xref: /llvm-project/clang/test/OpenMP/distribute_parallel_for_firstprivate_codegen.cpp (revision 94473f4db6a6f5f12d7c4081455b5b596094eac5)
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 -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -DLAMBDA -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 -DLAMBDA -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK1
5 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK3
6 // RUN: %clang_cc1 -DLAMBDA -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 -DLAMBDA -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK3
8 
9 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
10 // RUN: %clang_cc1 -DLAMBDA -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 - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
11 // RUN: %clang_cc1 -DLAMBDA -verify -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
12 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
13 // RUN: %clang_cc1 -DLAMBDA -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
14 
15 // RUN: %clang_cc1  -verify -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK8
16 // RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
17 // RUN: %clang_cc1  -fopenmp -x c++ -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK8
18 // RUN: %clang_cc1  -verify -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK10
19 // RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
20 // RUN: %clang_cc1  -fopenmp -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --check-prefix=CHECK10
21 
22 // RUN: %clang_cc1  -verify -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
23 // RUN: %clang_cc1  -fopenmp-simd -x c++ -std=c++11 -triple powerpc64le-unknown-unknown -fopenmp-targets=powerpc64le-ibm-linux-gnu -emit-pch -o %t %s
24 // RUN: %clang_cc1  -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 - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
25 // RUN: %clang_cc1  -verify -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-llvm %s -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
26 // RUN: %clang_cc1  -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -emit-pch -o %t %s
27 // RUN: %clang_cc1  -fopenmp-simd -x c++ -std=c++11 -triple i386-unknown-unknown -fopenmp-targets=i386-pc-linux-gnu -std=c++11 -include-pch %t -verify %s -emit-llvm -o - -Wno-openmp-mapping | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
28 // expected-no-diagnostics
29 #ifndef HEADER
30 #define HEADER
31 
32 template <class T>
33 struct S {
34   T f;
35   S(T a) : f(a) {}
36   S() : f() {}
37   operator T() { return T(); }
38   ~S() {}
39 };
40 
41 template <typename T>
42 T tmain() {
43   S<T> test;
44   T t_var = T();
45   T vec[] = {1, 2};
46   S<T> s_arr[] = {1, 2};
47   S<T> &var = test;
48   #pragma omp target
49   #pragma omp teams
50   #pragma omp distribute parallel for firstprivate(t_var, vec, s_arr, s_arr, var, var)
51   for (int i = 0; i < 2; ++i) {
52     vec[i] = t_var;
53     s_arr[i] = var;
54   }
55   return T();
56 }
57 
58 int main() {
59   static int svar;
60   volatile double g;
61   volatile double &g1 = g;
62 
63   #ifdef LAMBDA
64   [&]() {
65     static float sfvar;
66 
67     #pragma omp target
68     #pragma omp teams
69     #pragma omp distribute parallel for firstprivate(g, g1, svar, sfvar)
70     for (int i = 0; i < 2; ++i) {
71 
72       // addr alloca's
73 
74       // private alloca's
75 
76       // transfer input parameters into addr alloca's
77 
78 
79       // init private alloca's with addr alloca's
80       // g
81 
82       // g1
83 
84       // svar
85 
86       // sfvar
87 
88       // pass firstprivate parameters to parallel outlined function
89       // g
90 
91       // g1
92 
93       // svar
94 
95       // sfvar
96 
97 
98 
99       // skip initial params
100 
101       // addr alloca's
102 
103       // private alloca's (only for 32-bit)
104 
105       // transfer input parameters into addr alloca's
106 
107       // prepare parameters for lambda
108       // g
109 
110       // g1
111 
112       // svar
113 
114       // sfvar
115 
116       g = 1;
117       g1 = 1;
118       svar = 3;
119       sfvar = 4.0;
120 
121       // pass params to inner lambda
122       [&]() {
123 	g = 2;
124 	g1 = 2;
125 	svar = 4;
126 	sfvar = 8.0;
127 
128       }();
129     }
130   }();
131   return 0;
132   #else
133   S<float> test;
134   int t_var = 0;
135   int vec[] = {1, 2};
136   S<float> s_arr[] = {1, 2};
137   S<float> &var = test;
138 
139   #pragma omp target
140   #pragma omp teams
141   #pragma omp distribute parallel for firstprivate(t_var, vec, s_arr, s_arr, var, var, svar)
142   for (int i = 0; i < 2; ++i) {
143     vec[i] = t_var;
144     s_arr[i] = var;
145   }
146   return tmain<int>();
147   #endif
148 }
149 
150 
151 
152 
153 // addr alloca's
154 
155 // skip loop alloca's
156 
157 // private alloca's
158 
159 
160 // init addr alloca's with input values
161 
162 // init private alloca's with addr alloca's
163 // t-var
164 
165 // vec
166 
167 // s_arr
168 
169 // var
170 
171 // svar
172 
173 // pass private alloca's to fork
174 // not dag to distinguish with S_VAR_CAST
175 
176 // call destructors: var..
177 
178 // ..and s_arr
179 
180 
181 // By OpenMP specifications, 'firstprivate' applies to both distribute and parallel for.
182 // However, the support for 'firstprivate' of 'parallel' is only used when 'parallel'
183 // is found alone. Therefore we only have one 'firstprivate' support for 'parallel for'
184 // in combination
185 
186 // addr alloca's
187 
188 // skip loop alloca's
189 
190 // private alloca's
191 
192 
193 // init addr alloca's with input values
194 
195 // init private alloca's with addr alloca's
196 // vec
197 
198 // s_arr
199 
200 // var
201 
202 
203 // call destructors: var..
204 
205 // ..and s_arr
206 
207 
208 // template tmain with S_INT_TY
209 
210 
211 
212 // addr alloca's
213 
214 // skip loop alloca's
215 
216 // private alloca's
217 
218 
219 // init addr alloca's with input values
220 
221 // init private alloca's with addr alloca's
222 // t-var
223 
224 // vec
225 
226 // s_arr
227 
228 // var
229 
230 // pass private alloca's to fork
231 // not dag to distinguish with S_VAR_CAST
232 
233 // call destructors: var..
234 
235 // ..and s_arr
236 
237 
238 // By OpenMP specifications, 'firstprivate' applies to both distribute and parallel for.
239 // However, the support for 'firstprivate' of 'parallel' is only used when 'parallel'
240 // is found alone. Therefore we only have one 'firstprivate' support for 'parallel for'
241 // in combination
242 
243 // addr alloca's
244 
245 // skip loop alloca's
246 
247 // private alloca's
248 
249 
250 // init addr alloca's with input values
251 
252 // init private alloca's with addr alloca's
253 // vec
254 
255 // s_arr
256 
257 // var
258 
259 
260 // call destructors: var..
261 
262 // ..and s_arr
263 
264 
265 #endif
266 // CHECK1-LABEL: define {{[^@]+}}@main
267 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
268 // CHECK1-NEXT:  entry:
269 // CHECK1-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
270 // CHECK1-NEXT:    [[G:%.*]] = alloca double, align 8
271 // CHECK1-NEXT:    [[G1:%.*]] = alloca ptr, align 8
272 // CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 8
273 // CHECK1-NEXT:    store i32 0, ptr [[RETVAL]], align 4
274 // CHECK1-NEXT:    store ptr [[G]], ptr [[G1]], align 8
275 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[CLASS_ANON]], ptr [[REF_TMP]], i32 0, i32 0
276 // CHECK1-NEXT:    store ptr [[G]], ptr [[TMP0]], align 8
277 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[CLASS_ANON]], ptr [[REF_TMP]], i32 0, i32 1
278 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[G1]], align 8
279 // CHECK1-NEXT:    store ptr [[TMP2]], ptr [[TMP1]], align 8
280 // CHECK1-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 8 dereferenceable(16) [[REF_TMP]])
281 // CHECK1-NEXT:    ret i32 0
282 //
283 //
284 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67
285 // CHECK1-SAME: (i64 noundef [[G:%.*]], i64 noundef [[G1:%.*]], i64 noundef [[SVAR:%.*]], i64 noundef [[SFVAR:%.*]]) #[[ATTR2:[0-9]+]] {
286 // CHECK1-NEXT:  entry:
287 // CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca i64, align 8
288 // CHECK1-NEXT:    [[G1_ADDR:%.*]] = alloca i64, align 8
289 // CHECK1-NEXT:    [[SVAR_ADDR:%.*]] = alloca i64, align 8
290 // CHECK1-NEXT:    [[SFVAR_ADDR:%.*]] = alloca i64, align 8
291 // CHECK1-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
292 // CHECK1-NEXT:    store i64 [[G]], ptr [[G_ADDR]], align 8
293 // CHECK1-NEXT:    store i64 [[G1]], ptr [[G1_ADDR]], align 8
294 // CHECK1-NEXT:    store i64 [[SVAR]], ptr [[SVAR_ADDR]], align 8
295 // CHECK1-NEXT:    store i64 [[SFVAR]], ptr [[SFVAR_ADDR]], align 8
296 // CHECK1-NEXT:    store ptr [[G1_ADDR]], ptr [[TMP]], align 8
297 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[TMP]], align 8
298 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67.omp_outlined, ptr [[G_ADDR]], ptr [[TMP0]], ptr [[SVAR_ADDR]], ptr [[SFVAR_ADDR]])
299 // CHECK1-NEXT:    ret void
300 //
301 //
302 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67.omp_outlined
303 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[G:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[G1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SVAR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SFVAR:%.*]]) #[[ATTR2]] {
304 // CHECK1-NEXT:  entry:
305 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
306 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
307 // CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 8
308 // CHECK1-NEXT:    [[G1_ADDR:%.*]] = alloca ptr, align 8
309 // CHECK1-NEXT:    [[SVAR_ADDR:%.*]] = alloca ptr, align 8
310 // CHECK1-NEXT:    [[SFVAR_ADDR:%.*]] = alloca ptr, align 8
311 // CHECK1-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
312 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
313 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
314 // CHECK1-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
315 // CHECK1-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
316 // CHECK1-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
317 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
318 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
319 // CHECK1-NEXT:    [[G3:%.*]] = alloca double, align 8
320 // CHECK1-NEXT:    [[G14:%.*]] = alloca double, align 8
321 // CHECK1-NEXT:    [[_TMP5:%.*]] = alloca ptr, align 8
322 // CHECK1-NEXT:    [[SVAR6:%.*]] = alloca i32, align 4
323 // CHECK1-NEXT:    [[SFVAR7:%.*]] = alloca float, align 4
324 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
325 // CHECK1-NEXT:    [[G_CASTED:%.*]] = alloca i64, align 8
326 // CHECK1-NEXT:    [[G1_CASTED:%.*]] = alloca i64, align 8
327 // CHECK1-NEXT:    [[SVAR_CASTED:%.*]] = alloca i64, align 8
328 // CHECK1-NEXT:    [[SFVAR_CASTED:%.*]] = alloca i64, align 8
329 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
330 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
331 // CHECK1-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 8
332 // CHECK1-NEXT:    store ptr [[G1]], ptr [[G1_ADDR]], align 8
333 // CHECK1-NEXT:    store ptr [[SVAR]], ptr [[SVAR_ADDR]], align 8
334 // CHECK1-NEXT:    store ptr [[SFVAR]], ptr [[SFVAR_ADDR]], align 8
335 // CHECK1-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[G_ADDR]], align 8
336 // CHECK1-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[G1_ADDR]], align 8
337 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[SVAR_ADDR]], align 8
338 // CHECK1-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[SFVAR_ADDR]], align 8
339 // CHECK1-NEXT:    store ptr [[TMP1]], ptr [[TMP]], align 8
340 // CHECK1-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[TMP]], align 8
341 // CHECK1-NEXT:    store ptr [[TMP4]], ptr [[_TMP1]], align 8
342 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
343 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_COMB_UB]], align 4
344 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
345 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
346 // CHECK1-NEXT:    [[TMP5:%.*]] = load double, ptr [[TMP0]], align 8
347 // CHECK1-NEXT:    store double [[TMP5]], ptr [[G3]], align 8
348 // CHECK1-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[_TMP1]], align 8
349 // CHECK1-NEXT:    [[TMP7:%.*]] = load double, ptr [[TMP6]], align 8
350 // CHECK1-NEXT:    store double [[TMP7]], ptr [[G14]], align 8
351 // CHECK1-NEXT:    store ptr [[G14]], ptr [[_TMP5]], align 8
352 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP2]], align 4
353 // CHECK1-NEXT:    store i32 [[TMP8]], ptr [[SVAR6]], align 4
354 // CHECK1-NEXT:    [[TMP9:%.*]] = load float, ptr [[TMP3]], align 4
355 // CHECK1-NEXT:    store float [[TMP9]], ptr [[SFVAR7]], align 4
356 // CHECK1-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
357 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
358 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
359 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
360 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP12]], 1
361 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
362 // CHECK1:       cond.true:
363 // CHECK1-NEXT:    br label [[COND_END:%.*]]
364 // CHECK1:       cond.false:
365 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
366 // CHECK1-NEXT:    br label [[COND_END]]
367 // CHECK1:       cond.end:
368 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
369 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
370 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
371 // CHECK1-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
372 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
373 // CHECK1:       omp.inner.for.cond:
374 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
375 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
376 // CHECK1-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
377 // CHECK1-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
378 // CHECK1:       omp.inner.for.body:
379 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
380 // CHECK1-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
381 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
382 // CHECK1-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
383 // CHECK1-NEXT:    [[TMP21:%.*]] = load double, ptr [[G3]], align 8
384 // CHECK1-NEXT:    store double [[TMP21]], ptr [[G_CASTED]], align 8
385 // CHECK1-NEXT:    [[TMP22:%.*]] = load i64, ptr [[G_CASTED]], align 8
386 // CHECK1-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[_TMP5]], align 8
387 // CHECK1-NEXT:    [[TMP24:%.*]] = load volatile double, ptr [[TMP23]], align 8
388 // CHECK1-NEXT:    store double [[TMP24]], ptr [[G1_CASTED]], align 8
389 // CHECK1-NEXT:    [[TMP25:%.*]] = load i64, ptr [[G1_CASTED]], align 8
390 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, ptr [[SVAR6]], align 4
391 // CHECK1-NEXT:    store i32 [[TMP26]], ptr [[SVAR_CASTED]], align 4
392 // CHECK1-NEXT:    [[TMP27:%.*]] = load i64, ptr [[SVAR_CASTED]], align 8
393 // CHECK1-NEXT:    [[TMP28:%.*]] = load float, ptr [[SFVAR7]], align 4
394 // CHECK1-NEXT:    store float [[TMP28]], ptr [[SFVAR_CASTED]], align 4
395 // CHECK1-NEXT:    [[TMP29:%.*]] = load i64, ptr [[SFVAR_CASTED]], align 8
396 // CHECK1-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 6, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67.omp_outlined.omp_outlined, i64 [[TMP18]], i64 [[TMP20]], i64 [[TMP22]], i64 [[TMP25]], i64 [[TMP27]], i64 [[TMP29]])
397 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
398 // CHECK1:       omp.inner.for.inc:
399 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
400 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
401 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP30]], [[TMP31]]
402 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
403 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
404 // CHECK1:       omp.inner.for.end:
405 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
406 // CHECK1:       omp.loop.exit:
407 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP11]])
408 // CHECK1-NEXT:    ret void
409 //
410 //
411 // CHECK1-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67.omp_outlined.omp_outlined
412 // CHECK1-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[G:%.*]], i64 noundef [[G1:%.*]], i64 noundef [[SVAR:%.*]], i64 noundef [[SFVAR:%.*]]) #[[ATTR2]] {
413 // CHECK1-NEXT:  entry:
414 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
415 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
416 // CHECK1-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
417 // CHECK1-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
418 // CHECK1-NEXT:    [[G_ADDR:%.*]] = alloca i64, align 8
419 // CHECK1-NEXT:    [[G1_ADDR:%.*]] = alloca i64, align 8
420 // CHECK1-NEXT:    [[SVAR_ADDR:%.*]] = alloca i64, align 8
421 // CHECK1-NEXT:    [[SFVAR_ADDR:%.*]] = alloca i64, align 8
422 // CHECK1-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
423 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
424 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
425 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
426 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
427 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
428 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
429 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
430 // CHECK1-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 8
431 // CHECK1-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
432 // CHECK1-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
433 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
434 // CHECK1-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
435 // CHECK1-NEXT:    store i64 [[G]], ptr [[G_ADDR]], align 8
436 // CHECK1-NEXT:    store i64 [[G1]], ptr [[G1_ADDR]], align 8
437 // CHECK1-NEXT:    store i64 [[SVAR]], ptr [[SVAR_ADDR]], align 8
438 // CHECK1-NEXT:    store i64 [[SFVAR]], ptr [[SFVAR_ADDR]], align 8
439 // CHECK1-NEXT:    store ptr [[G1_ADDR]], ptr [[TMP]], align 8
440 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
441 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
442 // CHECK1-NEXT:    [[TMP0:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
443 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
444 // CHECK1-NEXT:    [[TMP1:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
445 // CHECK1-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP1]] to i32
446 // CHECK1-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
447 // CHECK1-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
448 // CHECK1-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
449 // CHECK1-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
450 // CHECK1-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
451 // CHECK1-NEXT:    [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4
452 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP3]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
453 // CHECK1-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
454 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 1
455 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
456 // CHECK1:       cond.true:
457 // CHECK1-NEXT:    br label [[COND_END:%.*]]
458 // CHECK1:       cond.false:
459 // CHECK1-NEXT:    [[TMP5:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
460 // CHECK1-NEXT:    br label [[COND_END]]
461 // CHECK1:       cond.end:
462 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
463 // CHECK1-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
464 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
465 // CHECK1-NEXT:    store i32 [[TMP6]], ptr [[DOTOMP_IV]], align 4
466 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
467 // CHECK1:       omp.inner.for.cond:
468 // CHECK1-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
469 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
470 // CHECK1-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
