1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
2 // RUN: %clang_cc1 -no-opaque-pointers -DCHECK -verify -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -no-opaque-pointers -DCHECK -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
4 // RUN: %clang_cc1 -no-opaque-pointers -DCHECK -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK1
5 // RUN: %clang_cc1 -no-opaque-pointers -DCHECK -verify -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
6 // RUN: %clang_cc1 -no-opaque-pointers -DCHECK -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
7 // RUN: %clang_cc1 -no-opaque-pointers -DCHECK -fopenmp -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
8 
9 // RUN: %clang_cc1 -no-opaque-pointers -DCHECK -verify -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
10 // RUN: %clang_cc1 -no-opaque-pointers -DCHECK -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
11 // RUN: %clang_cc1 -no-opaque-pointers -DCHECK -fopenmp-simd -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
12 // RUN: %clang_cc1 -no-opaque-pointers -DCHECK -verify -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
13 // RUN: %clang_cc1 -no-opaque-pointers -DCHECK -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
14 // RUN: %clang_cc1 -no-opaque-pointers -DCHECK -fopenmp-simd -x c++ -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
15 
16 // RUN: %clang_cc1 -no-opaque-pointers -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK9
17 // RUN: %clang_cc1 -no-opaque-pointers -DLAMBDA -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
18 // RUN: %clang_cc1 -no-opaque-pointers -DLAMBDA -fopenmp -x c++  -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK9
19 
20 // RUN: %clang_cc1 -no-opaque-pointers -DLAMBDA -verify -fopenmp-simd -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}}"
21 // RUN: %clang_cc1 -no-opaque-pointers -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
22 // RUN: %clang_cc1 -no-opaque-pointers -DLAMBDA -fopenmp-simd -x c++  -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
23 
24 // expected-no-diagnostics
25 #ifndef HEADER
26 #define HEADER
27 
28 int x;
29 #pragma omp threadprivate(x)
30 
31 template <typename T>
32 T tmain() {
33   int a[2];
34 #pragma omp target
35 #pragma omp teams distribute parallel for copyin(x)
36   for (int i = 0; i < 2; ++i) {
37     a[i] = x;
38   }
39   return T();
40 }
41 
42 int main() {
43   int a[2];
44 #ifdef LAMBDA
45   [&]() {
46 #pragma omp target
47 #pragma omp teams distribute parallel for copyin(x)
48   for (int i = 0; i < 2; ++i) {
49 
50     // Skip global, bound tid and loop vars
51     a[i] = x;
52 
53 
54     // Skip global, bound tid and loop vars
55 
56 
57     [&]() {
58       a[i] = x;
59     }();
60   }
61   }();
62   return 0;
63 #else
64 #pragma omp target
65 #pragma omp teams distribute parallel for copyin(x)
66   for (int i = 0; i < 2; ++i) {
67     a[i] = x;
68   }
69   return tmain<int>();
70   //return 0;
71 #endif
72 }
73 
74 
75 
76 // Skip global, bound tid and loop vars
77 
78 
79 // Skip global, bound tid and loop vars
80 
81 
82 
83 
84 // Skip global, bound tid and loop vars
85 
86 
87 // Skip global, bound tid and loop vars
88 // prev lb and ub
89 
90 // iter variables
91 
92 
93 
94 #endif
95 // CHECK1-LABEL: define {{[^@]+}}@main
96 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
97 // CHECK1-NEXT:  entry:
98 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
99 // CHECK1-NEXT:    [[A:%.*]] = alloca [2 x i32], align 4
100 // CHECK1-NEXT:    [[X_CASTED:%.*]] = alloca i64, align 8
101 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 8
102 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 8
103 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 8
104 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
105 // CHECK1-NEXT:    store i32 0, i32* [[RETVAL]], align 4
106 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* @x, align 4
107 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32*
108 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
109 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[X_CASTED]], align 8
110 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
111 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i64*
112 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP3]], align 8
113 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
114 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
115 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP5]], align 8
116 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
117 // CHECK1-NEXT:    store i8* null, i8** [[TMP6]], align 8
118 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
119 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [2 x i32]**
120 // CHECK1-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[TMP8]], align 8
121 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
122 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [2 x i32]**
123 // CHECK1-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[TMP10]], align 8
124 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
125 // CHECK1-NEXT:    store i8* null, i8** [[TMP11]], align 8
126 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
127 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
128 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2)
129 // CHECK1-NEXT:    [[TMP14:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.region_id, i32 2, i8** [[TMP12]], i8** [[TMP13]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
130 // CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
131 // CHECK1-NEXT:    br i1 [[TMP15]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
132 // CHECK1:       omp_offload.failed:
133 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64(i64 [[TMP1]], [2 x i32]* [[A]]) #[[ATTR2:[0-9]+]]
134 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
135 // CHECK1:       omp_offload.cont:
136 // CHECK1-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiET_v()
137 // CHECK1-NEXT:    ret i32 [[CALL]]
138 //
139 //
140 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64
141 // CHECK1-SAME: (i64 noundef [[X:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
142 // CHECK1-NEXT:  entry:
143 // CHECK1-NEXT:    [[X_ADDR:%.*]] = alloca i64, align 8
144 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8
145 // CHECK1-NEXT:    store i64 [[X]], i64* [[X_ADDR]], align 8
146 // CHECK1-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8
147 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32*
148 // CHECK1-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8
149 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [2 x i32]*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), [2 x i32]* [[TMP0]], i32* [[CONV]])
150 // CHECK1-NEXT:    ret void
151 //
152 //
153 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
154 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
155 // CHECK1-NEXT:  entry:
156 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
157 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
158 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8
159 // CHECK1-NEXT:    [[X_ADDR:%.*]] = alloca i32*, align 8
160 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
161 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
162 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
163 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
164 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
165 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
166 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
167 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
168 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
169 // CHECK1-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8
170 // CHECK1-NEXT:    store i32* [[X]], i32** [[X_ADDR]], align 8
171 // CHECK1-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8
172 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[X_ADDR]], align 8
173 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
174 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
175 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
176 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
177 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
178 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
179 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
180 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
181 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
182 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
183 // CHECK1:       cond.true:
184 // CHECK1-NEXT:    br label [[COND_END:%.*]]
