1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
2 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
4 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK2
5 // RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
6 // RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK4
7 
8 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
9 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
10 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
11 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DLAMBDA -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
12 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -fblocks -DBLOCKS -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
13 // expected-no-diagnostics
14 #ifndef HEADER
15 #define HEADER
16 
foo(float * c)17 void foo(float *c) {
18 #pragma omp parallel for simd aligned(c)
19   for (int i = 0; i < 10; ++i);
20 }
21 
22 #endif
23 
24 // CHECK1-LABEL: define {{[^@]+}}@_Z3fooPf
25 // CHECK1-SAME: (float* [[C:%.*]]) #[[ATTR0:[0-9]+]] {
26 // CHECK1-NEXT:  entry:
27 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
28 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
29 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**)* @.omp_outlined. to void (i32*, i32*, ...)*), float** [[C_ADDR]])
30 // CHECK1-NEXT:    ret void
31 //
32 //
33 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
34 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], float** nonnull align 8 dereferenceable(8) [[C:%.*]]) #[[ATTR1:[0-9]+]] {
35 // CHECK1-NEXT:  entry:
36 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
37 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
38 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca float**, align 8
39 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
40 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
41 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
42 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
43 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
44 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
45 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
46 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
47 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
48 // CHECK1-NEXT:    store float** [[C]], float*** [[C_ADDR]], align 8
49 // CHECK1-NEXT:    [[TMP0:%.*]] = load float**, float*** [[C_ADDR]], align 8
50 // CHECK1-NEXT:    [[TMP1:%.*]] = load float*, float** [[TMP0]], align 8
51 // CHECK1-NEXT:    call void @llvm.assume(i1 true) [ "align"(float* [[TMP1]], i64 16) ]
52 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
53 // CHECK1-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
54 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
55 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
56 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
57 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
58 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
59 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
60 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
61 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
62 // CHECK1:       cond.true:
63 // CHECK1-NEXT:    br label [[COND_END:%.*]]
64 // CHECK1:       cond.false:
65 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
66 // CHECK1-NEXT:    br label [[COND_END]]
67 // CHECK1:       cond.end:
68 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
69 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
70 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
71 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
72 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
73 // CHECK1:       omp.inner.for.cond:
74 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
75 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
76 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
77 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
78 // CHECK1:       omp.inner.for.body:
79 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
80 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
81 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
82 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
83 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
84 // CHECK1:       omp.body.continue:
85 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
86 // CHECK1:       omp.inner.for.inc:
87 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
88 // CHECK1-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
89 // CHECK1-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
90 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
91 // CHECK1:       omp.inner.for.end:
92 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
93 // CHECK1:       omp.loop.exit:
94 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
95 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
96 // CHECK1-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
97 // CHECK1-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
98 // CHECK1:       .omp.final.then:
99 // CHECK1-NEXT:    store i32 10, i32* [[I]], align 4
100 // CHECK1-NEXT:    br label [[DOTOMP_FINAL_DONE]]
101 // CHECK1:       .omp.final.done:
102 // CHECK1-NEXT:    ret void
103 //
104 //
105 // CHECK2-LABEL: define {{[^@]+}}@_Z3fooPf
106 // CHECK2-SAME: (float* [[C:%.*]]) #[[ATTR0:[0-9]+]] {
107 // CHECK2-NEXT:  entry:
108 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
109 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
110 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**)* @.omp_outlined. to void (i32*, i32*, ...)*), float** [[C_ADDR]])
111 // CHECK2-NEXT:    ret void
112 //
113 //
114 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
115 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], float** nonnull align 8 dereferenceable(8) [[C:%.*]]) #[[ATTR1:[0-9]+]] {
116 // CHECK2-NEXT:  entry:
117 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
118 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
119 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca float**, align 8
120 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
121 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
122 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
123 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
124 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
125 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
126 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
127 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
128 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
129 // CHECK2-NEXT:    store float** [[C]], float*** [[C_ADDR]], align 8
130 // CHECK2-NEXT:    [[TMP0:%.*]] = load float**, float*** [[C_ADDR]], align 8
131 // CHECK2-NEXT:    [[TMP1:%.*]] = load float*, float** [[TMP0]], align 8
132 // CHECK2-NEXT:    call void @llvm.assume(i1 true) [ "align"(float* [[TMP1]], i64 16) ]
133 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
134 // CHECK2-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
135 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
136 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
137 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
138 // CHECK2-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
139 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
140 // CHECK2-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
141 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
142 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
143 // CHECK2:       cond.true:
144 // CHECK2-NEXT:    br label [[COND_END:%.*]]
145 // CHECK2:       cond.false:
146 // CHECK2-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
147 // CHECK2-NEXT:    br label [[COND_END]]
148 // CHECK2:       cond.end:
149 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
150 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
151 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
152 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
153 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
154 // CHECK2:       omp.inner.for.cond:
155 // CHECK2-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
156 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
157 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
158 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
159 // CHECK2:       omp.inner.for.body:
160 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
161 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
162 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
163 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
164 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
165 // CHECK2:       omp.body.continue:
166 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
167 // CHECK2:       omp.inner.for.inc:
168 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
169 // CHECK2-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
170 // CHECK2-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
171 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
172 // CHECK2:       omp.inner.for.end:
173 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
174 // CHECK2:       omp.loop.exit:
175 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