471 // CHECK1-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
472 // CHECK1:       omp.inner.for.body:
473 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
474 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
475 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
476 // CHECK1-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
477 // CHECK1-NEXT:    store double 1.000000e+00, ptr [[G_ADDR]], align 8
478 // CHECK1-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[TMP]], align 8
479 // CHECK1-NEXT:    store volatile double 1.000000e+00, ptr [[TMP10]], align 8
480 // CHECK1-NEXT:    store i32 3, ptr [[SVAR_ADDR]], align 4
481 // CHECK1-NEXT:    store float 4.000000e+00, ptr [[SFVAR_ADDR]], align 4
482 // CHECK1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
483 // CHECK1-NEXT:    store ptr [[G_ADDR]], ptr [[TMP11]], align 8
484 // CHECK1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 1
485 // CHECK1-NEXT:    [[TMP13:%.*]] = load ptr, ptr [[TMP]], align 8
486 // CHECK1-NEXT:    store ptr [[TMP13]], ptr [[TMP12]], align 8
487 // CHECK1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 2
488 // CHECK1-NEXT:    store ptr [[SVAR_ADDR]], ptr [[TMP14]], align 8
489 // CHECK1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 3
490 // CHECK1-NEXT:    store ptr [[SFVAR_ADDR]], ptr [[TMP15]], align 8
491 // CHECK1-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr noundef nonnull align 8 dereferenceable(32) [[REF_TMP]])
492 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
493 // CHECK1:       omp.body.continue:
494 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
495 // CHECK1:       omp.inner.for.inc:
496 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
497 // CHECK1-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP16]], 1
498 // CHECK1-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
499 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
500 // CHECK1:       omp.inner.for.end:
501 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
502 // CHECK1:       omp.loop.exit:
503 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP3]])
504 // CHECK1-NEXT:    ret void
505 //
506 //
507 // CHECK3-LABEL: define {{[^@]+}}@main
508 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
509 // CHECK3-NEXT:  entry:
510 // CHECK3-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
511 // CHECK3-NEXT:    [[G:%.*]] = alloca double, align 8
512 // CHECK3-NEXT:    [[G1:%.*]] = alloca ptr, align 4
513 // CHECK3-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON:%.*]], align 4
514 // CHECK3-NEXT:    store i32 0, ptr [[RETVAL]], align 4
515 // CHECK3-NEXT:    store ptr [[G]], ptr [[G1]], align 4
516 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds nuw [[CLASS_ANON]], ptr [[REF_TMP]], i32 0, i32 0
517 // CHECK3-NEXT:    store ptr [[G]], ptr [[TMP0]], align 4
518 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [[CLASS_ANON]], ptr [[REF_TMP]], i32 0, i32 1
519 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[G1]], align 4
520 // CHECK3-NEXT:    store ptr [[TMP2]], ptr [[TMP1]], align 4
521 // CHECK3-NEXT:    call void @"_ZZ4mainENK3$_0clEv"(ptr noundef nonnull align 4 dereferenceable(8) [[REF_TMP]])
522 // CHECK3-NEXT:    ret i32 0
523 //
524 //
525 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67
526 // CHECK3-SAME: (ptr noundef nonnull align 4 dereferenceable(8) [[G:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[G1:%.*]], i32 noundef [[SVAR:%.*]], i32 noundef [[SFVAR:%.*]]) #[[ATTR2:[0-9]+]] {
527 // CHECK3-NEXT:  entry:
528 // CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
529 // CHECK3-NEXT:    [[G1_ADDR:%.*]] = alloca ptr, align 4
530 // CHECK3-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32, align 4
531 // CHECK3-NEXT:    [[SFVAR_ADDR:%.*]] = alloca i32, align 4
532 // CHECK3-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
533 // CHECK3-NEXT:    [[G2:%.*]] = alloca double, align 8
534 // CHECK3-NEXT:    [[G13:%.*]] = alloca double, align 8
535 // CHECK3-NEXT:    [[_TMP4:%.*]] = alloca ptr, align 4
536 // CHECK3-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
537 // CHECK3-NEXT:    store ptr [[G1]], ptr [[G1_ADDR]], align 4
538 // CHECK3-NEXT:    store i32 [[SVAR]], ptr [[SVAR_ADDR]], align 4
539 // CHECK3-NEXT:    store i32 [[SFVAR]], ptr [[SFVAR_ADDR]], align 4
540 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[G_ADDR]], align 4
541 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[G1_ADDR]], align 4
542 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[TMP]], align 4
543 // CHECK3-NEXT:    [[TMP2:%.*]] = load double, ptr [[TMP0]], align 8
544 // CHECK3-NEXT:    store double [[TMP2]], ptr [[G2]], align 8
545 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 4
546 // CHECK3-NEXT:    [[TMP4:%.*]] = load volatile double, ptr [[TMP3]], align 4
547 // CHECK3-NEXT:    store double [[TMP4]], ptr [[G13]], align 8
548 // CHECK3-NEXT:    store ptr [[G13]], ptr [[_TMP4]], align 4
549 // CHECK3-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[_TMP4]], align 4
550 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3:[0-9]+]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67.omp_outlined, ptr [[G2]], ptr [[TMP5]], ptr [[SVAR_ADDR]], ptr [[SFVAR_ADDR]])
551 // CHECK3-NEXT:    ret void
552 //
553 //
554 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67.omp_outlined
555 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[G:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[G1:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SVAR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SFVAR:%.*]]) #[[ATTR2]] {
556 // CHECK3-NEXT:  entry:
557 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
558 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
559 // CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
560 // CHECK3-NEXT:    [[G1_ADDR:%.*]] = alloca ptr, align 4
561 // CHECK3-NEXT:    [[SVAR_ADDR:%.*]] = alloca ptr, align 4
562 // CHECK3-NEXT:    [[SFVAR_ADDR:%.*]] = alloca ptr, align 4
563 // CHECK3-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
564 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
565 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
566 // CHECK3-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
567 // CHECK3-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
568 // CHECK3-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
569 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
570 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
571 // CHECK3-NEXT:    [[G3:%.*]] = alloca double, align 8
572 // CHECK3-NEXT:    [[G14:%.*]] = alloca double, align 8
573 // CHECK3-NEXT:    [[_TMP5:%.*]] = alloca ptr, align 4
574 // CHECK3-NEXT:    [[SVAR6:%.*]] = alloca i32, align 4
575 // CHECK3-NEXT:    [[SFVAR7:%.*]] = alloca float, align 4
576 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
577 // CHECK3-NEXT:    [[G1_CASTED:%.*]] = alloca i32, align 4
578 // CHECK3-NEXT:    [[SVAR_CASTED:%.*]] = alloca i32, align 4
579 // CHECK3-NEXT:    [[SFVAR_CASTED:%.*]] = alloca i32, align 4
580 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
581 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
582 // CHECK3-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
583 // CHECK3-NEXT:    store ptr [[G1]], ptr [[G1_ADDR]], align 4
584 // CHECK3-NEXT:    store ptr [[SVAR]], ptr [[SVAR_ADDR]], align 4
585 // CHECK3-NEXT:    store ptr [[SFVAR]], ptr [[SFVAR_ADDR]], align 4
586 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[G_ADDR]], align 4
587 // CHECK3-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[G1_ADDR]], align 4
588 // CHECK3-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[SVAR_ADDR]], align 4
589 // CHECK3-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[SFVAR_ADDR]], align 4
590 // CHECK3-NEXT:    store ptr [[TMP1]], ptr [[TMP]], align 4
591 // CHECK3-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[TMP]], align 4
592 // CHECK3-NEXT:    store ptr [[TMP4]], ptr [[_TMP1]], align 4
593 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
594 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_COMB_UB]], align 4
595 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
596 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
597 // CHECK3-NEXT:    [[TMP5:%.*]] = load double, ptr [[TMP0]], align 8
598 // CHECK3-NEXT:    store double [[TMP5]], ptr [[G3]], align 8
599 // CHECK3-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[_TMP1]], align 4
600 // CHECK3-NEXT:    [[TMP7:%.*]] = load double, ptr [[TMP6]], align 4
601 // CHECK3-NEXT:    store double [[TMP7]], ptr [[G14]], align 8
602 // CHECK3-NEXT:    store ptr [[G14]], ptr [[_TMP5]], align 4
603 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP2]], align 4
604 // CHECK3-NEXT:    store i32 [[TMP8]], ptr [[SVAR6]], align 4
605 // CHECK3-NEXT:    [[TMP9:%.*]] = load float, ptr [[TMP3]], align 4
606 // CHECK3-NEXT:    store float [[TMP9]], ptr [[SFVAR7]], align 4
607 // CHECK3-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
608 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
609 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
610 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
611 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP12]], 1
612 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
613 // CHECK3:       cond.true:
614 // CHECK3-NEXT:    br label [[COND_END:%.*]]
615 // CHECK3:       cond.false:
616 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
617 // CHECK3-NEXT:    br label [[COND_END]]
618 // CHECK3:       cond.end:
619 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
620 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
621 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
622 // CHECK3-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
623 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
624 // CHECK3:       omp.inner.for.cond:
625 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
626 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
627 // CHECK3-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
628 // CHECK3-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
629 // CHECK3:       omp.inner.for.body:
630 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
631 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
632 // CHECK3-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[_TMP5]], align 4
633 // CHECK3-NEXT:    [[TMP20:%.*]] = load volatile double, ptr [[TMP19]], align 4
634 // CHECK3-NEXT:    store double [[TMP20]], ptr [[G1_CASTED]], align 4
635 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, ptr [[G1_CASTED]], align 4
636 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, ptr [[SVAR6]], align 4
637 // CHECK3-NEXT:    store i32 [[TMP22]], ptr [[SVAR_CASTED]], align 4
638 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, ptr [[SVAR_CASTED]], align 4
639 // CHECK3-NEXT:    [[TMP24:%.*]] = load float, ptr [[SFVAR7]], align 4
640 // CHECK3-NEXT:    store float [[TMP24]], ptr [[SFVAR_CASTED]], align 4
641 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, ptr [[SFVAR_CASTED]], align 4
642 // CHECK3-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 6, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67.omp_outlined.omp_outlined, i32 [[TMP17]], i32 [[TMP18]], ptr [[G3]], i32 [[TMP21]], i32 [[TMP23]], i32 [[TMP25]])
643 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
644 // CHECK3:       omp.inner.for.inc:
645 // CHECK3-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
646 // CHECK3-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
647 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP26]], [[TMP27]]
648 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
649 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
650 // CHECK3:       omp.inner.for.end:
651 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
652 // CHECK3:       omp.loop.exit:
653 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP11]])
654 // CHECK3-NEXT:    ret void
655 //
656 //
657 // CHECK3-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l67.omp_outlined.omp_outlined
658 // CHECK3-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[G:%.*]], i32 noundef [[G1:%.*]], i32 noundef [[SVAR:%.*]], i32 noundef [[SFVAR:%.*]]) #[[ATTR2]] {
659 // CHECK3-NEXT:  entry:
660 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
661 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
662 // CHECK3-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
663 // CHECK3-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
664 // CHECK3-NEXT:    [[G_ADDR:%.*]] = alloca ptr, align 4
665 // CHECK3-NEXT:    [[G1_ADDR:%.*]] = alloca i32, align 4
666 // CHECK3-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32, align 4
667 // CHECK3-NEXT:    [[SFVAR_ADDR:%.*]] = alloca i32, align 4
668 // CHECK3-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
669 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
670 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
671 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
672 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
673 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
674 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
675 // CHECK3-NEXT:    [[G2:%.*]] = alloca double, align 8
676 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
677 // CHECK3-NEXT:    [[REF_TMP:%.*]] = alloca [[CLASS_ANON_0:%.*]], align 4
678 // CHECK3-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
679 // CHECK3-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
680 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
681 // CHECK3-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
682 // CHECK3-NEXT:    store ptr [[G]], ptr [[G_ADDR]], align 4
683 // CHECK3-NEXT:    store i32 [[G1]], ptr [[G1_ADDR]], align 4
684 // CHECK3-NEXT:    store i32 [[SVAR]], ptr [[SVAR_ADDR]], align 4
685 // CHECK3-NEXT:    store i32 [[SFVAR]], ptr [[SFVAR_ADDR]], align 4
686 // CHECK3-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[G_ADDR]], align 4
687 // CHECK3-NEXT:    store ptr [[G1_ADDR]], ptr [[TMP]], align 4
688 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
689 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
690 // CHECK3-NEXT:    [[TMP1:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
691 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
692 // CHECK3-NEXT:    store i32 [[TMP1]], ptr [[DOTOMP_LB]], align 4
693 // CHECK3-NEXT:    store i32 [[TMP2]], ptr [[DOTOMP_UB]], align 4
694 // CHECK3-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
695 // CHECK3-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
696 // CHECK3-NEXT:    [[TMP3:%.*]] = load double, ptr [[TMP0]], align 8
697 // CHECK3-NEXT:    store double [[TMP3]], ptr [[G2]], align 8
698 // CHECK3-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
699 // CHECK3-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP4]], align 4
700 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP5]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
701 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
702 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP6]], 1
703 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
704 // CHECK3:       cond.true:
705 // CHECK3-NEXT:    br label [[COND_END:%.*]]
706 // CHECK3:       cond.false:
707 // CHECK3-NEXT:    [[TMP7:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
708 // CHECK3-NEXT:    br label [[COND_END]]
709 // CHECK3:       cond.end:
710 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP7]], [[COND_FALSE]] ]
711 // CHECK3-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
712 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
713 // CHECK3-NEXT:    store i32 [[TMP8]], ptr [[DOTOMP_IV]], align 4
714 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
715 // CHECK3:       omp.inner.for.cond:
716 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
717 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
718 // CHECK3-NEXT:    [[CMP3:%.*]] = icmp sle i32 [[TMP9]], [[TMP10]]
719 // CHECK3-NEXT:    br i1 [[CMP3]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
720 // CHECK3:       omp.inner.for.body:
721 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
722 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP11]], 1
723 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
724 // CHECK3-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
725 // CHECK3-NEXT:    store double 1.000000e+00, ptr [[G2]], align 8
726 // CHECK3-NEXT:    [[TMP12:%.*]] = load ptr, ptr [[TMP]], align 4
727 // CHECK3-NEXT:    store volatile double 1.000000e+00, ptr [[TMP12]], align 4
728 // CHECK3-NEXT:    store i32 3, ptr [[SVAR_ADDR]], align 4
729 // CHECK3-NEXT:    store float 4.000000e+00, ptr [[SFVAR_ADDR]], align 4
730 // CHECK3-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 0
731 // CHECK3-NEXT:    store ptr [[G2]], ptr [[TMP13]], align 4
732 // CHECK3-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 1
733 // CHECK3-NEXT:    [[TMP15:%.*]] = load ptr, ptr [[TMP]], align 4
734 // CHECK3-NEXT:    store ptr [[TMP15]], ptr [[TMP14]], align 4
735 // CHECK3-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 2
736 // CHECK3-NEXT:    store ptr [[SVAR_ADDR]], ptr [[TMP16]], align 4
737 // CHECK3-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw [[CLASS_ANON_0]], ptr [[REF_TMP]], i32 0, i32 3
738 // CHECK3-NEXT:    store ptr [[SFVAR_ADDR]], ptr [[TMP17]], align 4
739 // CHECK3-NEXT:    call void @"_ZZZ4mainENK3$_0clEvENKUlvE_clEv"(ptr noundef nonnull align 4 dereferenceable(16) [[REF_TMP]])
740 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
741 // CHECK3:       omp.body.continue:
742 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
743 // CHECK3:       omp.inner.for.inc:
744 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
745 // CHECK3-NEXT:    [[ADD4:%.*]] = add nsw i32 [[TMP18]], 1
746 // CHECK3-NEXT:    store i32 [[ADD4]], ptr [[DOTOMP_IV]], align 4
747 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
748 // CHECK3:       omp.inner.for.end:
749 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
750 // CHECK3:       omp.loop.exit:
751 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP5]])
752 // CHECK3-NEXT:    ret void
753 //
754 //
755 // CHECK8-LABEL: define {{[^@]+}}@main
756 // CHECK8-SAME: () #[[ATTR0:[0-9]+]] {
757 // CHECK8-NEXT:  entry:
758 // CHECK8-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
759 // CHECK8-NEXT:    [[G:%.*]] = alloca double, align 8
760 // CHECK8-NEXT:    [[G1:%.*]] = alloca ptr, align 8
761 // CHECK8-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
762 // CHECK8-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
763 // CHECK8-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
764 // CHECK8-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
765 // CHECK8-NEXT:    [[VAR:%.*]] = alloca ptr, align 8
766 // CHECK8-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
767 // CHECK8-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
768 // CHECK8-NEXT:    [[SVAR_CASTED:%.*]] = alloca i64, align 8
769 // CHECK8-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 8
770 // CHECK8-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 8
771 // CHECK8-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 8
772 // CHECK8-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
773 // CHECK8-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
774 // CHECK8-NEXT:    store i32 0, ptr [[RETVAL]], align 4
775 // CHECK8-NEXT:    store ptr [[G]], ptr [[G1]], align 8
776 // CHECK8-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
777 // CHECK8-NEXT:    store i32 0, ptr [[T_VAR]], align 4
778 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC]], ptr align 4 @__const.main.vec, i64 8, i1 false)
779 // CHECK8-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[S_ARR]], float noundef 1.000000e+00)
780 // CHECK8-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[S_ARR]], i64 1
781 // CHECK8-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float noundef 2.000000e+00)