185 // CHECK1:       cond.false:
186 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
187 // CHECK1-NEXT:    br label [[COND_END]]
188 // CHECK1:       cond.end:
189 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
190 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
191 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
192 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
193 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
194 // CHECK1:       omp.inner.for.cond:
195 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
196 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
197 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
198 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
199 // CHECK1:       omp.inner.for.body:
200 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
201 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
202 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
203 // CHECK1-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP11]] to i64
204 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [2 x i32]*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP10]], i64 [[TMP12]], [2 x i32]* [[TMP0]], i32* [[TMP1]])
205 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
206 // CHECK1:       omp.inner.for.inc:
207 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
208 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
209 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
210 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
211 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
212 // CHECK1:       omp.inner.for.end:
213 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
214 // CHECK1:       omp.loop.exit:
215 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
216 // CHECK1-NEXT:    ret void
217 //
218 //
219 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
220 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
221 // CHECK1-NEXT:  entry:
222 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
223 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
224 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
225 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
226 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8
227 // CHECK1-NEXT:    [[X_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_LB:%.*]] = alloca i32, align 4
231 // CHECK1-NEXT:    [[DOTOMP_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 i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
238 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
239 // CHECK1-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8
240 // CHECK1-NEXT:    store i32* [[X]], i32** [[X_ADDR]], align 8
241 // CHECK1-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8
242 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[X_ADDR]], align 8
243 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
244 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
245 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
246 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP2]] to i32
247 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
248 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP3]] to i32
249 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
250 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
251 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
252 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
253 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
254 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
255 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
256 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
257 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 1
258 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
259 // CHECK1:       cond.true:
260 // CHECK1-NEXT:    br label [[COND_END:%.*]]
261 // CHECK1:       cond.false:
262 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
263 // CHECK1-NEXT:    br label [[COND_END]]
264 // CHECK1:       cond.end:
265 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
266 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
267 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
268 // CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
269 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
270 // CHECK1:       omp.inner.for.cond:
271 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
272 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
273 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
274 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
275 // CHECK1:       omp.inner.for.body:
276 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
277 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
278 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
279 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
280 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP1]], align 4
281 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4
282 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
283 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
284 // CHECK1-NEXT:    store i32 [[TMP12]], i32* [[ARRAYIDX]], align 4
285 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
286 // CHECK1:       omp.body.continue:
287 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
288 // CHECK1:       omp.inner.for.inc:
289 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
290 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
291 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
292 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
293 // CHECK1:       omp.inner.for.end:
294 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
295 // CHECK1:       omp.loop.exit:
296 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
297 // CHECK1-NEXT:    ret void
298 //
299 //
300 // CHECK1-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
301 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] comdat {
302 // CHECK1-NEXT:  entry:
303 // CHECK1-NEXT:    [[A:%.*]] = alloca [2 x i32], align 4
304 // CHECK1-NEXT:    [[X_CASTED:%.*]] = alloca i64, align 8
305 // CHECK1-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 8
306 // CHECK1-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 8
307 // CHECK1-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 8
308 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
309 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* @x, align 4
310 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[X_CASTED]] to i32*
311 // CHECK1-NEXT:    store i32 [[TMP0]], i32* [[CONV]], align 4
312 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, i64* [[X_CASTED]], align 8
313 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
314 // CHECK1-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i64*
315 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP3]], align 8
316 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
317 // CHECK1-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i64*
318 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP5]], align 8
319 // CHECK1-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
320 // CHECK1-NEXT:    store i8* null, i8** [[TMP6]], align 8
321 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
322 // CHECK1-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [2 x i32]**
323 // CHECK1-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[TMP8]], align 8
324 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
325 // CHECK1-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [2 x i32]**
326 // CHECK1-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[TMP10]], align 8
327 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
328 // CHECK1-NEXT:    store i8* null, i8** [[TMP11]], align 8
329 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
330 // CHECK1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
331 // CHECK1-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2)
332 // CHECK1-NEXT:    [[TMP14:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34.region_id, i32 2, i8** [[TMP12]], i8** [[TMP13]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
333 // CHECK1-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
334 // CHECK1-NEXT:    br i1 [[TMP15]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
335 // CHECK1:       omp_offload.failed:
336 // CHECK1-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34(i64 [[TMP1]], [2 x i32]* [[A]]) #[[ATTR2]]
337 // CHECK1-NEXT:    br label [[OMP_OFFLOAD_CONT]]
338 // CHECK1:       omp_offload.cont:
339 // CHECK1-NEXT:    ret i32 0
340 //
341 //
342 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34
