1:       omp.inner.for.inc:
425 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
426 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
427 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
428 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
429 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
430 // CHECK1:       omp.inner.for.end:
431 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
432 // CHECK1:       omp.loop.exit:
433 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
434 // CHECK1-NEXT:    ret void
435 //
436 //
437 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..5
438 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
439 // CHECK1-NEXT:  entry:
440 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
441 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
442 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
443 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
444 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
445 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
446 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
447 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
448 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
449 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
450 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
451 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
452 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
453 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
454 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
455 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
456 // CHECK1-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
457 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
458 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
459 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
460 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
461 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
462 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
463 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
464 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
465 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
466 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
467 // 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)
468 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
469 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
470 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
471 // CHECK1:       cond.true:
472 // CHECK1-NEXT:    br label [[COND_END:%.*]]
473 // CHECK1:       cond.false:
474 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
475 // CHECK1-NEXT:    br label [[COND_END]]
476 // CHECK1:       cond.end:
477 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
478 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
479 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
480 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
481 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
482 // CHECK1:       omp.inner.for.cond:
483 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
484 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
485 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
486 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
487 // CHECK1:       omp.inner.for.body:
488 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
489 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
490 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
491 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
492 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
493 // CHECK1:       omp.body.continue:
494 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
495 // CHECK1:       omp.inner.for.inc:
496 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
497 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
498 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
499 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
500 // CHECK1:       omp.inner.for.end:
501 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
502 // CHECK1:       omp.loop.exit:
503 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
504 // CHECK1-NEXT:    ret void
505 //
506 //
507 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
508 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] {
509 // CHECK1-NEXT:  entry:
510 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
511 // CHECK1-NEXT:    ret void
512 //
513 //
514 // CHECK2-LABEL: define {{[^@]+}}@main
515 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
516 // CHECK2-NEXT:  entry:
517 // CHECK2-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
518 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
519 // CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
520 // CHECK2-NEXT:    store i32 0, i32* [[RETVAL]], align 4
521 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 1000)
522 // 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)
523 // CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
524 // CHECK2-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
525 // CHECK2:       omp_offload.failed:
526 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36() #[[ATTR2:[0-9]+]]
527 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
528 // CHECK2:       omp_offload.cont:
529 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
530 // 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_l38.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
531 // CHECK2-NEXT:    [[TMP3:%.*]] = icmp ne i32 [[TMP2]], 0
532 // CHECK2-NEXT:    br i1 [[TMP3]], label [[OMP_OFFLOAD_FAILED2:%.*]], label [[OMP_OFFLOAD_CONT3:%.*]]
533 // CHECK2:       omp_offload.failed2:
534 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38() #[[ATTR2]]
535 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT3]]
536 // CHECK2:       omp_offload.cont3:
537 // CHECK2-NEXT:    [[CALL:%.*]] = call signext i32 @_Z5tmainIiET_v()
538 // CHECK2-NEXT:    ret i32 [[CALL]]
539 //
540 //
541 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l36
542 // CHECK2-SAME: () #[[ATTR1:[0-9]+]] {
543 // CHECK2-NEXT:  entry:
544 // 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*, ...)*))
545 // CHECK2-NEXT:    ret void
546 //
547 //
548 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
549 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
550 // CHECK2-NEXT:  entry:
551 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
552 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
553 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
554 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
555 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
556 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
557 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
558 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
559 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
560 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
561 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
562 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
563 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
564 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
565 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
566 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
567 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
568 // 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)
569 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
570 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
571 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
572 // CHECK2:       cond.true:
573 // CHECK2-NEXT:    br label [[COND_END:%.*]]
574 // CHECK2:       cond.false:
575 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
576 // CHECK2-NEXT:    br label [[COND_END]]
577 // CHECK2:       cond.end:
578 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
579 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
580 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
581 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
582 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
583 // CHECK2:       omp.inner.for.cond:
584 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
585 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
586 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
587 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
588 // CHECK2:       omp.inner.for.body:
589 // CHECK2-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 4)
590 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
591 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
592 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
593 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
594 // 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]])
595 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
596 // CHECK2:       omp.inner.for.inc:
597 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
598 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
599 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
600 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
601 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
602 // CHECK2:       omp.inner.for.end:
603 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
604 // CHECK2:       omp.loop.exit:
605 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
606 // CHECK2-NEXT:    ret void
607 //
608 //
609 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
610 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
611 // CHECK2-NEXT:  entry:
612 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