782 // CHECK8-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 8
783 // CHECK8-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VAR]], align 8
784 // CHECK8-NEXT:    store ptr [[TMP0]], ptr [[TMP]], align 8
785 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, ptr [[T_VAR]], align 4
786 // CHECK8-NEXT:    store i32 [[TMP1]], ptr [[T_VAR_CASTED]], align 4
787 // CHECK8-NEXT:    [[TMP2:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
788 // CHECK8-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 8
789 // CHECK8-NEXT:    [[TMP4:%.*]] = load i32, ptr @_ZZ4mainE4svar, align 4
790 // CHECK8-NEXT:    store i32 [[TMP4]], ptr [[SVAR_CASTED]], align 4
791 // CHECK8-NEXT:    [[TMP5:%.*]] = load i64, ptr [[SVAR_CASTED]], align 8
792 // CHECK8-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[TMP]], align 8
793 // CHECK8-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP]], align 8
794 // CHECK8-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
795 // CHECK8-NEXT:    store i64 [[TMP2]], ptr [[TMP8]], align 8
796 // CHECK8-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
797 // CHECK8-NEXT:    store i64 [[TMP2]], ptr [[TMP9]], align 8
798 // CHECK8-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
799 // CHECK8-NEXT:    store ptr null, ptr [[TMP10]], align 8
800 // CHECK8-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
801 // CHECK8-NEXT:    store ptr [[VEC]], ptr [[TMP11]], align 8
802 // CHECK8-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
803 // CHECK8-NEXT:    store ptr [[VEC]], ptr [[TMP12]], align 8
804 // CHECK8-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
805 // CHECK8-NEXT:    store ptr null, ptr [[TMP13]], align 8
806 // CHECK8-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
807 // CHECK8-NEXT:    store ptr [[S_ARR]], ptr [[TMP14]], align 8
808 // CHECK8-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
809 // CHECK8-NEXT:    store ptr [[S_ARR]], ptr [[TMP15]], align 8
810 // CHECK8-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
811 // CHECK8-NEXT:    store ptr null, ptr [[TMP16]], align 8
812 // CHECK8-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
813 // CHECK8-NEXT:    store ptr [[TMP6]], ptr [[TMP17]], align 8
814 // CHECK8-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
815 // CHECK8-NEXT:    store ptr [[TMP7]], ptr [[TMP18]], align 8
816 // CHECK8-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
817 // CHECK8-NEXT:    store ptr null, ptr [[TMP19]], align 8
818 // CHECK8-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
819 // CHECK8-NEXT:    store i64 [[TMP5]], ptr [[TMP20]], align 8
820 // CHECK8-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
821 // CHECK8-NEXT:    store i64 [[TMP5]], ptr [[TMP21]], align 8
822 // CHECK8-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 4
823 // CHECK8-NEXT:    store ptr null, ptr [[TMP22]], align 8
824 // CHECK8-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
825 // CHECK8-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
826 // CHECK8-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
827 // CHECK8-NEXT:    store i32 3, ptr [[TMP25]], align 4
828 // CHECK8-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
829 // CHECK8-NEXT:    store i32 5, ptr [[TMP26]], align 4
830 // CHECK8-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
831 // CHECK8-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 8
832 // CHECK8-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
833 // CHECK8-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 8
834 // CHECK8-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
835 // CHECK8-NEXT:    store ptr @.offload_sizes, ptr [[TMP29]], align 8
836 // CHECK8-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
837 // CHECK8-NEXT:    store ptr @.offload_maptypes, ptr [[TMP30]], align 8
838 // CHECK8-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
839 // CHECK8-NEXT:    store ptr null, ptr [[TMP31]], align 8
840 // CHECK8-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
841 // CHECK8-NEXT:    store ptr null, ptr [[TMP32]], align 8
842 // CHECK8-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
843 // CHECK8-NEXT:    store i64 2, ptr [[TMP33]], align 8
844 // CHECK8-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
845 // CHECK8-NEXT:    store i64 0, ptr [[TMP34]], align 8
846 // CHECK8-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
847 // CHECK8-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
848 // CHECK8-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
849 // CHECK8-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
850 // CHECK8-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
851 // CHECK8-NEXT:    store i32 0, ptr [[TMP37]], align 4
852 // CHECK8-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.region_id, ptr [[KERNEL_ARGS]])
853 // CHECK8-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
854 // CHECK8-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
855 // CHECK8:       omp_offload.failed:
856 // CHECK8-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139(i64 [[TMP2]], ptr [[VEC]], ptr [[S_ARR]], ptr [[TMP3]], i64 [[TMP5]]) #[[ATTR4:[0-9]+]]
857 // CHECK8-NEXT:    br label [[OMP_OFFLOAD_CONT]]
858 // CHECK8:       omp_offload.cont:
859 // CHECK8-NEXT:    [[CALL:%.*]] = call noundef signext i32 @_Z5tmainIiET_v()
860 // CHECK8-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
861 // CHECK8-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
862 // CHECK8-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
863 // CHECK8-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
864 // CHECK8:       arraydestroy.body:
865 // CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP40]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
866 // CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
867 // CHECK8-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
868 // CHECK8-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
869 // CHECK8-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
870 // CHECK8:       arraydestroy.done2:
871 // CHECK8-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
872 // CHECK8-NEXT:    [[TMP41:%.*]] = load i32, ptr [[RETVAL]], align 4
873 // CHECK8-NEXT:    ret i32 [[TMP41]]
874 //
875 //
876 // CHECK8-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
877 // CHECK8-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat {
878 // CHECK8-NEXT:  entry:
879 // CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
880 // CHECK8-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
881 // CHECK8-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
882 // CHECK8-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
883 // CHECK8-NEXT:    ret void
884 //
885 //
886 // CHECK8-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
887 // CHECK8-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
888 // CHECK8-NEXT:  entry:
889 // CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
890 // CHECK8-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
891 // CHECK8-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
892 // CHECK8-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
893 // CHECK8-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
894 // CHECK8-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
895 // CHECK8-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
896 // CHECK8-NEXT:    ret void
897 //
898 //
899 // CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139
900 // CHECK8-SAME: (i64 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]], i64 noundef [[SVAR:%.*]]) #[[ATTR3:[0-9]+]] {
901 // CHECK8-NEXT:  entry:
902 // CHECK8-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
903 // CHECK8-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
904 // CHECK8-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
905 // CHECK8-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
906 // CHECK8-NEXT:    [[SVAR_ADDR:%.*]] = alloca i64, align 8
907 // CHECK8-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
908 // CHECK8-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
909 // CHECK8-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
910 // CHECK8-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
911 // CHECK8-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
912 // CHECK8-NEXT:    store i64 [[SVAR]], ptr [[SVAR_ADDR]], align 8
913 // CHECK8-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
914 // CHECK8-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
915 // CHECK8-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
916 // CHECK8-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 8
917 // CHECK8-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 8
918 // CHECK8-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.omp_outlined, ptr [[T_VAR_ADDR]], ptr [[TMP0]], ptr [[TMP1]], ptr [[TMP3]], ptr [[SVAR_ADDR]])
919 // CHECK8-NEXT:    ret void
920 //
921 //
922 // CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.omp_outlined
923 // CHECK8-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SVAR:%.*]]) #[[ATTR3]] {
924 // CHECK8-NEXT:  entry:
925 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
926 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
927 // CHECK8-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
928 // CHECK8-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
929 // CHECK8-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
930 // CHECK8-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
931 // CHECK8-NEXT:    [[SVAR_ADDR:%.*]] = alloca ptr, align 8
932 // CHECK8-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
933 // CHECK8-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
934 // CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
935 // CHECK8-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
936 // CHECK8-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
937 // CHECK8-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
938 // CHECK8-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
939 // CHECK8-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
940 // CHECK8-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
941 // CHECK8-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
942 // CHECK8-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S], align 4
943 // CHECK8-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S:%.*]], align 4
944 // CHECK8-NEXT:    [[_TMP8:%.*]] = alloca ptr, align 8
945 // CHECK8-NEXT:    [[SVAR9:%.*]] = alloca i32, align 4
946 // CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
947 // CHECK8-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
948 // CHECK8-NEXT:    [[SVAR_CASTED:%.*]] = alloca i64, align 8
949 // CHECK8-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
950 // CHECK8-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
951 // CHECK8-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
952 // CHECK8-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
953 // CHECK8-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
954 // CHECK8-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
955 // CHECK8-NEXT:    store ptr [[SVAR]], ptr [[SVAR_ADDR]], align 8
956 // CHECK8-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
957 // CHECK8-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
958 // CHECK8-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
959 // CHECK8-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
960 // CHECK8-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[SVAR_ADDR]], align 8
961 // CHECK8-NEXT:    store ptr [[TMP3]], ptr [[TMP]], align 8
962 // CHECK8-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP]], align 8
963 // CHECK8-NEXT:    store ptr [[TMP5]], ptr [[_TMP1]], align 8
964 // CHECK8-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
965 // CHECK8-NEXT:    store i32 1, ptr [[DOTOMP_COMB_UB]], align 4
966 // CHECK8-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
967 // CHECK8-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
968 // CHECK8-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP0]], align 4
969 // CHECK8-NEXT:    store i32 [[TMP6]], ptr [[T_VAR3]], align 4
970 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC4]], ptr align 4 [[TMP1]], i64 8, i1 false)
971 // CHECK8-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR5]], i32 0, i32 0
972 // CHECK8-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
973 // CHECK8-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP7]]
974 // CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
975 // CHECK8:       omp.arraycpy.body:
976 // CHECK8-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
977 // CHECK8-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
978 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr align 4 [[OMP_ARRAYCPY_SRCELEMENTPAST]], i64 4, i1 false)
979 // CHECK8-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
980 // CHECK8-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
981 // CHECK8-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP7]]
982 // CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
983 // CHECK8:       omp.arraycpy.done6:
984 // CHECK8-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[_TMP1]], align 8
985 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VAR7]], ptr align 4 [[TMP8]], i64 4, i1 false)
986 // CHECK8-NEXT:    store ptr [[VAR7]], ptr [[_TMP8]], align 8
987 // CHECK8-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP4]], align 4
988 // CHECK8-NEXT:    store i32 [[TMP9]], ptr [[SVAR9]], align 4
989 // CHECK8-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
990 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
991 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
992 // CHECK8-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
993 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP12]], 1
994 // CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
995 // CHECK8:       cond.true:
996 // CHECK8-NEXT:    br label [[COND_END:%.*]]
997 // CHECK8:       cond.false:
998 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
999 // CHECK8-NEXT:    br label [[COND_END]]
1000 // CHECK8:       cond.end:
1001 // CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
1002 // CHECK8-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1003 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1004 // CHECK8-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
1005 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1006 // CHECK8:       omp.inner.for.cond:
1007 // CHECK8-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1008 // CHECK8-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1009 // CHECK8-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
1010 // CHECK8-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
1011 // CHECK8:       omp.inner.for.cond.cleanup:
1012 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
1013 // CHECK8:       omp.inner.for.body:
1014 // CHECK8-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1015 // CHECK8-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
1016 // CHECK8-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1017 // CHECK8-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
1018 // CHECK8-NEXT:    [[TMP21:%.*]] = load i32, ptr [[T_VAR3]], align 4
1019 // CHECK8-NEXT:    store i32 [[TMP21]], ptr [[T_VAR_CASTED]], align 4
1020 // CHECK8-NEXT:    [[TMP22:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
1021 // CHECK8-NEXT:    [[TMP23:%.*]] = load ptr, ptr [[_TMP8]], align 8
1022 // CHECK8-NEXT:    [[TMP24:%.*]] = load i32, ptr [[SVAR9]], align 4
1023 // CHECK8-NEXT:    store i32 [[TMP24]], ptr [[SVAR_CASTED]], align 4
1024 // CHECK8-NEXT:    [[TMP25:%.*]] = load i64, ptr [[SVAR_CASTED]], align 8
1025 // CHECK8-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.omp_outlined.omp_outlined, i64 [[TMP18]], i64 [[TMP20]], ptr [[VEC4]], i64 [[TMP22]], ptr [[S_ARR5]], ptr [[TMP23]], i64 [[TMP25]])
1026 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1027 // CHECK8:       omp.inner.for.inc:
1028 // CHECK8-NEXT:    [[TMP26:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1029 // CHECK8-NEXT:    [[TMP27:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1030 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP26]], [[TMP27]]
1031 // CHECK8-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1032 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
1033 // CHECK8:       omp.inner.for.end:
1034 // CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1035 // CHECK8:       omp.loop.exit:
1036 // CHECK8-NEXT:    [[TMP28:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1037 // CHECK8-NEXT:    [[TMP29:%.*]] = load i32, ptr [[TMP28]], align 4
1038 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP29]])
1039 // CHECK8-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
1040 // CHECK8-NEXT:    [[ARRAY_BEGIN11:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR5]], i32 0, i32 0
1041 // CHECK8-NEXT:    [[TMP30:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN11]], i64 2
1042 // CHECK8-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
1043 // CHECK8:       arraydestroy.body:
1044 // CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP30]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
1045 // CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
1046 // CHECK8-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
1047 // CHECK8-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN11]]
1048 // CHECK8-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE12:%.*]], label [[ARRAYDESTROY_BODY]]
1049 // CHECK8:       arraydestroy.done12:
1050 // CHECK8-NEXT:    ret void
1051 //
1052 //
1053 // CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.omp_outlined.omp_outlined
1054 // CHECK8-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], i64 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]], i64 noundef [[SVAR:%.*]]) #[[ATTR3]] {
1055 // CHECK8-NEXT:  entry:
1056 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1057 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1058 // CHECK8-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1059 // CHECK8-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1060 // CHECK8-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
1061 // CHECK8-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
1062 // CHECK8-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
1063 // CHECK8-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
1064 // CHECK8-NEXT:    [[SVAR_ADDR:%.*]] = alloca i64, align 8
1065 // CHECK8-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
1066 // CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1067 // CHECK8-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1068 // CHECK8-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1069 // CHECK8-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1070 // CHECK8-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1071 // CHECK8-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1072 // CHECK8-NEXT:    [[VEC3:%.*]] = alloca [2 x i32], align 4
1073 // CHECK8-NEXT:    [[S_ARR4:%.*]] = alloca [2 x %struct.S], align 4
1074 // CHECK8-NEXT:    [[VAR6:%.*]] = alloca [[STRUCT_S:%.*]], align 4
1075 // CHECK8-NEXT:    [[_TMP7:%.*]] = alloca ptr, align 8
1076 // CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
1077 // CHECK8-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1078 // CHECK8-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1079 // CHECK8-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1080 // CHECK8-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1081 // CHECK8-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
1082 // CHECK8-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
1083 // CHECK8-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
1084 // CHECK8-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
1085 // CHECK8-NEXT:    store i64 [[SVAR]], ptr [[SVAR_ADDR]], align 8