343 // CHECK1-SAME: (i64 noundef [[X:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
344 // CHECK1-NEXT:  entry:
345 // CHECK1-NEXT:    [[X_ADDR:%.*]] = alloca i64, align 8
346 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8
347 // CHECK1-NEXT:    store i64 [[X]], i64* [[X_ADDR]], align 8
348 // CHECK1-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8
349 // CHECK1-NEXT:    [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32*
350 // CHECK1-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8
351 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [2 x i32]*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), [2 x i32]* [[TMP0]], i32* [[CONV]])
352 // CHECK1-NEXT:    ret void
353 //
354 //
355 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
356 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
357 // CHECK1-NEXT:  entry:
358 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
359 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
360 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8
361 // CHECK1-NEXT:    [[X_ADDR:%.*]] = alloca i32*, align 8
362 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
363 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
364 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
365 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
366 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
367 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
368 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
369 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
370 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
371 // CHECK1-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8
372 // CHECK1-NEXT:    store i32* [[X]], i32** [[X_ADDR]], align 8
373 // CHECK1-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8
374 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[X_ADDR]], align 8
375 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
376 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
377 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
378 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
379 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
380 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
381 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
382 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
383 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
384 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
385 // CHECK1:       cond.true:
386 // CHECK1-NEXT:    br label [[COND_END:%.*]]
387 // CHECK1:       cond.false:
388 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
389 // CHECK1-NEXT:    br label [[COND_END]]
390 // CHECK1:       cond.end:
391 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
392 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
393 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
394 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
395 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
396 // CHECK1:       omp.inner.for.cond:
397 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
398 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
399 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
400 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
401 // CHECK1:       omp.inner.for.body:
402 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
403 // CHECK1-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
404 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
405 // CHECK1-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP11]] to i64
406 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [2 x i32]*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i64 [[TMP10]], i64 [[TMP12]], [2 x i32]* [[TMP0]], i32* [[TMP1]])
407 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
408 // CHECK1:       omp.inner.for.inc:
409 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
410 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
411 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
412 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
413 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
414 // CHECK1:       omp.inner.for.end:
415 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
416 // CHECK1:       omp.loop.exit:
417 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
418 // CHECK1-NEXT:    ret void
419 //
420 //
421 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
422 // CHECK1-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
423 // CHECK1-NEXT:  entry:
424 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
425 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
426 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
427 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
428 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8
429 // CHECK1-NEXT:    [[X_ADDR:%.*]] = alloca i32*, align 8
430 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
431 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
432 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
433 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
434 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
435 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
436 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
437 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
438 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
439 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
440 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
441 // CHECK1-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8
442 // CHECK1-NEXT:    store i32* [[X]], i32** [[X_ADDR]], align 8
443 // CHECK1-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8
444 // CHECK1-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[X_ADDR]], align 8
445 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
446 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
447 // CHECK1-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
448 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP2]] to i32
449 // CHECK1-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
450 // CHECK1-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP3]] to i32
451 // CHECK1-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
452 // CHECK1-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
453 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
454 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
455 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
456 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
457 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
458 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
459 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 1
460 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
461 // CHECK1:       cond.true:
462 // CHECK1-NEXT:    br label [[COND_END:%.*]]
463 // CHECK1:       cond.false:
464 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
465 // CHECK1-NEXT:    br label [[COND_END]]
466 // CHECK1:       cond.end:
467 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
468 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
469 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
470 // CHECK1-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
471 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
472 // CHECK1:       omp.inner.for.cond:
473 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
474 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
475 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
476 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
477 // CHECK1:       omp.inner.for.body:
478 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
479 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
480 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
481 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
482 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP1]], align 4
483 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4
484 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
485 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
486 // CHECK1-NEXT:    store i32 [[TMP12]], i32* [[ARRAYIDX]], align 4
487 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
488 // CHECK1:       omp.body.continue:
489 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
490 // CHECK1:       omp.inner.for.inc:
491 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
492 // CHECK1-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP14]], 1
493 // CHECK1-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
