1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
2 // RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK1
3 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK2
4 
5 // RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
6 // RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK3
7 
8 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
9 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK4
10 
11 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp -fexceptions -fcxx-exceptions -debug-info-kind=line-tables-only -gno-column-info -x c++ -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK5
12 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK6
13 
14 // RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
15 // RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
16 // RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
17 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp-simd -fexceptions -fcxx-exceptions -debug-info-kind=line-tables-only -x c++ -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
18 // RUN: %clang_cc1 -verify -triple x86_64-apple-darwin10 -fopenmp-simd -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
19 // expected-no-diagnostics
20 
21 // RUN: %clang_cc1 -verify -fopenmp -DOMP5 -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --check-prefix=CHECK11
22 // RUN: %clang_cc1 -fopenmp -DOMP5 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
23 // RUN: %clang_cc1 -fopenmp -DOMP5 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK12
24 // RUN: %clang_cc1 -verify -fopenmp-simd -DOMP5 -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -fexceptions -fcxx-exceptions -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
25 // RUN: %clang_cc1 -fopenmp-simd -DOMP5 -x c++ -std=c++11 -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -emit-pch -o %t %s
26 // RUN: %clang_cc1 -fopenmp-simd -DOMP5 -x c++ -triple x86_64-unknown-unknown -fexceptions -fcxx-exceptions -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
27 #ifndef HEADER
28 #define HEADER
29 
30 #ifndef OMP5
31 
32 void with_var_schedule() {
33   double a = 5;
34 
35 #pragma omp parallel for schedule(static, char(a)) private(a)
36   for (unsigned long long i = 1; i < 2 + a; ++i) {
37   }
38 }
39 
40 void without_schedule_clause(float *a, float *b, float *c, float *d) {
41   #pragma omp parallel for
42 // UB = min(UB, GlobalUB)
43 // Loop header
44   for (int i = 33; i < 32000000; i += 7) {
45 // Start of body: calculate i from IV:
46 // ... loop body ...
47 // End of body: store into a[i]:
48     a[i] = b[i] * c[i] * d[i];
49   }
50 }
51 
52 void static_not_chunked(float *a, float *b, float *c, float *d) {
53   #pragma omp parallel for schedule(static)
54 // UB = min(UB, GlobalUB)
55 // Loop header
56   for (int i = 32000000; i > 33; i += -7) {
57 // Start of body: calculate i from IV:
58 // ... loop body ...
59 // End of body: store into a[i]:
60     a[i] = b[i] * c[i] * d[i];
61   }
62 }
63 
64 void static_chunked(float *a, float *b, float *c, float *d) {
65   #pragma omp parallel for schedule(static, 5)
66 // UB = min(UB, GlobalUB)
67 
68 // Outer loop header
69 
70 // Loop header
71   for (unsigned i = 131071; i <= 2147483647; i += 127) {
72 // Start of body: calculate i from IV:
73 // ... loop body ...
74 // End of body: store into a[i]:
75     a[i] = b[i] * c[i] * d[i];
76   }
77 // Update the counters, adding stride
78 
79 }
80 
81 void dynamic1(float *a, float *b, float *c, float *d) {
82   #pragma omp parallel for schedule(dynamic)
83 
84 // Loop header
85 
86   for (unsigned long long i = 131071; i < 2147483647; i += 127) {
87 // Start of body: calculate i from IV:
88 // ... loop body ...
89 // End of body: store into a[i]:
90     a[i] = b[i] * c[i] * d[i];
91   }
92 }
93 
94 void guided7(float *a, float *b, float *c, float *d) {
95   #pragma omp parallel for schedule(guided, 7)
96 
97 // Loop header
98 
99   for (unsigned long long i = 131071; i < 2147483647; i += 127) {
100 // Start of body: calculate i from IV:
101 // ... loop body ...
102 // End of body: store into a[i]:
103     a[i] = b[i] * c[i] * d[i];
104   }
105 }
106 
107 void test_auto(float *a, float *b, float *c, float *d) {
108   unsigned int x = 0;
109   unsigned int y = 0;
110   #pragma omp parallel for schedule(auto) collapse(2)
111 
112 // Loop header
113 
114 // FIXME: When the iteration count of some nested loop is not a known constant,
115 // we should pre-calculate it, like we do for the total number of iterations!
116   for (char i = static_cast<char>(y); i <= '9'; ++i)
117     for (x = 11; x > 0; --x) {
118 // Start of body: indices are calculated from IV:
119 // ... loop body ...