1086 // CHECK8-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
1087 // CHECK8-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
1088 // CHECK8-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
1089 // CHECK8-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 8
1090 // CHECK8-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1091 // CHECK8-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
1092 // CHECK8-NEXT:    [[TMP3:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1093 // CHECK8-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP3]] to i32
1094 // CHECK8-NEXT:    [[TMP4:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1095 // CHECK8-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP4]] to i32
1096 // CHECK8-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
1097 // CHECK8-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
1098 // CHECK8-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1099 // CHECK8-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1100 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC3]], ptr align 4 [[TMP0]], i64 8, i1 false)
1101 // CHECK8-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR4]], i32 0, i32 0
1102 // CHECK8-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i64 2
1103 // CHECK8-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP5]]
1104 // CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1105 // CHECK8:       omp.arraycpy.body:
1106 // CHECK8-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1107 // CHECK8-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1108 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr align 4 [[OMP_ARRAYCPY_SRCELEMENTPAST]], i64 4, i1 false)
1109 // CHECK8-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1110 // CHECK8-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1111 // CHECK8-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
1112 // CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
1113 // CHECK8:       omp.arraycpy.done5:
1114 // CHECK8-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[TMP]], align 8
1115 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VAR6]], ptr align 4 [[TMP6]], i64 4, i1 false)
1116 // CHECK8-NEXT:    store ptr [[VAR6]], ptr [[_TMP7]], align 8
1117 // CHECK8-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1118 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
1119 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1120 // CHECK8-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1121 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1
1122 // CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1123 // CHECK8:       cond.true:
1124 // CHECK8-NEXT:    br label [[COND_END:%.*]]
1125 // CHECK8:       cond.false:
1126 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1127 // CHECK8-NEXT:    br label [[COND_END]]
1128 // CHECK8:       cond.end:
1129 // CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
1130 // CHECK8-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1131 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1132 // CHECK8-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
1133 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1134 // CHECK8:       omp.inner.for.cond:
1135 // CHECK8-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1136 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1137 // CHECK8-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
1138 // CHECK8-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
1139 // CHECK8:       omp.inner.for.cond.cleanup:
1140 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
1141 // CHECK8:       omp.inner.for.body:
1142 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1143 // CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
1144 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1145 // CHECK8-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1146 // CHECK8-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
1147 // CHECK8-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4
1148 // CHECK8-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
1149 // CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC3]], i64 0, i64 [[IDXPROM]]
1150 // CHECK8-NEXT:    store i32 [[TMP15]], ptr [[ARRAYIDX]], align 4
1151 // CHECK8-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[_TMP7]], align 8
1152 // CHECK8-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I]], align 4
1153 // CHECK8-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP18]] to i64
1154 // CHECK8-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR4]], i64 0, i64 [[IDXPROM9]]
1155 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX10]], ptr align 4 [[TMP17]], i64 4, i1 false)
1156 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1157 // CHECK8:       omp.body.continue:
1158 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1159 // CHECK8:       omp.inner.for.inc:
1160 // CHECK8-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1161 // CHECK8-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP19]], 1
1162 // CHECK8-NEXT:    store i32 [[ADD11]], ptr [[DOTOMP_IV]], align 4
1163 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
1164 // CHECK8:       omp.inner.for.end:
1165 // CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1166 // CHECK8:       omp.loop.exit:
1167 // CHECK8-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1168 // CHECK8-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
1169 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
1170 // CHECK8-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR6]]) #[[ATTR4]]
1171 // CHECK8-NEXT:    [[ARRAY_BEGIN12:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR4]], i32 0, i32 0
1172 // CHECK8-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN12]], i64 2
1173 // CHECK8-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
1174 // CHECK8:       arraydestroy.body:
1175 // CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP22]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
1176 // CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
1177 // CHECK8-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
1178 // CHECK8-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN12]]
1179 // CHECK8-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE13:%.*]], label [[ARRAYDESTROY_BODY]]
1180 // CHECK8:       arraydestroy.done13:
1181 // CHECK8-NEXT:    ret void
1182 //
1183 //
1184 // CHECK8-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
1185 // CHECK8-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
1186 // CHECK8-NEXT:  entry:
1187 // CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1188 // CHECK8-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
1189 // CHECK8-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
1190 // CHECK8-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
1191 // CHECK8-NEXT:    ret void
1192 //
1193 //
1194 // CHECK8-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
1195 // CHECK8-SAME: () #[[ATTR1]] comdat {
1196 // CHECK8-NEXT:  entry:
1197 // CHECK8-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1198 // CHECK8-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
1199 // CHECK8-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
1200 // CHECK8-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
1201 // CHECK8-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
1202 // CHECK8-NEXT:    [[VAR:%.*]] = alloca ptr, align 8
1203 // CHECK8-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
1204 // CHECK8-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
1205 // CHECK8-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x ptr], align 8
1206 // CHECK8-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x ptr], align 8
1207 // CHECK8-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x ptr], align 8
1208 // CHECK8-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1209 // CHECK8-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1210 // CHECK8-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
1211 // CHECK8-NEXT:    store i32 0, ptr [[T_VAR]], align 4
1212 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC]], ptr align 4 @__const._Z5tmainIiET_v.vec, i64 8, i1 false)
1213 // CHECK8-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[S_ARR]], i32 noundef signext 1)
1214 // CHECK8-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[S_ARR]], i64 1
1215 // CHECK8-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 noundef signext 2)
1216 // CHECK8-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 8
1217 // CHECK8-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VAR]], align 8
1218 // CHECK8-NEXT:    store ptr [[TMP0]], ptr [[TMP]], align 8
1219 // CHECK8-NEXT:    [[TMP1:%.*]] = load i32, ptr [[T_VAR]], align 4
1220 // CHECK8-NEXT:    store i32 [[TMP1]], ptr [[T_VAR_CASTED]], align 4
1221 // CHECK8-NEXT:    [[TMP2:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
1222 // CHECK8-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 8
1223 // CHECK8-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[TMP]], align 8
1224 // CHECK8-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP]], align 8
1225 // CHECK8-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1226 // CHECK8-NEXT:    store i64 [[TMP2]], ptr [[TMP6]], align 8
1227 // CHECK8-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1228 // CHECK8-NEXT:    store i64 [[TMP2]], ptr [[TMP7]], align 8
1229 // CHECK8-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 0
1230 // CHECK8-NEXT:    store ptr null, ptr [[TMP8]], align 8
1231 // CHECK8-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1232 // CHECK8-NEXT:    store ptr [[VEC]], ptr [[TMP9]], align 8
1233 // CHECK8-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1234 // CHECK8-NEXT:    store ptr [[VEC]], ptr [[TMP10]], align 8
1235 // CHECK8-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 1
1236 // CHECK8-NEXT:    store ptr null, ptr [[TMP11]], align 8
1237 // CHECK8-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1238 // CHECK8-NEXT:    store ptr [[S_ARR]], ptr [[TMP12]], align 8
1239 // CHECK8-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1240 // CHECK8-NEXT:    store ptr [[S_ARR]], ptr [[TMP13]], align 8
1241 // CHECK8-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 2
1242 // CHECK8-NEXT:    store ptr null, ptr [[TMP14]], align 8
1243 // CHECK8-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1244 // CHECK8-NEXT:    store ptr [[TMP4]], ptr [[TMP15]], align 8
1245 // CHECK8-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1246 // CHECK8-NEXT:    store ptr [[TMP5]], ptr [[TMP16]], align 8
1247 // CHECK8-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i64 0, i64 3
1248 // CHECK8-NEXT:    store ptr null, ptr [[TMP17]], align 8
1249 // CHECK8-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1250 // CHECK8-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1251 // CHECK8-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1252 // CHECK8-NEXT:    store i32 3, ptr [[TMP20]], align 4
1253 // CHECK8-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1254 // CHECK8-NEXT:    store i32 4, ptr [[TMP21]], align 4
1255 // CHECK8-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1256 // CHECK8-NEXT:    store ptr [[TMP18]], ptr [[TMP22]], align 8
1257 // CHECK8-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1258 // CHECK8-NEXT:    store ptr [[TMP19]], ptr [[TMP23]], align 8
1259 // CHECK8-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1260 // CHECK8-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP24]], align 8
1261 // CHECK8-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1262 // CHECK8-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP25]], align 8
1263 // CHECK8-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1264 // CHECK8-NEXT:    store ptr null, ptr [[TMP26]], align 8
1265 // CHECK8-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1266 // CHECK8-NEXT:    store ptr null, ptr [[TMP27]], align 8
1267 // CHECK8-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1268 // CHECK8-NEXT:    store i64 2, ptr [[TMP28]], align 8
1269 // CHECK8-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1270 // CHECK8-NEXT:    store i64 0, ptr [[TMP29]], align 8
1271 // CHECK8-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1272 // CHECK8-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP30]], align 4
1273 // CHECK8-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1274 // CHECK8-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP31]], align 4
1275 // CHECK8-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1276 // CHECK8-NEXT:    store i32 0, ptr [[TMP32]], align 4
1277 // CHECK8-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.region_id, ptr [[KERNEL_ARGS]])
1278 // CHECK8-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
1279 // CHECK8-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1280 // CHECK8:       omp_offload.failed:
1281 // CHECK8-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48(i64 [[TMP2]], ptr [[VEC]], ptr [[S_ARR]], ptr [[TMP3]]) #[[ATTR4]]
1282 // CHECK8-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1283 // CHECK8:       omp_offload.cont:
1284 // CHECK8-NEXT:    store i32 0, ptr [[RETVAL]], align 4
1285 // CHECK8-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
1286 // CHECK8-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
1287 // CHECK8-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
1288 // CHECK8:       arraydestroy.body:
1289 // CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP35]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
1290 // CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
1291 // CHECK8-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
1292 // CHECK8-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
1293 // CHECK8-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
1294 // CHECK8:       arraydestroy.done2:
1295 // CHECK8-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
1296 // CHECK8-NEXT:    [[TMP36:%.*]] = load i32, ptr [[RETVAL]], align 4
1297 // CHECK8-NEXT:    ret i32 [[TMP36]]
1298 //
1299 //
1300 // CHECK8-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
1301 // CHECK8-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
1302 // CHECK8-NEXT:  entry:
1303 // CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1304 // CHECK8-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
1305 // CHECK8-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
1306 // CHECK8-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
1307 // CHECK8-NEXT:    store float 0.000000e+00, ptr [[F]], align 4
1308 // CHECK8-NEXT:    ret void
1309 //
1310 //
1311 // CHECK8-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
1312 // CHECK8-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
1313 // CHECK8-NEXT:  entry:
1314 // CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1315 // CHECK8-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
1316 // CHECK8-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
1317 // CHECK8-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
1318 // CHECK8-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
1319 // CHECK8-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
1320 // CHECK8-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
1321 // CHECK8-NEXT:    store float [[TMP0]], ptr [[F]], align 4
1322 // CHECK8-NEXT:    ret void
1323 //
1324 //
1325 // CHECK8-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
1326 // CHECK8-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
1327 // CHECK8-NEXT:  entry:
1328 // CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1329 // CHECK8-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
1330 // CHECK8-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
1331 // CHECK8-NEXT:    ret void
1332 //
1333 //
1334 // CHECK8-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
1335 // CHECK8-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
1336 // CHECK8-NEXT:  entry:
1337 // CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1338 // CHECK8-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
1339 // CHECK8-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
1340 // CHECK8-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
1341 // CHECK8-NEXT:    ret void
1342 //
1343 //
1344 // CHECK8-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
1345 // CHECK8-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
1346 // CHECK8-NEXT:  entry:
1347 // CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1348 // CHECK8-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1349 // CHECK8-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
1350 // CHECK8-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
1351 // CHECK8-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
1352 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
1353 // CHECK8-NEXT:    call void @_ZN1SIiEC2Ei(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], i32 noundef signext [[TMP0]])
1354 // CHECK8-NEXT:    ret void
1355 //
1356 //
1357 // CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48
1358 // CHECK8-SAME: (i64 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
1359 // CHECK8-NEXT:  entry:
1360 // CHECK8-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
1361 // CHECK8-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
1362 // CHECK8-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
1363 // CHECK8-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
1364 // CHECK8-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
1365 // CHECK8-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
1366 // CHECK8-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
1367 // CHECK8-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
1368 // CHECK8-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
1369 // CHECK8-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
1370 // CHECK8-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
1371 // CHECK8-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
1372 // CHECK8-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 8
1373 // CHECK8-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 8
1374 // CHECK8-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.omp_outlined, ptr [[T_VAR_ADDR]], ptr [[TMP0]], ptr [[TMP1]], ptr [[TMP3]])
1375 // CHECK8-NEXT:    ret void
1376 //
1377 //
1378 // CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.omp_outlined
1379 // CHECK8-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
1380 // CHECK8-NEXT:  entry:
1381 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1382 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1383 // CHECK8-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 8
1384 // CHECK8-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
1385 // CHECK8-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
1386 // CHECK8-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
1387 // CHECK8-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
1388 // CHECK8-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 8
1389 // CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1390 // CHECK8-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
1391 // CHECK8-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1392 // CHECK8-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1393 // CHECK8-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1394 // CHECK8-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1395 // CHECK8-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
1396 // CHECK8-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
1397 // CHECK8-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S.0], align 4
1398 // CHECK8-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
1399 // CHECK8-NEXT:    [[_TMP8:%.*]] = alloca ptr, align 8
1400 // CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
1401 // CHECK8-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i64, align 8