494 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
495 // CHECK1:       omp.inner.for.end:
496 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
497 // CHECK1:       omp.loop.exit:
498 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
499 // CHECK1-NEXT:    ret void
500 //
501 //
502 // CHECK1-LABEL: define {{[^@]+}}@_ZTW1x
503 // CHECK1-SAME: () #[[ATTR4:[0-9]+]] comdat {
504 // CHECK1-NEXT:    ret i32* @x
505 //
506 //
507 // CHECK1-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
508 // CHECK1-SAME: () #[[ATTR5:[0-9]+]] {
509 // CHECK1-NEXT:  entry:
510 // CHECK1-NEXT:    call void @__tgt_register_requires(i64 1)
511 // CHECK1-NEXT:    ret void
512 //
513 //
514 // CHECK3-LABEL: define {{[^@]+}}@main
515 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
516 // CHECK3-NEXT:  entry:
517 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
518 // CHECK3-NEXT:    [[A:%.*]] = alloca [2 x i32], align 4
519 // CHECK3-NEXT:    [[X_CASTED:%.*]] = alloca i32, align 4
520 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 4
521 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 4
522 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 4
523 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
524 // CHECK3-NEXT:    store i32 0, i32* [[RETVAL]], align 4
525 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* @x, align 4
526 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[X_CASTED]], align 4
527 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[X_CASTED]], align 4
528 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
529 // CHECK3-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32*
530 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP3]], align 4
531 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
532 // CHECK3-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32*
533 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP5]], align 4
534 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
535 // CHECK3-NEXT:    store i8* null, i8** [[TMP6]], align 4
536 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
537 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [2 x i32]**
538 // CHECK3-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[TMP8]], align 4
539 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
540 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [2 x i32]**
541 // CHECK3-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[TMP10]], align 4
542 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
543 // CHECK3-NEXT:    store i8* null, i8** [[TMP11]], align 4
544 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
545 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
546 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3:[0-9]+]], i64 -1, i64 2)
547 // CHECK3-NEXT:    [[TMP14:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64.region_id, i32 2, i8** [[TMP12]], i8** [[TMP13]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
548 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
549 // CHECK3-NEXT:    br i1 [[TMP15]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
550 // CHECK3:       omp_offload.failed:
551 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64(i32 [[TMP1]], [2 x i32]* [[A]]) #[[ATTR2:[0-9]+]]
552 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
553 // CHECK3:       omp_offload.cont:
554 // CHECK3-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiET_v()
555 // CHECK3-NEXT:    ret i32 [[CALL]]
556 //
557 //
558 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l64
559 // CHECK3-SAME: (i32 noundef [[X:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
560 // CHECK3-NEXT:  entry:
561 // CHECK3-NEXT:    [[X_ADDR:%.*]] = alloca i32, align 4
562 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 4
563 // CHECK3-NEXT:    store i32 [[X]], i32* [[X_ADDR]], align 4
564 // CHECK3-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 4
565 // CHECK3-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 4
566 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [2 x i32]*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), [2 x i32]* [[TMP0]], i32* [[X_ADDR]])
567 // CHECK3-NEXT:    ret void
568 //
569 //
570 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
571 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
572 // CHECK3-NEXT:  entry:
573 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
574 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
575 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 4
576 // CHECK3-NEXT:    [[X_ADDR:%.*]] = alloca i32*, align 4
577 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
578 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
579 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
580 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
581 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
582 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
583 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
584 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
585 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
586 // CHECK3-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 4
587 // CHECK3-NEXT:    store i32* [[X]], i32** [[X_ADDR]], align 4
588 // CHECK3-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 4
589 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[X_ADDR]], align 4
590 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
591 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
592 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
593 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
594 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
595 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
596 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
597 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
598 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
599 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
600 // CHECK3:       cond.true:
601 // CHECK3-NEXT:    br label [[COND_END:%.*]]
602 // CHECK3:       cond.false:
603 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
604 // CHECK3-NEXT:    br label [[COND_END]]
605 // CHECK3:       cond.end:
606 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
607 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
608 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
609 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
610 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
611 // CHECK3:       omp.inner.for.cond:
612 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
613 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
614 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
615 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
616 // CHECK3:       omp.inner.for.body:
617 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
618 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
619 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [2 x i32]*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i32 [[TMP9]], i32 [[TMP10]], [2 x i32]* [[TMP0]], i32* [[TMP1]])
620 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
621 // CHECK3:       omp.inner.for.inc:
622 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
623 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
624 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
625 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
626 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
627 // CHECK3:       omp.inner.for.end:
628 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
629 // CHECK3:       omp.loop.exit:
630 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
631 // CHECK3-NEXT:    ret void
632 //
633 //
634 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
635 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
636 // CHECK3-NEXT:  entry:
637 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
638 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
639 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
640 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
641 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 4
642 // CHECK3-NEXT:    [[X_ADDR:%.*]] = alloca i32*, align 4
643 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
644 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
645 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
646 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
647 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
648 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
649 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
650 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