176 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
177 // CHECK2-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
178 // CHECK2-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
179 // CHECK2:       .omp.final.then:
180 // CHECK2-NEXT:    store i32 10, i32* [[I]], align 4
181 // CHECK2-NEXT:    br label [[DOTOMP_FINAL_DONE]]
182 // CHECK2:       .omp.final.done:
183 // CHECK2-NEXT:    ret void
184 //
185 //
186 // CHECK3-LABEL: define {{[^@]+}}@_Z3fooPf
187 // CHECK3-SAME: (float* [[C:%.*]]) #[[ATTR0:[0-9]+]] {
188 // CHECK3-NEXT:  entry:
189 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
190 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
191 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**)* @.omp_outlined. to void (i32*, i32*, ...)*), float** [[C_ADDR]])
192 // CHECK3-NEXT:    ret void
193 //
194 //
195 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
196 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], float** nonnull align 8 dereferenceable(8) [[C:%.*]]) #[[ATTR1:[0-9]+]] {
197 // CHECK3-NEXT:  entry:
198 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
199 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
200 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float**, align 8
201 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
202 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
203 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
204 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
205 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
206 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
207 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
208 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
209 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
210 // CHECK3-NEXT:    store float** [[C]], float*** [[C_ADDR]], align 8
211 // CHECK3-NEXT:    [[TMP0:%.*]] = load float**, float*** [[C_ADDR]], align 8
212 // CHECK3-NEXT:    [[TMP1:%.*]] = load float*, float** [[TMP0]], align 8
213 // CHECK3-NEXT:    call void @llvm.assume(i1 true) [ "align"(float* [[TMP1]], i64 16) ]
214 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
215 // CHECK3-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
216 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
217 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
218 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
219 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
220 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
221 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
222 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
223 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
224 // CHECK3:       cond.true:
225 // CHECK3-NEXT:    br label [[COND_END:%.*]]
226 // CHECK3:       cond.false:
227 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
228 // CHECK3-NEXT:    br label [[COND_END]]
229 // CHECK3:       cond.end:
230 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
231 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
232 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
233 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
234 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
235 // CHECK3:       omp.inner.for.cond:
236 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
237 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
238 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
239 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
240 // CHECK3:       omp.inner.for.body:
241 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
242 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
243 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
244 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
245 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
246 // CHECK3:       omp.body.continue:
247 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
248 // CHECK3:       omp.inner.for.inc:
249 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
250 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
251 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
252 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
253 // CHECK3:       omp.inner.for.end:
254 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
255 // CHECK3:       omp.loop.exit:
256 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
257 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
258 // CHECK3-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
259 // CHECK3-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
260 // CHECK3:       .omp.final.then:
261 // CHECK3-NEXT:    store i32 10, i32* [[I]], align 4
262 // CHECK3-NEXT:    br label [[DOTOMP_FINAL_DONE]]
263 // CHECK3:       .omp.final.done:
264 // CHECK3-NEXT:    ret void
265 //
266 //
267 // CHECK4-LABEL: define {{[^@]+}}@_Z3fooPf
268 // CHECK4-SAME: (float* [[C:%.*]]) #[[ATTR0:[0-9]+]] {
269 // CHECK4-NEXT:  entry:
270 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
271 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
272 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**)* @.omp_outlined. to void (i32*, i32*, ...)*), float** [[C_ADDR]])
273 // CHECK4-NEXT:    ret void
274 //
275 //
276 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
277 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], float** nonnull align 8 dereferenceable(8) [[C:%.*]]) #[[ATTR1:[0-9]+]] {
278 // CHECK4-NEXT:  entry:
279 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
280 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
281 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca float**, align 8
282 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
283 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
284 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
285 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
286 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
287 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
288 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
289 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
290 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
291 // CHECK4-NEXT:    store float** [[C]], float*** [[C_ADDR]], align 8
292 // CHECK4-NEXT:    [[TMP0:%.*]] = load float**, float*** [[C_ADDR]], align 8
293 // CHECK4-NEXT:    [[TMP1:%.*]] = load float*, float** [[TMP0]], align 8
294 // CHECK4-NEXT:    call void @llvm.assume(i1 true) [ "align"(float* [[TMP1]], i64 16) ]
295 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
296 // CHECK4-NEXT:    store i32 9, i32* [[DOTOMP_UB]], align 4
297 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
298 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
299 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
300 // CHECK4-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
301 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
302 // CHECK4-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
303 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
304 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
305 // CHECK4:       cond.true:
306 // CHECK4-NEXT:    br label [[COND_END:%.*]]
307 // CHECK4:       cond.false:
308 // CHECK4-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
309 // CHECK4-NEXT:    br label [[COND_END]]
310 // CHECK4:       cond.end:
311 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
312 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
313 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
314 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
315 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
316 // CHECK4:       omp.inner.for.cond:
317 // CHECK4-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
318 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
319 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
320 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
321 // CHECK4:       omp.inner.for.body:
322 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
323 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
324 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
325 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
326 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
327 // CHECK4:       omp.body.continue:
328 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
329 // CHECK4:       omp.inner.for.inc:
330 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
331 // CHECK4-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
332 // CHECK4-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
333 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
334 // CHECK4:       omp.inner.for.end:
335 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
336 // CHECK4:       omp.loop.exit:
337 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
338 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
339 // CHECK4-NEXT:    [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
340 // CHECK4-NEXT:    br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
341 // CHECK4:       .omp.final.then:
342 // CHECK4-NEXT:    store i32 10, i32* [[I]], align 4
343 // CHECK4-NEXT:    br label [[DOTOMP_FINAL_DONE]]
344 // CHECK4:       .omp.final.done:
345 // CHECK4-NEXT:    ret void
346 //
347