1402 // CHECK8-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1403 // CHECK8-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1404 // CHECK8-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
1405 // CHECK8-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
1406 // CHECK8-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
1407 // CHECK8-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
1408 // CHECK8-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 8
1409 // CHECK8-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
1410 // CHECK8-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
1411 // CHECK8-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
1412 // CHECK8-NEXT:    store ptr [[TMP3]], ptr [[TMP]], align 8
1413 // CHECK8-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[TMP]], align 8
1414 // CHECK8-NEXT:    store ptr [[TMP4]], ptr [[_TMP1]], align 8
1415 // CHECK8-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1416 // CHECK8-NEXT:    store i32 1, ptr [[DOTOMP_COMB_UB]], align 4
1417 // CHECK8-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1418 // CHECK8-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1419 // CHECK8-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
1420 // CHECK8-NEXT:    store i32 [[TMP5]], ptr [[T_VAR3]], align 4
1421 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC4]], ptr align 4 [[TMP1]], i64 8, i1 false)
1422 // CHECK8-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR5]], i32 0, i32 0
1423 // CHECK8-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
1424 // CHECK8-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP6]]
1425 // CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1426 // CHECK8:       omp.arraycpy.body:
1427 // CHECK8-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1428 // CHECK8-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1429 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr align 4 [[OMP_ARRAYCPY_SRCELEMENTPAST]], i64 4, i1 false)
1430 // CHECK8-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1431 // CHECK8-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1432 // CHECK8-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP6]]
1433 // CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
1434 // CHECK8:       omp.arraycpy.done6:
1435 // CHECK8-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[_TMP1]], align 8
1436 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VAR7]], ptr align 4 [[TMP7]], i64 4, i1 false)
1437 // CHECK8-NEXT:    store ptr [[VAR7]], ptr [[_TMP8]], align 8
1438 // CHECK8-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1439 // CHECK8-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
1440 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP9]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1441 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1442 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP10]], 1
1443 // CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1444 // CHECK8:       cond.true:
1445 // CHECK8-NEXT:    br label [[COND_END:%.*]]
1446 // CHECK8:       cond.false:
1447 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1448 // CHECK8-NEXT:    br label [[COND_END]]
1449 // CHECK8:       cond.end:
1450 // CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1451 // CHECK8-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1452 // CHECK8-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1453 // CHECK8-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
1454 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1455 // CHECK8:       omp.inner.for.cond:
1456 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1457 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1458 // CHECK8-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1459 // CHECK8-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
1460 // CHECK8:       omp.inner.for.cond.cleanup:
1461 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
1462 // CHECK8:       omp.inner.for.body:
1463 // CHECK8-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1464 // CHECK8-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP15]] to i64
1465 // CHECK8-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1466 // CHECK8-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
1467 // CHECK8-NEXT:    [[TMP19:%.*]] = load i32, ptr [[T_VAR3]], align 4
1468 // CHECK8-NEXT:    store i32 [[TMP19]], ptr [[T_VAR_CASTED]], align 4
1469 // CHECK8-NEXT:    [[TMP20:%.*]] = load i64, ptr [[T_VAR_CASTED]], align 8
1470 // CHECK8-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[_TMP8]], align 8
1471 // CHECK8-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 6, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.omp_outlined.omp_outlined, i64 [[TMP16]], i64 [[TMP18]], ptr [[VEC4]], i64 [[TMP20]], ptr [[S_ARR5]], ptr [[TMP21]])
1472 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1473 // CHECK8:       omp.inner.for.inc:
1474 // CHECK8-NEXT:    [[TMP22:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1475 // CHECK8-NEXT:    [[TMP23:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1476 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP22]], [[TMP23]]
1477 // CHECK8-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1478 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
1479 // CHECK8:       omp.inner.for.end:
1480 // CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1481 // CHECK8:       omp.loop.exit:
1482 // CHECK8-NEXT:    [[TMP24:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1483 // CHECK8-NEXT:    [[TMP25:%.*]] = load i32, ptr [[TMP24]], align 4
1484 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP25]])
1485 // CHECK8-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
1486 // CHECK8-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR5]], i32 0, i32 0
1487 // CHECK8-NEXT:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN10]], i64 2
1488 // CHECK8-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
1489 // CHECK8:       arraydestroy.body:
1490 // CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP26]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
1491 // CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
1492 // CHECK8-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
1493 // CHECK8-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
1494 // CHECK8-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
1495 // CHECK8:       arraydestroy.done11:
1496 // CHECK8-NEXT:    ret void
1497 //
1498 //
1499 // CHECK8-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.omp_outlined.omp_outlined
1500 // CHECK8-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], i64 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
1501 // CHECK8-NEXT:  entry:
1502 // CHECK8-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8
1503 // CHECK8-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8
1504 // CHECK8-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8
1505 // CHECK8-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8
1506 // CHECK8-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 8
1507 // CHECK8-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i64, align 8
1508 // CHECK8-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 8
1509 // CHECK8-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 8
1510 // CHECK8-NEXT:    [[TMP:%.*]] = alloca ptr, align 8
1511 // CHECK8-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1512 // CHECK8-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1513 // CHECK8-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1514 // CHECK8-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1515 // CHECK8-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1516 // CHECK8-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1517 // CHECK8-NEXT:    [[VEC3:%.*]] = alloca [2 x i32], align 4
1518 // CHECK8-NEXT:    [[S_ARR4:%.*]] = alloca [2 x %struct.S.0], align 4
1519 // CHECK8-NEXT:    [[VAR6:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
1520 // CHECK8-NEXT:    [[_TMP7:%.*]] = alloca ptr, align 8
1521 // CHECK8-NEXT:    [[I:%.*]] = alloca i32, align 4
1522 // CHECK8-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 8
1523 // CHECK8-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 8
1524 // CHECK8-NEXT:    store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1525 // CHECK8-NEXT:    store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1526 // CHECK8-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 8
1527 // CHECK8-NEXT:    store i64 [[T_VAR]], ptr [[T_VAR_ADDR]], align 8
1528 // CHECK8-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 8
1529 // CHECK8-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 8
1530 // CHECK8-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 8
1531 // CHECK8-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 8
1532 // CHECK8-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 8
1533 // CHECK8-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 8
1534 // CHECK8-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
1535 // CHECK8-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
1536 // CHECK8-NEXT:    [[TMP3:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR]], align 8
1537 // CHECK8-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP3]] to i32
1538 // CHECK8-NEXT:    [[TMP4:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR]], align 8
1539 // CHECK8-NEXT:    [[CONV2:%.*]] = trunc i64 [[TMP4]] to i32
1540 // CHECK8-NEXT:    store i32 [[CONV]], ptr [[DOTOMP_LB]], align 4
1541 // CHECK8-NEXT:    store i32 [[CONV2]], ptr [[DOTOMP_UB]], align 4
1542 // CHECK8-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1543 // CHECK8-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1544 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VEC3]], ptr align 4 [[TMP0]], i64 8, i1 false)
1545 // CHECK8-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR4]], i32 0, i32 0
1546 // CHECK8-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i64 2
1547 // CHECK8-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP5]]
1548 // CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE5:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1549 // CHECK8:       omp.arraycpy.body:
1550 // CHECK8-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1551 // CHECK8-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1552 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr align 4 [[OMP_ARRAYCPY_SRCELEMENTPAST]], i64 4, i1 false)
1553 // CHECK8-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1554 // CHECK8-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1555 // CHECK8-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
1556 // CHECK8-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE5]], label [[OMP_ARRAYCPY_BODY]]
1557 // CHECK8:       omp.arraycpy.done5:
1558 // CHECK8-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[TMP]], align 8
1559 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[VAR6]], ptr align 4 [[TMP6]], i64 4, i1 false)
1560 // CHECK8-NEXT:    store ptr [[VAR6]], ptr [[_TMP7]], align 8
1561 // CHECK8-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1562 // CHECK8-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
1563 // CHECK8-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1564 // CHECK8-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1565 // CHECK8-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1
1566 // CHECK8-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1567 // CHECK8:       cond.true:
1568 // CHECK8-NEXT:    br label [[COND_END:%.*]]
1569 // CHECK8:       cond.false:
1570 // CHECK8-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1571 // CHECK8-NEXT:    br label [[COND_END]]
1572 // CHECK8:       cond.end:
1573 // CHECK8-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
1574 // CHECK8-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
1575 // CHECK8-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
1576 // CHECK8-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
1577 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1578 // CHECK8:       omp.inner.for.cond:
1579 // CHECK8-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1580 // CHECK8-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
1581 // CHECK8-NEXT:    [[CMP8:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
1582 // CHECK8-NEXT:    br i1 [[CMP8]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
1583 // CHECK8:       omp.inner.for.cond.cleanup:
1584 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
1585 // CHECK8:       omp.inner.for.body:
1586 // CHECK8-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1587 // CHECK8-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
1588 // CHECK8-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
1589 // CHECK8-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
1590 // CHECK8-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
1591 // CHECK8-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4
1592 // CHECK8-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP16]] to i64
1593 // CHECK8-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC3]], i64 0, i64 [[IDXPROM]]
1594 // CHECK8-NEXT:    store i32 [[TMP15]], ptr [[ARRAYIDX]], align 4
1595 // CHECK8-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[_TMP7]], align 8
1596 // CHECK8-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I]], align 4
1597 // CHECK8-NEXT:    [[IDXPROM9:%.*]] = sext i32 [[TMP18]] to i64
1598 // CHECK8-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR4]], i64 0, i64 [[IDXPROM9]]
1599 // CHECK8-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[ARRAYIDX10]], ptr align 4 [[TMP17]], i64 4, i1 false)
1600 // CHECK8-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1601 // CHECK8:       omp.body.continue:
1602 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1603 // CHECK8:       omp.inner.for.inc:
1604 // CHECK8-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1605 // CHECK8-NEXT:    [[ADD11:%.*]] = add nsw i32 [[TMP19]], 1
1606 // CHECK8-NEXT:    store i32 [[ADD11]], ptr [[DOTOMP_IV]], align 4
1607 // CHECK8-NEXT:    br label [[OMP_INNER_FOR_COND]]
1608 // CHECK8:       omp.inner.for.end:
1609 // CHECK8-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1610 // CHECK8:       omp.loop.exit:
1611 // CHECK8-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 8
1612 // CHECK8-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
1613 // CHECK8-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
1614 // CHECK8-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR6]]) #[[ATTR4]]
1615 // CHECK8-NEXT:    [[ARRAY_BEGIN12:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR4]], i32 0, i32 0
1616 // CHECK8-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN12]], i64 2
1617 // CHECK8-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
1618 // CHECK8:       arraydestroy.body:
1619 // CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP22]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
1620 // CHECK8-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i64 -1
1621 // CHECK8-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
1622 // CHECK8-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN12]]
1623 // CHECK8-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE13:%.*]], label [[ARRAYDESTROY_BODY]]
1624 // CHECK8:       arraydestroy.done13:
1625 // CHECK8-NEXT:    ret void
1626 //
1627 //
1628 // CHECK8-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
1629 // CHECK8-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
1630 // CHECK8-NEXT:  entry:
1631 // CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1632 // CHECK8-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
1633 // CHECK8-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
1634 // CHECK8-NEXT:    call void @_ZN1SIiED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
1635 // CHECK8-NEXT:    ret void
1636 //
1637 //
1638 // CHECK8-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
1639 // CHECK8-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
1640 // CHECK8-NEXT:  entry:
1641 // CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1642 // CHECK8-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
1643 // CHECK8-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
1644 // CHECK8-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
1645 // CHECK8-NEXT:    store i32 0, ptr [[F]], align 4
1646 // CHECK8-NEXT:    ret void
1647 //
1648 //
1649 // CHECK8-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
1650 // CHECK8-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef signext [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat {
1651 // CHECK8-NEXT:  entry:
1652 // CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1653 // CHECK8-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
1654 // CHECK8-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
1655 // CHECK8-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
1656 // CHECK8-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
1657 // CHECK8-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
1658 // CHECK8-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
1659 // CHECK8-NEXT:    store i32 [[TMP0]], ptr [[F]], align 4
1660 // CHECK8-NEXT:    ret void
1661 //
1662 //
1663 // CHECK8-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
1664 // CHECK8-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat {
1665 // CHECK8-NEXT:  entry:
1666 // CHECK8-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 8
1667 // CHECK8-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 8
1668 // CHECK8-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 8
1669 // CHECK8-NEXT:    ret void
1670 //
1671 //
1672 // CHECK10-LABEL: define {{[^@]+}}@main
1673 // CHECK10-SAME: () #[[ATTR0:[0-9]+]] {
1674 // CHECK10-NEXT:  entry:
1675 // CHECK10-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
1676 // CHECK10-NEXT:    [[G:%.*]] = alloca double, align 8
1677 // CHECK10-NEXT:    [[G1:%.*]] = alloca ptr, align 4
1678 // CHECK10-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S:%.*]], align 4
1679 // CHECK10-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
1680 // CHECK10-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
1681 // CHECK10-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S], align 4
1682 // CHECK10-NEXT:    [[VAR:%.*]] = alloca ptr, align 4
1683 // CHECK10-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
1684 // CHECK10-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
1685 // CHECK10-NEXT:    [[SVAR_CASTED:%.*]] = alloca i32, align 4
1686 // CHECK10-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [5 x ptr], align 4
1687 // CHECK10-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [5 x ptr], align 4
1688 // CHECK10-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [5 x ptr], align 4
1689 // CHECK10-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1690 // CHECK10-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
1691 // CHECK10-NEXT:    store i32 0, ptr [[RETVAL]], align 4
1692 // CHECK10-NEXT:    store ptr [[G]], ptr [[G1]], align 4
1693 // CHECK10-NEXT:    call void @_ZN1SIfEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
1694 // CHECK10-NEXT:    store i32 0, ptr [[T_VAR]], align 4
1695 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC]], ptr align 4 @__const.main.vec, i32 8, i1 false)
1696 // CHECK10-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[S_ARR]], float noundef 1.000000e+00)
1697 // CHECK10-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[S_ARR]], i32 1
1698 // CHECK10-NEXT:    call void @_ZN1SIfEC1Ef(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], float noundef 2.000000e+00)
1699 // CHECK10-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 4
1700 // CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VAR]], align 4
1701 // CHECK10-NEXT:    store ptr [[TMP0]], ptr [[TMP]], align 4
1702 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, ptr [[T_VAR]], align 4
1703 // CHECK10-NEXT:    store i32 [[TMP1]], ptr [[T_VAR_CASTED]], align 4