613 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
614 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
615 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
616 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
617 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
618 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
619 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
620 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
621 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
622 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
623 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
624 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
625 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
626 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
627 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
628 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
629 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
630 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
631 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
632 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
633 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
634 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
635 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
636 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
637 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
638 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
639 // 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)
640 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
641 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
642 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
643 // CHECK2:       cond.true:
644 // CHECK2-NEXT:    br label [[COND_END:%.*]]
645 // CHECK2:       cond.false:
646 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
647 // CHECK2-NEXT:    br label [[COND_END]]
648 // CHECK2:       cond.end:
649 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
650 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
651 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
652 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
653 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
654 // CHECK2:       omp.inner.for.cond:
655 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
656 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
657 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
658 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
659 // CHECK2:       omp.inner.for.body:
660 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
661 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
662 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
663 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
664 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
665 // CHECK2:       omp.body.continue:
666 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
667 // CHECK2:       omp.inner.for.inc:
668 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
669 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
670 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
671 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
672 // CHECK2:       omp.inner.for.end:
673 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
674 // CHECK2:       omp.loop.exit:
675 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
676 // CHECK2-NEXT:    ret void
677 //
678 //
679 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38
680 // CHECK2-SAME: () #[[ATTR1]] {
681 // CHECK2-NEXT:  entry:
682 // 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*, ...)*))
683 // CHECK2-NEXT:    ret void
684 //
685 //
686 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
687 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
688 // CHECK2-NEXT:  entry:
689 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
690 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
691 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
692 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
693 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
694 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
695 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
696 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
697 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
698 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
699 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
700 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
701 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
702 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
703 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
704 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
705 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
706 // 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)
707 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
708 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
709 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
710 // CHECK2:       cond.true:
711 // CHECK2-NEXT:    br label [[COND_END:%.*]]
712 // CHECK2:       cond.false:
713 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
714 // CHECK2-NEXT:    br label [[COND_END]]
715 // CHECK2:       cond.end:
716 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
717 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
718 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
719 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
720 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
721 // CHECK2:       omp.inner.for.cond:
722 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
723 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
724 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
725 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
726 // CHECK2:       omp.inner.for.body:
727 // CHECK2-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 3)
728 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
729 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
730 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
731 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
732 // 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]])
733 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
734 // CHECK2:       omp.inner.for.inc:
735 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
736 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
737 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
738 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
739 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
740 // CHECK2:       omp.inner.for.end:
741 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
742 // CHECK2:       omp.loop.exit:
743 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
744 // CHECK2-NEXT:    ret void
745 //
746 //
747 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
748 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
749 // CHECK2-NEXT:  entry:
750 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
751 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
752 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
753 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
754 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
755 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
756 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
757 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
758 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
759 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
760 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
761 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
762 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
763 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
764 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
765 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
766 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
767 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
768 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
769 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
770 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
771 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
772 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
773 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
774 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
775 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
776 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
777 // 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)
778 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
779 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
780 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
781 // CHECK2:       cond.true:
782 // CHECK2-NEXT:    br label [[COND_END:%.*]]
783 // CHECK2:       cond.false:
784 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
785 // CHECK2-NEXT:    br label [[COND_END]]
786 // CHECK2:       cond.end:
787 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
788 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
789 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
790 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
791 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
792 // CHECK2:       omp.inner.for.cond:
793 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
794 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
795 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
796 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