651 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
652 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
653 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
654 // CHECK3-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 4
655 // CHECK3-NEXT:    store i32* [[X]], i32** [[X_ADDR]], align 4
656 // CHECK3-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 4
657 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[X_ADDR]], align 4
658 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
659 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
660 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
661 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
662 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_LB]], align 4
663 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_UB]], align 4
664 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
665 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
666 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
667 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
668 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
669 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
670 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 1
671 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
672 // CHECK3:       cond.true:
673 // CHECK3-NEXT:    br label [[COND_END:%.*]]
674 // CHECK3:       cond.false:
675 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
676 // CHECK3-NEXT:    br label [[COND_END]]
677 // CHECK3:       cond.end:
678 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
679 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
680 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
681 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
682 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
683 // CHECK3:       omp.inner.for.cond:
684 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
685 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
686 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
687 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
688 // CHECK3:       omp.inner.for.body:
689 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
690 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
691 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
692 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
693 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP1]], align 4
694 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4
695 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
696 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[ARRAYIDX]], align 4
697 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
698 // CHECK3:       omp.body.continue:
699 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
700 // CHECK3:       omp.inner.for.inc:
701 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
702 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], 1
703 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
704 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
705 // CHECK3:       omp.inner.for.end:
706 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
707 // CHECK3:       omp.loop.exit:
708 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
709 // CHECK3-NEXT:    ret void
710 //
711 //
712 // CHECK3-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
713 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] comdat {
714 // CHECK3-NEXT:  entry:
715 // CHECK3-NEXT:    [[A:%.*]] = alloca [2 x i32], align 4
716 // CHECK3-NEXT:    [[X_CASTED:%.*]] = alloca i32, align 4
717 // CHECK3-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [2 x i8*], align 4
718 // CHECK3-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [2 x i8*], align 4
719 // CHECK3-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [2 x i8*], align 4
720 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
721 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* @x, align 4
722 // CHECK3-NEXT:    store i32 [[TMP0]], i32* [[X_CASTED]], align 4
723 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, i32* [[X_CASTED]], align 4
724 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
725 // CHECK3-NEXT:    [[TMP3:%.*]] = bitcast i8** [[TMP2]] to i32*
726 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP3]], align 4
727 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
728 // CHECK3-NEXT:    [[TMP5:%.*]] = bitcast i8** [[TMP4]] to i32*
729 // CHECK3-NEXT:    store i32 [[TMP1]], i32* [[TMP5]], align 4
730 // CHECK3-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
731 // CHECK3-NEXT:    store i8* null, i8** [[TMP6]], align 4
732 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
733 // CHECK3-NEXT:    [[TMP8:%.*]] = bitcast i8** [[TMP7]] to [2 x i32]**
734 // CHECK3-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[TMP8]], align 4
735 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 1
736 // CHECK3-NEXT:    [[TMP10:%.*]] = bitcast i8** [[TMP9]] to [2 x i32]**
737 // CHECK3-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[TMP10]], align 4
738 // CHECK3-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
739 // CHECK3-NEXT:    store i8* null, i8** [[TMP11]], align 4
740 // CHECK3-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
741 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [2 x i8*], [2 x i8*]* [[DOTOFFLOAD_PTRS]], i32 0, i32 0
742 // CHECK3-NEXT:    call void @__kmpc_push_target_tripcount_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i64 2)
743 // CHECK3-NEXT:    [[TMP14:%.*]] = call i32 @__tgt_target_teams_mapper(%struct.ident_t* @[[GLOB3]], i64 -1, i8* @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34.region_id, i32 2, i8** [[TMP12]], i8** [[TMP13]], i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_sizes.4, i32 0, i32 0), i64* getelementptr inbounds ([2 x i64], [2 x i64]* @.offload_maptypes.5, i32 0, i32 0), i8** null, i8** null, i32 0, i32 0)
744 // CHECK3-NEXT:    [[TMP15:%.*]] = icmp ne i32 [[TMP14]], 0
745 // CHECK3-NEXT:    br i1 [[TMP15]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
746 // CHECK3:       omp_offload.failed:
747 // CHECK3-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34(i32 [[TMP1]], [2 x i32]* [[A]]) #[[ATTR2]]
748 // CHECK3-NEXT:    br label [[OMP_OFFLOAD_CONT]]
749 // CHECK3:       omp_offload.cont:
750 // CHECK3-NEXT:    ret i32 0
751 //
752 //
753 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l34
754 // CHECK3-SAME: (i32 noundef [[X:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
755 // CHECK3-NEXT:  entry:
756 // CHECK3-NEXT:    [[X_ADDR:%.*]] = alloca i32, align 4
757 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 4
758 // CHECK3-NEXT:    store i32 [[X]], i32* [[X_ADDR]], align 4
759 // CHECK3-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 4
760 // CHECK3-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 4
761 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [2 x i32]*, i32*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), [2 x i32]* [[TMP0]], i32* [[X_ADDR]])
762 // CHECK3-NEXT:    ret void
763 //
764 //
765 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
766 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
767 // CHECK3-NEXT:  entry:
768 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
769 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
770 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 4
771 // CHECK3-NEXT:    [[X_ADDR:%.*]] = alloca i32*, align 4
772 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
773 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
774 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
775 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
776 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
777 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
778 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
779 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
780 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
781 // CHECK3-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 4
782 // CHECK3-NEXT:    store i32* [[X]], i32** [[X_ADDR]], align 4
783 // CHECK3-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 4
784 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[X_ADDR]], align 4
785 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
786 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
787 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
788 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
789 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
790 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