1704 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
1705 // CHECK10-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 4
1706 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, ptr @_ZZ4mainE4svar, align 4
1707 // CHECK10-NEXT:    store i32 [[TMP4]], ptr [[SVAR_CASTED]], align 4
1708 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, ptr [[SVAR_CASTED]], align 4
1709 // CHECK10-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[TMP]], align 4
1710 // CHECK10-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[TMP]], align 4
1711 // CHECK10-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1712 // CHECK10-NEXT:    store i32 [[TMP2]], ptr [[TMP8]], align 4
1713 // CHECK10-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1714 // CHECK10-NEXT:    store i32 [[TMP2]], ptr [[TMP9]], align 4
1715 // CHECK10-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
1716 // CHECK10-NEXT:    store ptr null, ptr [[TMP10]], align 4
1717 // CHECK10-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
1718 // CHECK10-NEXT:    store ptr [[VEC]], ptr [[TMP11]], align 4
1719 // CHECK10-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
1720 // CHECK10-NEXT:    store ptr [[VEC]], ptr [[TMP12]], align 4
1721 // CHECK10-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
1722 // CHECK10-NEXT:    store ptr null, ptr [[TMP13]], align 4
1723 // CHECK10-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
1724 // CHECK10-NEXT:    store ptr [[S_ARR]], ptr [[TMP14]], align 4
1725 // CHECK10-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
1726 // CHECK10-NEXT:    store ptr [[S_ARR]], ptr [[TMP15]], align 4
1727 // CHECK10-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
1728 // CHECK10-NEXT:    store ptr null, ptr [[TMP16]], align 4
1729 // CHECK10-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
1730 // CHECK10-NEXT:    store ptr [[TMP6]], ptr [[TMP17]], align 4
1731 // CHECK10-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
1732 // CHECK10-NEXT:    store ptr [[TMP7]], ptr [[TMP18]], align 4
1733 // CHECK10-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
1734 // CHECK10-NEXT:    store ptr null, ptr [[TMP19]], align 4
1735 // CHECK10-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 4
1736 // CHECK10-NEXT:    store i32 [[TMP5]], ptr [[TMP20]], align 4
1737 // CHECK10-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 4
1738 // CHECK10-NEXT:    store i32 [[TMP5]], ptr [[TMP21]], align 4
1739 // CHECK10-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 4
1740 // CHECK10-NEXT:    store ptr null, ptr [[TMP22]], align 4
1741 // CHECK10-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
1742 // CHECK10-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [5 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
1743 // CHECK10-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
1744 // CHECK10-NEXT:    store i32 3, ptr [[TMP25]], align 4
1745 // CHECK10-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
1746 // CHECK10-NEXT:    store i32 5, ptr [[TMP26]], align 4
1747 // CHECK10-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
1748 // CHECK10-NEXT:    store ptr [[TMP23]], ptr [[TMP27]], align 4
1749 // CHECK10-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
1750 // CHECK10-NEXT:    store ptr [[TMP24]], ptr [[TMP28]], align 4
1751 // CHECK10-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
1752 // CHECK10-NEXT:    store ptr @.offload_sizes, ptr [[TMP29]], align 4
1753 // CHECK10-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
1754 // CHECK10-NEXT:    store ptr @.offload_maptypes, ptr [[TMP30]], align 4
1755 // CHECK10-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
1756 // CHECK10-NEXT:    store ptr null, ptr [[TMP31]], align 4
1757 // CHECK10-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
1758 // CHECK10-NEXT:    store ptr null, ptr [[TMP32]], align 4
1759 // CHECK10-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
1760 // CHECK10-NEXT:    store i64 2, ptr [[TMP33]], align 8
1761 // CHECK10-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
1762 // CHECK10-NEXT:    store i64 0, ptr [[TMP34]], align 8
1763 // CHECK10-NEXT:    [[TMP35:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
1764 // CHECK10-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP35]], align 4
1765 // CHECK10-NEXT:    [[TMP36:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
1766 // CHECK10-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP36]], align 4
1767 // CHECK10-NEXT:    [[TMP37:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
1768 // CHECK10-NEXT:    store i32 0, ptr [[TMP37]], align 4
1769 // CHECK10-NEXT:    [[TMP38:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3:[0-9]+]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.region_id, ptr [[KERNEL_ARGS]])
1770 // CHECK10-NEXT:    [[TMP39:%.*]] = icmp ne i32 [[TMP38]], 0
1771 // CHECK10-NEXT:    br i1 [[TMP39]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
1772 // CHECK10:       omp_offload.failed:
1773 // CHECK10-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139(i32 [[TMP2]], ptr [[VEC]], ptr [[S_ARR]], ptr [[TMP3]], i32 [[TMP5]]) #[[ATTR4:[0-9]+]]
1774 // CHECK10-NEXT:    br label [[OMP_OFFLOAD_CONT]]
1775 // CHECK10:       omp_offload.cont:
1776 // CHECK10-NEXT:    [[CALL:%.*]] = call noundef i32 @_Z5tmainIiET_v()
1777 // CHECK10-NEXT:    store i32 [[CALL]], ptr [[RETVAL]], align 4
1778 // CHECK10-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR]], i32 0, i32 0
1779 // CHECK10-NEXT:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
1780 // CHECK10-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
1781 // CHECK10:       arraydestroy.body:
1782 // CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP40]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
1783 // CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
1784 // CHECK10-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
1785 // CHECK10-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
1786 // CHECK10-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
1787 // CHECK10:       arraydestroy.done2:
1788 // CHECK10-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
1789 // CHECK10-NEXT:    [[TMP41:%.*]] = load i32, ptr [[RETVAL]], align 4
1790 // CHECK10-NEXT:    ret i32 [[TMP41]]
1791 //
1792 //
1793 // CHECK10-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ev
1794 // CHECK10-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1:[0-9]+]] comdat align 2 {
1795 // CHECK10-NEXT:  entry:
1796 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
1797 // CHECK10-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
1798 // CHECK10-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
1799 // CHECK10-NEXT:    call void @_ZN1SIfEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
1800 // CHECK10-NEXT:    ret void
1801 //
1802 //
1803 // CHECK10-LABEL: define {{[^@]+}}@_ZN1SIfEC1Ef
1804 // CHECK10-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
1805 // CHECK10-NEXT:  entry:
1806 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
1807 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
1808 // CHECK10-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
1809 // CHECK10-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
1810 // CHECK10-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
1811 // CHECK10-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
1812 // CHECK10-NEXT:    call void @_ZN1SIfEC2Ef(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], float noundef [[TMP0]])
1813 // CHECK10-NEXT:    ret void
1814 //
1815 //
1816 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139
1817 // CHECK10-SAME: (i32 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32 noundef [[SVAR:%.*]]) #[[ATTR3:[0-9]+]] {
1818 // CHECK10-NEXT:  entry:
1819 // CHECK10-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
1820 // CHECK10-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
1821 // CHECK10-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
1822 // CHECK10-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
1823 // CHECK10-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32, align 4
1824 // CHECK10-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
1825 // CHECK10-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
1826 // CHECK10-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
1827 // CHECK10-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
1828 // CHECK10-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
1829 // CHECK10-NEXT:    store i32 [[SVAR]], ptr [[SVAR_ADDR]], align 4
1830 // CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
1831 // CHECK10-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
1832 // CHECK10-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
1833 // CHECK10-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 4
1834 // CHECK10-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 4
1835 // CHECK10-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 5, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.omp_outlined, ptr [[T_VAR_ADDR]], ptr [[TMP0]], ptr [[TMP1]], ptr [[TMP3]], ptr [[SVAR_ADDR]])
1836 // CHECK10-NEXT:    ret void
1837 //
1838 //
1839 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.omp_outlined
1840 // CHECK10-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[SVAR:%.*]]) #[[ATTR3]] {
1841 // CHECK10-NEXT:  entry:
1842 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1843 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1844 // CHECK10-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
1845 // CHECK10-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
1846 // CHECK10-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
1847 // CHECK10-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
1848 // CHECK10-NEXT:    [[SVAR_ADDR:%.*]] = alloca ptr, align 4
1849 // CHECK10-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
1850 // CHECK10-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
1851 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1852 // CHECK10-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
1853 // CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
1854 // CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
1855 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1856 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1857 // CHECK10-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
1858 // CHECK10-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
1859 // CHECK10-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S], align 4
1860 // CHECK10-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S:%.*]], align 4
1861 // CHECK10-NEXT:    [[_TMP8:%.*]] = alloca ptr, align 4
1862 // CHECK10-NEXT:    [[SVAR9:%.*]] = alloca i32, align 4
1863 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
1864 // CHECK10-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
1865 // CHECK10-NEXT:    [[SVAR_CASTED:%.*]] = alloca i32, align 4
1866 // CHECK10-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1867 // CHECK10-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1868 // CHECK10-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
1869 // CHECK10-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
1870 // CHECK10-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
1871 // CHECK10-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
1872 // CHECK10-NEXT:    store ptr [[SVAR]], ptr [[SVAR_ADDR]], align 4
1873 // CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
1874 // CHECK10-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
1875 // CHECK10-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
1876 // CHECK10-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
1877 // CHECK10-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[SVAR_ADDR]], align 4
1878 // CHECK10-NEXT:    store ptr [[TMP3]], ptr [[TMP]], align 4
1879 // CHECK10-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP]], align 4
1880 // CHECK10-NEXT:    store ptr [[TMP5]], ptr [[_TMP1]], align 4
1881 // CHECK10-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
1882 // CHECK10-NEXT:    store i32 1, ptr [[DOTOMP_COMB_UB]], align 4
1883 // CHECK10-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
1884 // CHECK10-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
1885 // CHECK10-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP0]], align 4
1886 // CHECK10-NEXT:    store i32 [[TMP6]], ptr [[T_VAR3]], align 4
1887 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC4]], ptr align 4 [[TMP1]], i32 8, i1 false)
1888 // CHECK10-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR5]], i32 0, i32 0
1889 // CHECK10-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
1890 // CHECK10-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP7]]
1891 // CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
1892 // CHECK10:       omp.arraycpy.body:
1893 // CHECK10-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1894 // CHECK10-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
1895 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr align 4 [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 4, i1 false)
1896 // CHECK10-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
1897 // CHECK10-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
1898 // CHECK10-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP7]]
1899 // CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
1900 // CHECK10:       omp.arraycpy.done6:
1901 // CHECK10-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[_TMP1]], align 4
1902 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VAR7]], ptr align 4 [[TMP8]], i32 4, i1 false)
1903 // CHECK10-NEXT:    store ptr [[VAR7]], ptr [[_TMP8]], align 4
1904 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP4]], align 4
1905 // CHECK10-NEXT:    store i32 [[TMP9]], ptr [[SVAR9]], align 4
1906 // CHECK10-NEXT:    [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1907 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
1908 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1:[0-9]+]], i32 [[TMP11]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
1909 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1910 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP12]], 1
1911 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1912 // CHECK10:       cond.true:
1913 // CHECK10-NEXT:    br label [[COND_END:%.*]]
1914 // CHECK10:       cond.false:
1915 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1916 // CHECK10-NEXT:    br label [[COND_END]]
1917 // CHECK10:       cond.end:
1918 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
1919 // CHECK10-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
1920 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1921 // CHECK10-NEXT:    store i32 [[TMP14]], ptr [[DOTOMP_IV]], align 4
1922 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1923 // CHECK10:       omp.inner.for.cond:
1924 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1925 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1926 // CHECK10-NEXT:    [[CMP10:%.*]] = icmp sle i32 [[TMP15]], [[TMP16]]
1927 // CHECK10-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
1928 // CHECK10:       omp.inner.for.cond.cleanup:
1929 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
1930 // CHECK10:       omp.inner.for.body:
1931 // CHECK10-NEXT:    [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
1932 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
1933 // CHECK10-NEXT:    [[TMP19:%.*]] = load i32, ptr [[T_VAR3]], align 4
1934 // CHECK10-NEXT:    store i32 [[TMP19]], ptr [[T_VAR_CASTED]], align 4
1935 // CHECK10-NEXT:    [[TMP20:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
1936 // CHECK10-NEXT:    [[TMP21:%.*]] = load ptr, ptr [[_TMP8]], align 4
1937 // CHECK10-NEXT:    [[TMP22:%.*]] = load i32, ptr [[SVAR9]], align 4
1938 // CHECK10-NEXT:    store i32 [[TMP22]], ptr [[SVAR_CASTED]], align 4
1939 // CHECK10-NEXT:    [[TMP23:%.*]] = load i32, ptr [[SVAR_CASTED]], align 4
1940 // CHECK10-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 7, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.omp_outlined.omp_outlined, i32 [[TMP17]], i32 [[TMP18]], ptr [[VEC4]], i32 [[TMP20]], ptr [[S_ARR5]], ptr [[TMP21]], i32 [[TMP23]])
1941 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1942 // CHECK10:       omp.inner.for.inc:
1943 // CHECK10-NEXT:    [[TMP24:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
1944 // CHECK10-NEXT:    [[TMP25:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
1945 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP24]], [[TMP25]]
1946 // CHECK10-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
1947 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
1948 // CHECK10:       omp.inner.for.end:
1949 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1950 // CHECK10:       omp.loop.exit:
1951 // CHECK10-NEXT:    [[TMP26:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
1952 // CHECK10-NEXT:    [[TMP27:%.*]] = load i32, ptr [[TMP26]], align 4
1953 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP27]])
1954 // CHECK10-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
1955 // CHECK10-NEXT:    [[ARRAY_BEGIN11:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR5]], i32 0, i32 0
1956 // CHECK10-NEXT:    [[TMP28:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN11]], i32 2
1957 // CHECK10-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
1958 // CHECK10:       arraydestroy.body:
1959 // CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP28]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
1960 // CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
1961 // CHECK10-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
1962 // CHECK10-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN11]]
1963 // CHECK10-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE12:%.*]], label [[ARRAYDESTROY_BODY]]
1964 // CHECK10:       arraydestroy.done12:
1965 // CHECK10-NEXT:    ret void
1966 //
1967 //
1968 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l139.omp_outlined.omp_outlined
1969 // CHECK10-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]], i32 noundef [[SVAR:%.*]]) #[[ATTR3]] {
1970 // CHECK10-NEXT:  entry:
1971 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
1972 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
1973 // CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
1974 // CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
1975 // CHECK10-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
1976 // CHECK10-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
1977 // CHECK10-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
1978 // CHECK10-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
1979 // CHECK10-NEXT:    [[SVAR_ADDR:%.*]] = alloca i32, align 4
1980 // CHECK10-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
1981 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1982 // CHECK10-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1983 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1984 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1985 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1986 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1987 // CHECK10-NEXT:    [[VEC2:%.*]] = alloca [2 x i32], align 4
1988 // CHECK10-NEXT:    [[S_ARR3:%.*]] = alloca [2 x %struct.S], align 4
1989 // CHECK10-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S:%.*]], align 4
1990 // CHECK10-NEXT:    [[_TMP6:%.*]] = alloca ptr, align 4
1991 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
1992 // CHECK10-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
1993 // CHECK10-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
1994 // CHECK10-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
1995 // CHECK10-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
1996 // CHECK10-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