797 // CHECK2:       omp.inner.for.body:
798 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
799 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
800 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
801 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
802 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
803 // CHECK2:       omp.body.continue:
804 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
805 // CHECK2:       omp.inner.for.inc:
806 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
807 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
808 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
809 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
810 // CHECK2:       omp.inner.for.end:
811 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
812 // CHECK2:       omp.loop.exit:
813 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
814 // CHECK2-NEXT:    ret void
815 //
816 //
817 // CHECK2-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
818 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] comdat {
819 // CHECK2-NEXT:  entry:
820 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
821 // CHECK2-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 1000)
822 // 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_l30.region_id, i32 0, i8** null, i8** null, i64* null, i64* null, i8** null, i8** null, i32 0, i32 0)
823 // CHECK2-NEXT:    [[TMP1:%.*]] = icmp ne i32 [[TMP0]], 0
824 // CHECK2-NEXT:    br i1 [[TMP1]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
825 // CHECK2:       omp_offload.failed:
826 // CHECK2-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l30() #[[ATTR2]]
827 // CHECK2-NEXT:    br label [[OMP_OFFLOAD_CONT]]
828 // CHECK2:       omp_offload.cont:
829 // CHECK2-NEXT:    ret i32 0
830 //
831 //
832 // CHECK2-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l30
833 // CHECK2-SAME: () #[[ATTR1]] {
834 // CHECK2-NEXT:  entry:
835 // 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*, ...)*))
836 // CHECK2-NEXT:    ret void
837 //
838 //
839 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
840 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] {
841 // CHECK2-NEXT:  entry:
842 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
843 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
844 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
845 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
846 // CHECK2-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
847 // CHECK2-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
848 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
849 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
850 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
851 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
852 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
853 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
854 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_COMB_UB]], align 4
855 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
856 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
857 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
858 // CHECK2-NEXT:    [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 4
859 // 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)
860 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
861 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP2]], 999
862 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
863 // CHECK2:       cond.true:
864 // CHECK2-NEXT:    br label [[COND_END:%.*]]
865 // CHECK2:       cond.false:
866 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
867 // CHECK2-NEXT:    br label [[COND_END]]
868 // CHECK2:       cond.end:
869 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP3]], [[COND_FALSE]] ]
870 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
871 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
872 // CHECK2-NEXT:    store i32 [[TMP4]], i32* [[DOTOMP_IV]], align 4
873 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
874 // CHECK2:       omp.inner.for.cond:
875 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
876 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
877 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP5]], [[TMP6]]
878 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
879 // CHECK2:       omp.inner.for.body:
880 // CHECK2-NEXT:    call void @__kmpc_push_proc_bind(%struct.ident_t* @[[GLOB3]], i32 [[TMP1]], i32 2)
881 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
882 // CHECK2-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP7]] to i64
883 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
884 // CHECK2-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
885 // 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]])
886 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
887 // CHECK2:       omp.inner.for.inc:
888 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
889 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
890 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
891 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
892 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
893 // CHECK2:       omp.inner.for.end:
894 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
895 // CHECK2:       omp.loop.exit:
896 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP1]])
897 // CHECK2-NEXT:    ret void
898 //
899 //
900 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..5
901 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], i64 [[DOTPREVIOUS_LB_:%.*]], i64 [[DOTPREVIOUS_UB_:%.*]]) #[[ATTR1]] {
902 // CHECK2-NEXT:  entry:
903 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
904 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
905 // CHECK2-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
906 // CHECK2-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
907 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
908 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
909 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
910 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
911 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
912 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
913 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
914 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
915 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
916 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
917 // CHECK2-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
918 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
919 // CHECK2-NEXT:    store i32 999, i32* [[DOTOMP_UB]], align 4
920 // CHECK2-NEXT:    [[TMP0:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
921 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
922 // CHECK2-NEXT:    [[TMP1:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
923 // CHECK2-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP1]] to i32
924 // CHECK2-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
925 // CHECK2-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
926 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
927 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
928 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
929 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
930 // 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)
931 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
932 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 999
933 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
934 // CHECK2:       cond.true:
935 // CHECK2-NEXT:    br label [[COND_END:%.*]]
936 // CHECK2:       cond.false:
937 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
938 // CHECK2-NEXT:    br label [[COND_END]]
939 // CHECK2:       cond.end:
940 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 999, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
941 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
942 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
943 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
944 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
945 // CHECK2:       omp.inner.for.cond:
946 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
947 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
948 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
949 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
950 // CHECK2:       omp.inner.for.body:
951 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
952 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
953 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
954 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
955 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
956 // CHECK2:       omp.body.continue:
957 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
958 // CHECK2:       omp.inner.for.inc:
959 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
960 // CHECK2-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP10]], 1
961 // CHECK2-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
962 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
963 // CHECK2:       omp.inner.for.end:
964 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
965 // CHECK2:       omp.loop.exit:
966 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
967 // CHECK2-NEXT:    ret void
968 //
969 //
970 // CHECK2-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
971 // CHECK2-SAME: () #[[ATTR4:[0-9]+]] {
972 // CHECK2-NEXT:  entry:
973 // CHECK2-NEXT:    call void @__tgt_register_requires(i64 1)
974 // CHECK2-NEXT:    ret void
975 //
976 //