791 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
792 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
793 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
794 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
795 // CHECK3:       cond.true:
796 // CHECK3-NEXT:    br label [[COND_END:%.*]]
797 // CHECK3:       cond.false:
798 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
799 // CHECK3-NEXT:    br label [[COND_END]]
800 // CHECK3:       cond.end:
801 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
802 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
803 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
804 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
805 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
806 // CHECK3:       omp.inner.for.cond:
807 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
808 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
809 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
810 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
811 // CHECK3:       omp.inner.for.body:
812 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
813 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
814 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i32, i32, [2 x i32]*, i32*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), i32 [[TMP9]], i32 [[TMP10]], [2 x i32]* [[TMP0]], i32* [[TMP1]])
815 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
816 // CHECK3:       omp.inner.for.inc:
817 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
818 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
819 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP11]], [[TMP12]]
820 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
821 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
822 // CHECK3:       omp.inner.for.end:
823 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
824 // CHECK3:       omp.loop.exit:
825 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
826 // CHECK3-NEXT:    ret void
827 //
828 //
829 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
830 // CHECK3-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR1]] {
831 // CHECK3-NEXT:  entry:
832 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 4
833 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 4
834 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
835 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
836 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 4
837 // CHECK3-NEXT:    [[X_ADDR:%.*]] = alloca i32*, align 4
838 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
839 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
840 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
841 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
842 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
843 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
844 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
845 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 4
846 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 4
847 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], i32* [[DOTPREVIOUS_LB__ADDR]], align 4
848 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], i32* [[DOTPREVIOUS_UB__ADDR]], align 4
849 // CHECK3-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 4
850 // CHECK3-NEXT:    store i32* [[X]], i32** [[X_ADDR]], align 4
851 // CHECK3-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 4
852 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[X_ADDR]], align 4
853 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
854 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
855 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTPREVIOUS_LB__ADDR]], align 4
856 // CHECK3-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTPREVIOUS_UB__ADDR]], align 4
857 // CHECK3-NEXT:    store i32 [[TMP2]], i32* [[DOTOMP_LB]], align 4
858 // CHECK3-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_UB]], align 4
859 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
860 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
861 // CHECK3-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 4
862 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
863 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
864 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
865 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 1
866 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
867 // CHECK3:       cond.true:
868 // CHECK3-NEXT:    br label [[COND_END:%.*]]
869 // CHECK3:       cond.false:
870 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
871 // CHECK3-NEXT:    br label [[COND_END]]
872 // CHECK3:       cond.end:
873 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
874 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
875 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
876 // CHECK3-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
877 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
878 // CHECK3:       omp.inner.for.cond:
879 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
880 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
881 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
882 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
883 // CHECK3:       omp.inner.for.body:
884 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
885 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
886 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
887 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
888 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP1]], align 4
889 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4
890 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP0]], i32 0, i32 [[TMP13]]
891 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[ARRAYIDX]], align 4
892 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
893 // CHECK3:       omp.body.continue:
894 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
895 // CHECK3:       omp.inner.for.inc:
896 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
897 // CHECK3-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP14]], 1
898 // CHECK3-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
899 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
900 // CHECK3:       omp.inner.for.end:
901 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
902 // CHECK3:       omp.loop.exit:
903 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
904 // CHECK3-NEXT:    ret void
905 //
906 //
907 // CHECK3-LABEL: define {{[^@]+}}@_ZTW1x
908 // CHECK3-SAME: () #[[ATTR4:[0-9]+]] comdat {
909 // CHECK3-NEXT:    ret i32* @x
910 //
911 //
912 // CHECK3-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
913 // CHECK3-SAME: () #[[ATTR5:[0-9]+]] {
914 // CHECK3-NEXT:  entry:
915 // CHECK3-NEXT:    call void @__tgt_register_requires(i64 1)
916 // CHECK3-NEXT:    ret void
917 //
918 //
919 // CHECK9-LABEL: define {{[^@]+}}@main
920 // CHECK9-SAME: () #[[ATTR0:[0-9]+]] {
921 // CHECK9-NEXT:  entry:
922 // CHECK9-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
923 // CHECK9-NEXT:    [[A:%.*]] = alloca [2 x i32], align 4
924 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 8
925 // CHECK9-NEXT:    store i32 0, i32* [[RETVAL]], align 4
926 // CHECK9-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[CLASS_ANON]], %class.anon* [[REF_TMP]], i32 0, i32 0
927 // CHECK9-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[TMP0]], align 8
928 // CHECK9-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(%class.anon* noundef nonnull align 8 dereferenceable(8) [[REF_TMP]])
929 // CHECK9-NEXT:    ret i32 0
930 //
931 //
932 // CHECK9-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46
933 // CHECK9-SAME: (i64 noundef [[X:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]]) #[[ATTR2:[0-9]+]] {
934 // CHECK9-NEXT:  entry:
935 // CHECK9-NEXT:    [[X_ADDR:%.*]] = alloca i64, align 8
936 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8
937 // CHECK9-NEXT:    store i64 [[X]], i64* [[X_ADDR]], align 8
938 // CHECK9-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8
939 // CHECK9-NEXT:    [[CONV:%.*]] = bitcast i64* [[X_ADDR]] to i32*
940 // CHECK9-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8
941 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_teams(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 2, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, [2 x i32]*, i32*)* @.omp_outlined. to void (i32*, i32*, ...)*), [2 x i32]* [[TMP0]], i32* [[CONV]])
942 // CHECK9-NEXT:    ret void
943 //
944 //
945 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined.