1997 // CHECK10-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
1998 // CHECK10-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
1999 // CHECK10-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
2000 // CHECK10-NEXT:    store i32 [[SVAR]], ptr [[SVAR_ADDR]], align 4
2001 // CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
2002 // CHECK10-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
2003 // CHECK10-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
2004 // CHECK10-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 4
2005 // CHECK10-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2006 // CHECK10-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
2007 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
2008 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
2009 // CHECK10-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_LB]], align 4
2010 // CHECK10-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
2011 // CHECK10-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2012 // CHECK10-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2013 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC2]], ptr align 4 [[TMP0]], i32 8, i1 false)
2014 // CHECK10-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR3]], i32 0, i32 0
2015 // CHECK10-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN]], i32 2
2016 // CHECK10-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP5]]
2017 // CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
2018 // CHECK10:       omp.arraycpy.body:
2019 // CHECK10-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2020 // CHECK10-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2021 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr align 4 [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 4, i1 false)
2022 // CHECK10-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
2023 // CHECK10-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
2024 // CHECK10-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
2025 // CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
2026 // CHECK10:       omp.arraycpy.done4:
2027 // CHECK10-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[TMP]], align 4
2028 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VAR5]], ptr align 4 [[TMP6]], i32 4, i1 false)
2029 // CHECK10-NEXT:    store ptr [[VAR5]], ptr [[_TMP6]], align 4
2030 // CHECK10-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2031 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
2032 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2:[0-9]+]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2033 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2034 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1
2035 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2036 // CHECK10:       cond.true:
2037 // CHECK10-NEXT:    br label [[COND_END:%.*]]
2038 // CHECK10:       cond.false:
2039 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2040 // CHECK10-NEXT:    br label [[COND_END]]
2041 // CHECK10:       cond.end:
2042 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
2043 // CHECK10-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2044 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2045 // CHECK10-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
2046 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2047 // CHECK10:       omp.inner.for.cond:
2048 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2049 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2050 // CHECK10-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
2051 // CHECK10-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
2052 // CHECK10:       omp.inner.for.cond.cleanup:
2053 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
2054 // CHECK10:       omp.inner.for.body:
2055 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2056 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
2057 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2058 // CHECK10-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2059 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
2060 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4
2061 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC2]], i32 0, i32 [[TMP16]]
2062 // CHECK10-NEXT:    store i32 [[TMP15]], ptr [[ARRAYIDX]], align 4
2063 // CHECK10-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[_TMP6]], align 4
2064 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I]], align 4
2065 // CHECK10-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR3]], i32 0, i32 [[TMP18]]
2066 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX8]], ptr align 4 [[TMP17]], i32 4, i1 false)
2067 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2068 // CHECK10:       omp.body.continue:
2069 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2070 // CHECK10:       omp.inner.for.inc:
2071 // CHECK10-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2072 // CHECK10-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], 1
2073 // CHECK10-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
2074 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
2075 // CHECK10:       omp.inner.for.end:
2076 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2077 // CHECK10:       omp.loop.exit:
2078 // CHECK10-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2079 // CHECK10-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
2080 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
2081 // CHECK10-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR4]]
2082 // CHECK10-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S], ptr [[S_ARR3]], i32 0, i32 0
2083 // CHECK10-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAY_BEGIN10]], i32 2
2084 // CHECK10-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
2085 // CHECK10:       arraydestroy.body:
2086 // CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP22]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
2087 // CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
2088 // CHECK10-NEXT:    call void @_ZN1SIfED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
2089 // CHECK10-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
2090 // CHECK10-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
2091 // CHECK10:       arraydestroy.done11:
2092 // CHECK10-NEXT:    ret void
2093 //
2094 //
2095 // CHECK10-LABEL: define {{[^@]+}}@_ZN1SIfED1Ev
2096 // CHECK10-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
2097 // CHECK10-NEXT:  entry:
2098 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2099 // CHECK10-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2100 // CHECK10-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2101 // CHECK10-NEXT:    call void @_ZN1SIfED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
2102 // CHECK10-NEXT:    ret void
2103 //
2104 //
2105 // CHECK10-LABEL: define {{[^@]+}}@_Z5tmainIiET_v
2106 // CHECK10-SAME: () #[[ATTR1]] comdat {
2107 // CHECK10-NEXT:  entry:
2108 // CHECK10-NEXT:    [[RETVAL:%.*]] = alloca i32, align 4
2109 // CHECK10-NEXT:    [[TEST:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
2110 // CHECK10-NEXT:    [[T_VAR:%.*]] = alloca i32, align 4
2111 // CHECK10-NEXT:    [[VEC:%.*]] = alloca [2 x i32], align 4
2112 // CHECK10-NEXT:    [[S_ARR:%.*]] = alloca [2 x %struct.S.0], align 4
2113 // CHECK10-NEXT:    [[VAR:%.*]] = alloca ptr, align 4
2114 // CHECK10-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
2115 // CHECK10-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
2116 // CHECK10-NEXT:    [[DOTOFFLOAD_BASEPTRS:%.*]] = alloca [4 x ptr], align 4
2117 // CHECK10-NEXT:    [[DOTOFFLOAD_PTRS:%.*]] = alloca [4 x ptr], align 4
2118 // CHECK10-NEXT:    [[DOTOFFLOAD_MAPPERS:%.*]] = alloca [4 x ptr], align 4
2119 // CHECK10-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
2120 // CHECK10-NEXT:    [[KERNEL_ARGS:%.*]] = alloca [[STRUCT___TGT_KERNEL_ARGUMENTS:%.*]], align 8
2121 // CHECK10-NEXT:    call void @_ZN1SIiEC1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]])
2122 // CHECK10-NEXT:    store i32 0, ptr [[T_VAR]], align 4
2123 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC]], ptr align 4 @__const._Z5tmainIiET_v.vec, i32 8, i1 false)
2124 // CHECK10-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[S_ARR]], i32 noundef 1)
2125 // CHECK10-NEXT:    [[ARRAYINIT_ELEMENT:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[S_ARR]], i32 1
2126 // CHECK10-NEXT:    call void @_ZN1SIiEC1Ei(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYINIT_ELEMENT]], i32 noundef 2)
2127 // CHECK10-NEXT:    store ptr [[TEST]], ptr [[VAR]], align 4
2128 // CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VAR]], align 4
2129 // CHECK10-NEXT:    store ptr [[TMP0]], ptr [[TMP]], align 4
2130 // CHECK10-NEXT:    [[TMP1:%.*]] = load i32, ptr [[T_VAR]], align 4
2131 // CHECK10-NEXT:    store i32 [[TMP1]], ptr [[T_VAR_CASTED]], align 4
2132 // CHECK10-NEXT:    [[TMP2:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
2133 // CHECK10-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 4
2134 // CHECK10-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[TMP]], align 4
2135 // CHECK10-NEXT:    [[TMP5:%.*]] = load ptr, ptr [[TMP]], align 4
2136 // CHECK10-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2137 // CHECK10-NEXT:    store i32 [[TMP2]], ptr [[TMP6]], align 4
2138 // CHECK10-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2139 // CHECK10-NEXT:    store i32 [[TMP2]], ptr [[TMP7]], align 4
2140 // CHECK10-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 0
2141 // CHECK10-NEXT:    store ptr null, ptr [[TMP8]], align 4
2142 // CHECK10-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 1
2143 // CHECK10-NEXT:    store ptr [[VEC]], ptr [[TMP9]], align 4
2144 // CHECK10-NEXT:    [[TMP10:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 1
2145 // CHECK10-NEXT:    store ptr [[VEC]], ptr [[TMP10]], align 4
2146 // CHECK10-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 1
2147 // CHECK10-NEXT:    store ptr null, ptr [[TMP11]], align 4
2148 // CHECK10-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 2
2149 // CHECK10-NEXT:    store ptr [[S_ARR]], ptr [[TMP12]], align 4
2150 // CHECK10-NEXT:    [[TMP13:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 2
2151 // CHECK10-NEXT:    store ptr [[S_ARR]], ptr [[TMP13]], align 4
2152 // CHECK10-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 2
2153 // CHECK10-NEXT:    store ptr null, ptr [[TMP14]], align 4
2154 // CHECK10-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 3
2155 // CHECK10-NEXT:    store ptr [[TMP4]], ptr [[TMP15]], align 4
2156 // CHECK10-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 3
2157 // CHECK10-NEXT:    store ptr [[TMP5]], ptr [[TMP16]], align 4
2158 // CHECK10-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_MAPPERS]], i32 0, i32 3
2159 // CHECK10-NEXT:    store ptr null, ptr [[TMP17]], align 4
2160 // CHECK10-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_BASEPTRS]], i32 0, i32 0
2161 // CHECK10-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [4 x ptr], ptr [[DOTOFFLOAD_PTRS]], i32 0, i32 0
2162 // CHECK10-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 0
2163 // CHECK10-NEXT:    store i32 3, ptr [[TMP20]], align 4
2164 // CHECK10-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 1
2165 // CHECK10-NEXT:    store i32 4, ptr [[TMP21]], align 4
2166 // CHECK10-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 2
2167 // CHECK10-NEXT:    store ptr [[TMP18]], ptr [[TMP22]], align 4
2168 // CHECK10-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 3
2169 // CHECK10-NEXT:    store ptr [[TMP19]], ptr [[TMP23]], align 4
2170 // CHECK10-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 4
2171 // CHECK10-NEXT:    store ptr @.offload_sizes.1, ptr [[TMP24]], align 4
2172 // CHECK10-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 5
2173 // CHECK10-NEXT:    store ptr @.offload_maptypes.2, ptr [[TMP25]], align 4
2174 // CHECK10-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 6
2175 // CHECK10-NEXT:    store ptr null, ptr [[TMP26]], align 4
2176 // CHECK10-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 7
2177 // CHECK10-NEXT:    store ptr null, ptr [[TMP27]], align 4
2178 // CHECK10-NEXT:    [[TMP28:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 8
2179 // CHECK10-NEXT:    store i64 2, ptr [[TMP28]], align 8
2180 // CHECK10-NEXT:    [[TMP29:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 9
2181 // CHECK10-NEXT:    store i64 0, ptr [[TMP29]], align 8
2182 // CHECK10-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 10
2183 // CHECK10-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP30]], align 4
2184 // CHECK10-NEXT:    [[TMP31:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 11
2185 // CHECK10-NEXT:    store [3 x i32] zeroinitializer, ptr [[TMP31]], align 4
2186 // CHECK10-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw [[STRUCT___TGT_KERNEL_ARGUMENTS]], ptr [[KERNEL_ARGS]], i32 0, i32 12
2187 // CHECK10-NEXT:    store i32 0, ptr [[TMP32]], align 4
2188 // CHECK10-NEXT:    [[TMP33:%.*]] = call i32 @__tgt_target_kernel(ptr @[[GLOB3]], i64 -1, i32 0, i32 0, ptr @.{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.region_id, ptr [[KERNEL_ARGS]])
2189 // CHECK10-NEXT:    [[TMP34:%.*]] = icmp ne i32 [[TMP33]], 0
2190 // CHECK10-NEXT:    br i1 [[TMP34]], label [[OMP_OFFLOAD_FAILED:%.*]], label [[OMP_OFFLOAD_CONT:%.*]]
2191 // CHECK10:       omp_offload.failed:
2192 // CHECK10-NEXT:    call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48(i32 [[TMP2]], ptr [[VEC]], ptr [[S_ARR]], ptr [[TMP3]]) #[[ATTR4]]
2193 // CHECK10-NEXT:    br label [[OMP_OFFLOAD_CONT]]
2194 // CHECK10:       omp_offload.cont:
2195 // CHECK10-NEXT:    store i32 0, ptr [[RETVAL]], align 4
2196 // CHECK10-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR]], i32 0, i32 0
2197 // CHECK10-NEXT:    [[TMP35:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
2198 // CHECK10-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
2199 // CHECK10:       arraydestroy.body:
2200 // CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP35]], [[OMP_OFFLOAD_CONT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
2201 // CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
2202 // CHECK10-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
2203 // CHECK10-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN]]
2204 // CHECK10-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE2:%.*]], label [[ARRAYDESTROY_BODY]]
2205 // CHECK10:       arraydestroy.done2:
2206 // CHECK10-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[TEST]]) #[[ATTR4]]
2207 // CHECK10-NEXT:    [[TMP36:%.*]] = load i32, ptr [[RETVAL]], align 4
2208 // CHECK10-NEXT:    ret i32 [[TMP36]]
2209 //
2210 //
2211 // CHECK10-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ev
2212 // CHECK10-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
2213 // CHECK10-NEXT:  entry:
2214 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2215 // CHECK10-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2216 // CHECK10-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2217 // CHECK10-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
2218 // CHECK10-NEXT:    store float 0.000000e+00, ptr [[F]], align 4
2219 // CHECK10-NEXT:    ret void
2220 //
2221 //
2222 // CHECK10-LABEL: define {{[^@]+}}@_ZN1SIfEC2Ef
2223 // CHECK10-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], float noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
2224 // CHECK10-NEXT:  entry:
2225 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2226 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca float, align 4
2227 // CHECK10-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2228 // CHECK10-NEXT:    store float [[A]], ptr [[A_ADDR]], align 4
2229 // CHECK10-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2230 // CHECK10-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S:%.*]], ptr [[THIS1]], i32 0, i32 0
2231 // CHECK10-NEXT:    [[TMP0:%.*]] = load float, ptr [[A_ADDR]], align 4
2232 // CHECK10-NEXT:    store float [[TMP0]], ptr [[F]], align 4
2233 // CHECK10-NEXT:    ret void
2234 //
2235 //
2236 // CHECK10-LABEL: define {{[^@]+}}@_ZN1SIfED2Ev
2237 // CHECK10-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
2238 // CHECK10-NEXT:  entry:
2239 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2240 // CHECK10-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2241 // CHECK10-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2242 // CHECK10-NEXT:    ret void
2243 //
2244 //
2245 // CHECK10-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ev
2246 // CHECK10-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
2247 // CHECK10-NEXT:  entry:
2248 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2249 // CHECK10-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2250 // CHECK10-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2251 // CHECK10-NEXT:    call void @_ZN1SIiEC2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]])
2252 // CHECK10-NEXT:    ret void
2253 //
2254 //
2255 // CHECK10-LABEL: define {{[^@]+}}@_ZN1SIiEC1Ei
2256 // CHECK10-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
2257 // CHECK10-NEXT:  entry:
2258 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2259 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2260 // CHECK10-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2261 // CHECK10-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
2262 // CHECK10-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2263 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
2264 // CHECK10-NEXT:    call void @_ZN1SIiEC2Ei(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]], i32 noundef [[TMP0]])
2265 // CHECK10-NEXT:    ret void
2266 //
2267 //
2268 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48
2269 // CHECK10-SAME: (i32 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
2270 // CHECK10-NEXT:  entry:
2271 // CHECK10-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
2272 // CHECK10-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
2273 // CHECK10-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
2274 // CHECK10-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
2275 // CHECK10-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
2276 // CHECK10-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
2277 // CHECK10-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
2278 // CHECK10-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
2279 // CHECK10-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
2280 // CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
2281 // CHECK10-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
2282 // CHECK10-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
2283 // CHECK10-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 4
2284 // CHECK10-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[TMP]], align 4
2285 // CHECK10-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_teams(ptr @[[GLOB3]], i32 4, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.omp_outlined, ptr [[T_VAR_ADDR]], ptr [[TMP0]], ptr [[TMP1]], ptr [[TMP3]])
2286 // CHECK10-NEXT:    ret void
2287 //
2288 //
2289 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.omp_outlined
2290 // CHECK10-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
2291 // CHECK10-NEXT:  entry:
2292 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2293 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2294 // CHECK10-NEXT:    [[T_VAR_ADDR:%.*]] = alloca ptr, align 4
2295 // CHECK10-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
2296 // CHECK10-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
2297 // CHECK10-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
2298 // CHECK10-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
2299 // CHECK10-NEXT:    [[_TMP1:%.*]] = alloca ptr, align 4
2300 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2301 // CHECK10-NEXT:    [[_TMP2:%.*]] = alloca i32, align 4