946 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR2]] {
947 // CHECK9-NEXT:  entry:
948 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
949 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
950 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8
951 // CHECK9-NEXT:    [[X_ADDR:%.*]] = alloca i32*, align 8
952 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
953 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
954 // CHECK9-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
955 // CHECK9-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
956 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
957 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
958 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
959 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
960 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
961 // CHECK9-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8
962 // CHECK9-NEXT:    store i32* [[X]], i32** [[X_ADDR]], align 8
963 // CHECK9-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8
964 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[X_ADDR]], align 8
965 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_COMB_LB]], align 4
966 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_COMB_UB]], align 4
967 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
968 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
969 // CHECK9-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
970 // CHECK9-NEXT:    [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
971 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 92, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_COMB_LB]], i32* [[DOTOMP_COMB_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
972 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
973 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
974 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
975 // CHECK9:       cond.true:
976 // CHECK9-NEXT:    br label [[COND_END:%.*]]
977 // CHECK9:       cond.false:
978 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
979 // CHECK9-NEXT:    br label [[COND_END]]
980 // CHECK9:       cond.end:
981 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
982 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_COMB_UB]], align 4
983 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
984 // CHECK9-NEXT:    store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
985 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
986 // CHECK9:       omp.inner.for.cond:
987 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
988 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
989 // CHECK9-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
990 // CHECK9-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
991 // CHECK9:       omp.inner.for.body:
992 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_COMB_LB]], align 4
993 // CHECK9-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP9]] to i64
994 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_COMB_UB]], align 4
995 // CHECK9-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP11]] to i64
996 // CHECK9-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB3]], i32 4, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, i64, i64, [2 x i32]*, i32*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), i64 [[TMP10]], i64 [[TMP12]], [2 x i32]* [[TMP0]], i32* [[TMP1]])
997 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
998 // CHECK9:       omp.inner.for.inc:
999 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1000 // CHECK9-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1001 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP13]], [[TMP14]]
1002 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1003 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1004 // CHECK9:       omp.inner.for.end:
1005 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1006 // CHECK9:       omp.loop.exit:
1007 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
1008 // CHECK9-NEXT:    ret void
1009 //
1010 //
1011 // CHECK9-LABEL: define {{[^@]+}}@.omp_outlined..1
1012 // CHECK9-SAME: (i32* noalias noundef [[DOTGLOBAL_TID_:%.*]], i32* noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], [2 x i32]* noundef nonnull align 4 dereferenceable(8) [[A:%.*]], i32* noundef nonnull align 4 dereferenceable(4) [[X:%.*]]) #[[ATTR2]] {
1013 // CHECK9-NEXT:  entry:
1014 // CHECK9-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1015 // CHECK9-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1016 // CHECK9-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1017 // CHECK9-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1018 // CHECK9-NEXT:    [[A_ADDR:%.*]] = alloca [2 x i32]*, align 8
1019 // CHECK9-NEXT:    [[X_ADDR:%.*]] = alloca i32*, align 8
1020 // CHECK9-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1021 // CHECK9-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1022 // CHECK9-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1023 // CHECK9-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1024 // CHECK9-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1025 // CHECK9-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1026 // CHECK9-NEXT:    [[I:%.*]] = alloca i32, align 4
1027 // CHECK9-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
1028 // CHECK9-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1029 // CHECK9-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1030 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_LB_]], i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1031 // CHECK9-NEXT:    store i64 [[DOTPREVIOUS_UB_]], i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1032 // CHECK9-NEXT:    store [2 x i32]* [[A]], [2 x i32]** [[A_ADDR]], align 8
1033 // CHECK9-NEXT:    store i32* [[X]], i32** [[X_ADDR]], align 8
1034 // CHECK9-NEXT:    [[TMP0:%.*]] = load [2 x i32]*, [2 x i32]** [[A_ADDR]], align 8
1035 // CHECK9-NEXT:    [[TMP1:%.*]] = load i32*, i32** [[X_ADDR]], align 8
1036 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1037 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_UB]], align 4
1038 // CHECK9-NEXT:    [[TMP2:%.*]] = load i64, i64* [[DOTPREVIOUS_LB__ADDR]], align 8
1039 // CHECK9-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP2]] to i32
1040 // CHECK9-NEXT:    [[TMP3:%.*]] = load i64, i64* [[DOTPREVIOUS_UB__ADDR]], align 8
1041 // CHECK9-NEXT:    [[CONV1:%.*]] = trunc i64 [[TMP3]] to i32
1042 // CHECK9-NEXT:    store i32 [[CONV]], i32* [[DOTOMP_LB]], align 4
1043 // CHECK9-NEXT:    store i32 [[CONV1]], i32* [[DOTOMP_UB]], align 4
1044 // CHECK9-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1045 // CHECK9-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1046 // CHECK9-NEXT:    [[TMP4:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1047 // CHECK9-NEXT:    [[TMP5:%.*]] = load i32, i32* [[TMP4]], align 4
1048 // CHECK9-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 [[TMP5]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1049 // CHECK9-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1050 // CHECK9-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 1
1051 // CHECK9-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1052 // CHECK9:       cond.true:
1053 // CHECK9-NEXT:    br label [[COND_END:%.*]]
1054 // CHECK9:       cond.false:
1055 // CHECK9-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1056 // CHECK9-NEXT:    br label [[COND_END]]
1057 // CHECK9:       cond.end:
1058 // CHECK9-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
1059 // CHECK9-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1060 // CHECK9-NEXT:    [[TMP8:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1061 // CHECK9-NEXT:    store i32 [[TMP8]], i32* [[DOTOMP_IV]], align 4
1062 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1063 // CHECK9:       omp.inner.for.cond:
1064 // CHECK9-NEXT:    [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1065 // CHECK9-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1066 // CHECK9-NEXT:    [[CMP2:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
1067 // CHECK9-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1068 // CHECK9:       omp.inner.for.body:
1069 // CHECK9-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1070 // CHECK9-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
1071 // CHECK9-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1072 // CHECK9-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
1073 // CHECK9-NEXT:    [[TMP12:%.*]] = load i32, i32* [[TMP1]], align 4
1074 // CHECK9-NEXT:    [[TMP13:%.*]] = load i32, i32* [[I]], align 4
1075 // CHECK9-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP13]] to i64
1076 // CHECK9-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], [2 x i32]* [[TMP0]], i64 0, i64 [[IDXPROM]]
1077 // CHECK9-NEXT:    store i32 [[TMP12]], i32* [[ARRAYIDX]], align 4
1078 // CHECK9-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 0
1079 // CHECK9-NEXT:    store [2 x i32]* [[TMP0]], [2 x i32]** [[TMP14]], align 8
1080 // CHECK9-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 1
1081 // CHECK9-NEXT:    store i32* [[I]], i32** [[TMP15]], align 8
1082 // CHECK9-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [[CLASS_ANON_0]], %class.anon.0* [[REF_TMP]], i32 0, i32 2
1083 // CHECK9-NEXT:    store i32* [[TMP1]], i32** [[TMP16]], align 8
1084 // CHECK9-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(%class.anon.0* noundef nonnull align 8 dereferenceable(24) [[REF_TMP]])
1085 // CHECK9-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1086 // CHECK9:       omp.body.continue:
1087 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1088 // CHECK9:       omp.inner.for.inc:
1089 // CHECK9-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1090 // CHECK9-NEXT:    [[ADD3:%.*]] = add nsw i32 [[TMP17]], 1
1091 // CHECK9-NEXT:    store i32 [[ADD3]], i32* [[DOTOMP_IV]], align 4
1092 // CHECK9-NEXT:    br label [[OMP_INNER_FOR_COND]]
1093 // CHECK9:       omp.inner.for.end:
1094 // CHECK9-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1095 // CHECK9:       omp.loop.exit:
1096 // CHECK9-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP5]])
1097 // CHECK9-NEXT:    ret void
1098 //
1099 //
1100 // CHECK9-LABEL: define {{[^@]+}}@_ZTW1x
1101 // CHECK9-SAME: () #[[ATTR4:[0-9]+]] comdat {
1102 // CHECK9-NEXT:    ret i32* @x
1103 //
1104 //
1105 // CHECK9-LABEL: define {{[^@]+}}@.omp_offloading.requires_reg
1106 // CHECK9-SAME: () #[[ATTR5:[0-9]+]] {
1107 // CHECK9-NEXT:  entry:
1108 // CHECK9-NEXT:    call void @__tgt_register_requires(i64 1)
1109 // CHECK9-NEXT:    ret void
1110 //
1111