2302 // CHECK10-NEXT:    [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4
2303 // CHECK10-NEXT:    [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4
2304 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2305 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2306 // CHECK10-NEXT:    [[T_VAR3:%.*]] = alloca i32, align 4
2307 // CHECK10-NEXT:    [[VEC4:%.*]] = alloca [2 x i32], align 4
2308 // CHECK10-NEXT:    [[S_ARR5:%.*]] = alloca [2 x %struct.S.0], align 4
2309 // CHECK10-NEXT:    [[VAR7:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
2310 // CHECK10-NEXT:    [[_TMP8:%.*]] = alloca ptr, align 4
2311 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
2312 // CHECK10-NEXT:    [[T_VAR_CASTED:%.*]] = alloca i32, align 4
2313 // CHECK10-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2314 // CHECK10-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2315 // CHECK10-NEXT:    store ptr [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
2316 // CHECK10-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
2317 // CHECK10-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
2318 // CHECK10-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
2319 // CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[T_VAR_ADDR]], align 4
2320 // CHECK10-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
2321 // CHECK10-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
2322 // CHECK10-NEXT:    [[TMP3:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
2323 // CHECK10-NEXT:    store ptr [[TMP3]], ptr [[TMP]], align 4
2324 // CHECK10-NEXT:    [[TMP4:%.*]] = load ptr, ptr [[TMP]], align 4
2325 // CHECK10-NEXT:    store ptr [[TMP4]], ptr [[_TMP1]], align 4
2326 // CHECK10-NEXT:    store i32 0, ptr [[DOTOMP_COMB_LB]], align 4
2327 // CHECK10-NEXT:    store i32 1, ptr [[DOTOMP_COMB_UB]], align 4
2328 // CHECK10-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2329 // CHECK10-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2330 // CHECK10-NEXT:    [[TMP5:%.*]] = load i32, ptr [[TMP0]], align 4
2331 // CHECK10-NEXT:    store i32 [[TMP5]], ptr [[T_VAR3]], align 4
2332 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC4]], ptr align 4 [[TMP1]], i32 8, i1 false)
2333 // CHECK10-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR5]], i32 0, i32 0
2334 // CHECK10-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
2335 // CHECK10-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP6]]
2336 // CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE6:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
2337 // CHECK10:       omp.arraycpy.body:
2338 // CHECK10-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP2]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2339 // CHECK10-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2340 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr align 4 [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 4, i1 false)
2341 // CHECK10-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
2342 // CHECK10-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
2343 // CHECK10-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP6]]
2344 // CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE6]], label [[OMP_ARRAYCPY_BODY]]
2345 // CHECK10:       omp.arraycpy.done6:
2346 // CHECK10-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[_TMP1]], align 4
2347 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VAR7]], ptr align 4 [[TMP7]], i32 4, i1 false)
2348 // CHECK10-NEXT:    store ptr [[VAR7]], ptr [[_TMP8]], align 4
2349 // CHECK10-NEXT:    [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2350 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
2351 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB1]], i32 [[TMP9]], i32 92, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_COMB_LB]], ptr [[DOTOMP_COMB_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2352 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2353 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP10]], 1
2354 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2355 // CHECK10:       cond.true:
2356 // CHECK10-NEXT:    br label [[COND_END:%.*]]
2357 // CHECK10:       cond.false:
2358 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2359 // CHECK10-NEXT:    br label [[COND_END]]
2360 // CHECK10:       cond.end:
2361 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
2362 // CHECK10-NEXT:    store i32 [[COND]], ptr [[DOTOMP_COMB_UB]], align 4
2363 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2364 // CHECK10-NEXT:    store i32 [[TMP12]], ptr [[DOTOMP_IV]], align 4
2365 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2366 // CHECK10:       omp.inner.for.cond:
2367 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2368 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2369 // CHECK10-NEXT:    [[CMP9:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
2370 // CHECK10-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
2371 // CHECK10:       omp.inner.for.cond.cleanup:
2372 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
2373 // CHECK10:       omp.inner.for.body:
2374 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, ptr [[DOTOMP_COMB_LB]], align 4
2375 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_UB]], align 4
2376 // CHECK10-NEXT:    [[TMP17:%.*]] = load i32, ptr [[T_VAR3]], align 4
2377 // CHECK10-NEXT:    store i32 [[TMP17]], ptr [[T_VAR_CASTED]], align 4
2378 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, ptr [[T_VAR_CASTED]], align 4
2379 // CHECK10-NEXT:    [[TMP19:%.*]] = load ptr, ptr [[_TMP8]], align 4
2380 // CHECK10-NEXT:    call void (ptr, i32, ptr, ...) @__kmpc_fork_call(ptr @[[GLOB3]], i32 6, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.omp_outlined.omp_outlined, i32 [[TMP15]], i32 [[TMP16]], ptr [[VEC4]], i32 [[TMP18]], ptr [[S_ARR5]], ptr [[TMP19]])
2381 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2382 // CHECK10:       omp.inner.for.inc:
2383 // CHECK10-NEXT:    [[TMP20:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2384 // CHECK10-NEXT:    [[TMP21:%.*]] = load i32, ptr [[DOTOMP_STRIDE]], align 4
2385 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP20]], [[TMP21]]
2386 // CHECK10-NEXT:    store i32 [[ADD]], ptr [[DOTOMP_IV]], align 4
2387 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
2388 // CHECK10:       omp.inner.for.end:
2389 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2390 // CHECK10:       omp.loop.exit:
2391 // CHECK10-NEXT:    [[TMP22:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2392 // CHECK10-NEXT:    [[TMP23:%.*]] = load i32, ptr [[TMP22]], align 4
2393 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[TMP23]])
2394 // CHECK10-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR7]]) #[[ATTR4]]
2395 // CHECK10-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR5]], i32 0, i32 0
2396 // CHECK10-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN10]], i32 2
2397 // CHECK10-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
2398 // CHECK10:       arraydestroy.body:
2399 // CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP24]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
2400 // CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
2401 // CHECK10-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
2402 // CHECK10-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
2403 // CHECK10-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
2404 // CHECK10:       arraydestroy.done11:
2405 // CHECK10-NEXT:    ret void
2406 //
2407 //
2408 // CHECK10-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}__Z5tmainIiET_v_l48.omp_outlined.omp_outlined
2409 // CHECK10-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i32 noundef [[DOTPREVIOUS_LB_:%.*]], i32 noundef [[DOTPREVIOUS_UB_:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[VEC:%.*]], i32 noundef [[T_VAR:%.*]], ptr noundef nonnull align 4 dereferenceable(8) [[S_ARR:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[VAR:%.*]]) #[[ATTR3]] {
2410 // CHECK10-NEXT:  entry:
2411 // CHECK10-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 4
2412 // CHECK10-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 4
2413 // CHECK10-NEXT:    [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i32, align 4
2414 // CHECK10-NEXT:    [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i32, align 4
2415 // CHECK10-NEXT:    [[VEC_ADDR:%.*]] = alloca ptr, align 4
2416 // CHECK10-NEXT:    [[T_VAR_ADDR:%.*]] = alloca i32, align 4
2417 // CHECK10-NEXT:    [[S_ARR_ADDR:%.*]] = alloca ptr, align 4
2418 // CHECK10-NEXT:    [[VAR_ADDR:%.*]] = alloca ptr, align 4
2419 // CHECK10-NEXT:    [[TMP:%.*]] = alloca ptr, align 4
2420 // CHECK10-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2421 // CHECK10-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
2422 // CHECK10-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2423 // CHECK10-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2424 // CHECK10-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2425 // CHECK10-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2426 // CHECK10-NEXT:    [[VEC2:%.*]] = alloca [2 x i32], align 4
2427 // CHECK10-NEXT:    [[S_ARR3:%.*]] = alloca [2 x %struct.S.0], align 4
2428 // CHECK10-NEXT:    [[VAR5:%.*]] = alloca [[STRUCT_S_0:%.*]], align 4
2429 // CHECK10-NEXT:    [[_TMP6:%.*]] = alloca ptr, align 4
2430 // CHECK10-NEXT:    [[I:%.*]] = alloca i32, align 4
2431 // CHECK10-NEXT:    store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR]], align 4
2432 // CHECK10-NEXT:    store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR]], align 4
2433 // CHECK10-NEXT:    store i32 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR]], align 4
2434 // CHECK10-NEXT:    store i32 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR]], align 4
2435 // CHECK10-NEXT:    store ptr [[VEC]], ptr [[VEC_ADDR]], align 4
2436 // CHECK10-NEXT:    store i32 [[T_VAR]], ptr [[T_VAR_ADDR]], align 4
2437 // CHECK10-NEXT:    store ptr [[S_ARR]], ptr [[S_ARR_ADDR]], align 4
2438 // CHECK10-NEXT:    store ptr [[VAR]], ptr [[VAR_ADDR]], align 4
2439 // CHECK10-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[VEC_ADDR]], align 4
2440 // CHECK10-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[S_ARR_ADDR]], align 4
2441 // CHECK10-NEXT:    [[TMP2:%.*]] = load ptr, ptr [[VAR_ADDR]], align 4
2442 // CHECK10-NEXT:    store ptr [[TMP2]], ptr [[TMP]], align 4
2443 // CHECK10-NEXT:    store i32 0, ptr [[DOTOMP_LB]], align 4
2444 // CHECK10-NEXT:    store i32 1, ptr [[DOTOMP_UB]], align 4
2445 // CHECK10-NEXT:    [[TMP3:%.*]] = load i32, ptr [[DOTPREVIOUS_LB__ADDR]], align 4
2446 // CHECK10-NEXT:    [[TMP4:%.*]] = load i32, ptr [[DOTPREVIOUS_UB__ADDR]], align 4
2447 // CHECK10-NEXT:    store i32 [[TMP3]], ptr [[DOTOMP_LB]], align 4
2448 // CHECK10-NEXT:    store i32 [[TMP4]], ptr [[DOTOMP_UB]], align 4
2449 // CHECK10-NEXT:    store i32 1, ptr [[DOTOMP_STRIDE]], align 4
2450 // CHECK10-NEXT:    store i32 0, ptr [[DOTOMP_IS_LAST]], align 4
2451 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VEC2]], ptr align 4 [[TMP0]], i32 8, i1 false)
2452 // CHECK10-NEXT:    [[ARRAY_BEGIN:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR3]], i32 0, i32 0
2453 // CHECK10-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN]], i32 2
2454 // CHECK10-NEXT:    [[OMP_ARRAYCPY_ISEMPTY:%.*]] = icmp eq ptr [[ARRAY_BEGIN]], [[TMP5]]
2455 // CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_ISEMPTY]], label [[OMP_ARRAYCPY_DONE4:%.*]], label [[OMP_ARRAYCPY_BODY:%.*]]
2456 // CHECK10:       omp.arraycpy.body:
2457 // CHECK10-NEXT:    [[OMP_ARRAYCPY_SRCELEMENTPAST:%.*]] = phi ptr [ [[TMP1]], [[ENTRY:%.*]] ], [ [[OMP_ARRAYCPY_SRC_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2458 // CHECK10-NEXT:    [[OMP_ARRAYCPY_DESTELEMENTPAST:%.*]] = phi ptr [ [[ARRAY_BEGIN]], [[ENTRY]] ], [ [[OMP_ARRAYCPY_DEST_ELEMENT:%.*]], [[OMP_ARRAYCPY_BODY]] ]
2459 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[OMP_ARRAYCPY_DESTELEMENTPAST]], ptr align 4 [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 4, i1 false)
2460 // CHECK10-NEXT:    [[OMP_ARRAYCPY_DEST_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_DESTELEMENTPAST]], i32 1
2461 // CHECK10-NEXT:    [[OMP_ARRAYCPY_SRC_ELEMENT]] = getelementptr [[STRUCT_S_0]], ptr [[OMP_ARRAYCPY_SRCELEMENTPAST]], i32 1
2462 // CHECK10-NEXT:    [[OMP_ARRAYCPY_DONE:%.*]] = icmp eq ptr [[OMP_ARRAYCPY_DEST_ELEMENT]], [[TMP5]]
2463 // CHECK10-NEXT:    br i1 [[OMP_ARRAYCPY_DONE]], label [[OMP_ARRAYCPY_DONE4]], label [[OMP_ARRAYCPY_BODY]]
2464 // CHECK10:       omp.arraycpy.done4:
2465 // CHECK10-NEXT:    [[TMP6:%.*]] = load ptr, ptr [[TMP]], align 4
2466 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[VAR5]], ptr align 4 [[TMP6]], i32 4, i1 false)
2467 // CHECK10-NEXT:    store ptr [[VAR5]], ptr [[_TMP6]], align 4
2468 // CHECK10-NEXT:    [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2469 // CHECK10-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
2470 // CHECK10-NEXT:    call void @__kmpc_for_static_init_4(ptr @[[GLOB2]], i32 [[TMP8]], i32 34, ptr [[DOTOMP_IS_LAST]], ptr [[DOTOMP_LB]], ptr [[DOTOMP_UB]], ptr [[DOTOMP_STRIDE]], i32 1, i32 1)
2471 // CHECK10-NEXT:    [[TMP9:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2472 // CHECK10-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP9]], 1
2473 // CHECK10-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2474 // CHECK10:       cond.true:
2475 // CHECK10-NEXT:    br label [[COND_END:%.*]]
2476 // CHECK10:       cond.false:
2477 // CHECK10-NEXT:    [[TMP10:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2478 // CHECK10-NEXT:    br label [[COND_END]]
2479 // CHECK10:       cond.end:
2480 // CHECK10-NEXT:    [[COND:%.*]] = phi i32 [ 1, [[COND_TRUE]] ], [ [[TMP10]], [[COND_FALSE]] ]
2481 // CHECK10-NEXT:    store i32 [[COND]], ptr [[DOTOMP_UB]], align 4
2482 // CHECK10-NEXT:    [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB]], align 4
2483 // CHECK10-NEXT:    store i32 [[TMP11]], ptr [[DOTOMP_IV]], align 4
2484 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2485 // CHECK10:       omp.inner.for.cond:
2486 // CHECK10-NEXT:    [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2487 // CHECK10-NEXT:    [[TMP13:%.*]] = load i32, ptr [[DOTOMP_UB]], align 4
2488 // CHECK10-NEXT:    [[CMP7:%.*]] = icmp sle i32 [[TMP12]], [[TMP13]]
2489 // CHECK10-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
2490 // CHECK10:       omp.inner.for.cond.cleanup:
2491 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
2492 // CHECK10:       omp.inner.for.body:
2493 // CHECK10-NEXT:    [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2494 // CHECK10-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
2495 // CHECK10-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
2496 // CHECK10-NEXT:    store i32 [[ADD]], ptr [[I]], align 4
2497 // CHECK10-NEXT:    [[TMP15:%.*]] = load i32, ptr [[T_VAR_ADDR]], align 4
2498 // CHECK10-NEXT:    [[TMP16:%.*]] = load i32, ptr [[I]], align 4
2499 // CHECK10-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[VEC2]], i32 0, i32 [[TMP16]]
2500 // CHECK10-NEXT:    store i32 [[TMP15]], ptr [[ARRAYIDX]], align 4
2501 // CHECK10-NEXT:    [[TMP17:%.*]] = load ptr, ptr [[_TMP6]], align 4
2502 // CHECK10-NEXT:    [[TMP18:%.*]] = load i32, ptr [[I]], align 4
2503 // CHECK10-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR3]], i32 0, i32 [[TMP18]]
2504 // CHECK10-NEXT:    call void @llvm.memcpy.p0.p0.i32(ptr align 4 [[ARRAYIDX8]], ptr align 4 [[TMP17]], i32 4, i1 false)
2505 // CHECK10-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2506 // CHECK10:       omp.body.continue:
2507 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2508 // CHECK10:       omp.inner.for.inc:
2509 // CHECK10-NEXT:    [[TMP19:%.*]] = load i32, ptr [[DOTOMP_IV]], align 4
2510 // CHECK10-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP19]], 1
2511 // CHECK10-NEXT:    store i32 [[ADD9]], ptr [[DOTOMP_IV]], align 4
2512 // CHECK10-NEXT:    br label [[OMP_INNER_FOR_COND]]
2513 // CHECK10:       omp.inner.for.end:
2514 // CHECK10-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2515 // CHECK10:       omp.loop.exit:
2516 // CHECK10-NEXT:    [[TMP20:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR]], align 4
2517 // CHECK10-NEXT:    [[TMP21:%.*]] = load i32, ptr [[TMP20]], align 4
2518 // CHECK10-NEXT:    call void @__kmpc_for_static_fini(ptr @[[GLOB2]], i32 [[TMP21]])
2519 // CHECK10-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[VAR5]]) #[[ATTR4]]
2520 // CHECK10-NEXT:    [[ARRAY_BEGIN10:%.*]] = getelementptr inbounds [2 x %struct.S.0], ptr [[S_ARR3]], i32 0, i32 0
2521 // CHECK10-NEXT:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAY_BEGIN10]], i32 2
2522 // CHECK10-NEXT:    br label [[ARRAYDESTROY_BODY:%.*]]
2523 // CHECK10:       arraydestroy.body:
2524 // CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENTPAST:%.*]] = phi ptr [ [[TMP22]], [[OMP_LOOP_EXIT]] ], [ [[ARRAYDESTROY_ELEMENT:%.*]], [[ARRAYDESTROY_BODY]] ]
2525 // CHECK10-NEXT:    [[ARRAYDESTROY_ELEMENT]] = getelementptr inbounds [[STRUCT_S_0]], ptr [[ARRAYDESTROY_ELEMENTPAST]], i32 -1
2526 // CHECK10-NEXT:    call void @_ZN1SIiED1Ev(ptr noundef nonnull align 4 dereferenceable(4) [[ARRAYDESTROY_ELEMENT]]) #[[ATTR4]]
2527 // CHECK10-NEXT:    [[ARRAYDESTROY_DONE:%.*]] = icmp eq ptr [[ARRAYDESTROY_ELEMENT]], [[ARRAY_BEGIN10]]
2528 // CHECK10-NEXT:    br i1 [[ARRAYDESTROY_DONE]], label [[ARRAYDESTROY_DONE11:%.*]], label [[ARRAYDESTROY_BODY]]
2529 // CHECK10:       arraydestroy.done11:
2530 // CHECK10-NEXT:    ret void
2531 //
2532 //
2533 // CHECK10-LABEL: define {{[^@]+}}@_ZN1SIiED1Ev
2534 // CHECK10-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
2535 // CHECK10-NEXT:  entry:
2536 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2537 // CHECK10-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2538 // CHECK10-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2539 // CHECK10-NEXT:    call void @_ZN1SIiED2Ev(ptr noundef nonnull align 4 dereferenceable(4) [[THIS1]]) #[[ATTR4]]
2540 // CHECK10-NEXT:    ret void
2541 //
2542 //
2543 // CHECK10-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ev
2544 // CHECK10-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
2545 // CHECK10-NEXT:  entry:
2546 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2547 // CHECK10-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2548 // CHECK10-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2549 // CHECK10-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
2550 // CHECK10-NEXT:    store i32 0, ptr [[F]], align 4
2551 // CHECK10-NEXT:    ret void
2552 //
2553 //
2554 // CHECK10-LABEL: define {{[^@]+}}@_ZN1SIiEC2Ei
2555 // CHECK10-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]], i32 noundef [[A:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
2556 // CHECK10-NEXT:  entry:
2557 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2558 // CHECK10-NEXT:    [[A_ADDR:%.*]] = alloca i32, align 4
2559 // CHECK10-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2560 // CHECK10-NEXT:    store i32 [[A]], ptr [[A_ADDR]], align 4
2561 // CHECK10-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2562 // CHECK10-NEXT:    [[F:%.*]] = getelementptr inbounds nuw [[STRUCT_S_0:%.*]], ptr [[THIS1]], i32 0, i32 0
2563 // CHECK10-NEXT:    [[TMP0:%.*]] = load i32, ptr [[A_ADDR]], align 4
2564 // CHECK10-NEXT:    store i32 [[TMP0]], ptr [[F]], align 4
2565 // CHECK10-NEXT:    ret void
2566 //
2567 //
2568 // CHECK10-LABEL: define {{[^@]+}}@_ZN1SIiED2Ev
2569 // CHECK10-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[THIS:%.*]]) unnamed_addr #[[ATTR1]] comdat align 2 {
2570 // CHECK10-NEXT:  entry:
2571 // CHECK10-NEXT:    [[THIS_ADDR:%.*]] = alloca ptr, align 4
2572 // CHECK10-NEXT:    store ptr [[THIS]], ptr [[THIS_ADDR]], align 4
2573 // CHECK10-NEXT:    [[THIS1:%.*]] = load ptr, ptr [[THIS_ADDR]], align 4
2574 // CHECK10-NEXT:    ret void
2575 //
2576