120 // End of body: store into a[i]:
121     a[i] = b[i] * c[i] * d[i];
122   }
123 }
124 
125 void runtime(float *a, float *b, float *c, float *d) {
126   int x = 0;
127   #pragma omp parallel for collapse(2) schedule(runtime)
128 
129 // Loop header
130 
131   for (unsigned char i = '0' ; i <= '9'; ++i)
132     for (x = -10; x < 10; ++x) {
133 // Start of body: indices are calculated from IV:
134 // ... loop body ...
135 // End of body: store into a[i]:
136     a[i] = b[i] * c[i] * d[i];
137   }
138 }
139 
140 int foo() { extern void mayThrow(); mayThrow(); return 0; };
141 
142 void parallel_for(float *a, const int n) {
143   float arr[n];
144 #pragma omp parallel for schedule(static, 5) private(arr) default(none) firstprivate(n) shared(a)
145   for (unsigned i = 131071; i <= 2147483647; i += 127)
146     a[i] += foo() + arr[i] + n;
147 }
148 // Check source line corresponds to "#pragma omp parallel for schedule(static, 5)" above:
149 
150 #else // OMP5
151 int increment () {
152   #pragma omp for
153 // Determine UB = min(UB, GlobalUB)
154 
155 // Loop header
156 
157   for (int i = 0 ; i != 5; ++i)
158 // Start of body: calculate i from IV:
159     ;
160   return 0;
161 }
162 
163 int decrement_nowait () {
164   #pragma omp for nowait
165 // Determine UB = min(UB, GlobalUB)
166 
167 // Loop header
168   for (int j = 5 ; j != 0; --j)
169 // Start of body: calculate i from IV:
170     ;
171   return 0;
172 }
173 
174 void range_for_single() {
175   int arr[10] = {0};
176 #pragma omp parallel for
177   for (auto &a : arr)
178     (void)a;
179 }
180 
181 
182 // __range = arr;
183 
184 // __end = end(_range);
185 
186 
187 // calculate number of elements.
188 
189 // __begin = begin(range);
190 
191 // __begin >= __end ? goto then : goto exit;
192 
193 
194 // lb = 0;
195 
196 // ub = number of elements
197 
198 // stride = 1;
199 
200 // is_last = 0;
201 
202 // loop.
203 
204 // ub = (ub > number_of_elems ? number_of_elems : ub);
205 
206 
207 
208 // OMP%: store i64 [[MIN]], i64* [[UB]],
209 
210 // iv = lb;
211 
212 // goto loop;
213 // loop:
214 
215 
216 // iv <= ub ? goto body : goto end;
217 
218 // body:
219 // __begin = begin(arr) + iv * 1;
220 
221 // a = *__begin;
222 
223 // (void)a;
224 
225 // iv += 1;
226 
227 // goto loop;
228 
229 // end:
230 // exit:
231 
232 void range_for_collapsed() {
233   int arr[10] = {0};
234 #pragma omp parallel for collapse(2)
235   for (auto &a : arr)
236     for (auto b : arr)
237       a = b;
238 }
239 #endif // OMP5
240 
241 #endif // HEADER
242 
243 // CHECK1-LABEL: define {{[^@]+}}@_Z17with_var_schedulev
244 // CHECK1-SAME: () #[[ATTR0:[0-9]+]] {
245 // CHECK1-NEXT:  entry:
246 // CHECK1-NEXT:    [[A:%.*]] = alloca double, align 8
247 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
248 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
249 // CHECK1-NEXT:    store double 5.000000e+00, double* [[A]], align 8
250 // CHECK1-NEXT:    [[TMP0:%.*]] = load double, double* [[A]], align 8
251 // CHECK1-NEXT:    [[CONV:%.*]] = fptosi double [[TMP0]] to i8
252 // CHECK1-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
253 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
254 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
255 // CHECK1-NEXT:    store i8 [[TMP2]], i8* [[TMP1]], align 1
256 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
257 // CHECK1-NEXT:    ret void
258 //
259 //
260 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
261 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
262 // CHECK1-NEXT:  entry:
263 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
264 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
265 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
266 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
267 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
268 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i64, align 8
269 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca double, align 8
270 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
271 // CHECK1-NEXT:    [[I:%.*]] = alloca i64, align 8
272 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
273 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
274 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
275 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
276 // CHECK1-NEXT:    [[A:%.*]] = alloca double, align 8
277 // CHECK1-NEXT:    [[I4:%.*]] = alloca i64, align 8
278 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
279 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
280 // CHECK1-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
281 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
282 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
283 // CHECK1-NEXT:    [[TMP2:%.*]] = load i8, i8* [[TMP1]], align 1
284 // CHECK1-NEXT:    store i8 [[TMP2]], i8* [[DOTCAPTURE_EXPR_]], align 1
285 // CHECK1-NEXT:    [[TMP3:%.*]] = load double, double* undef, align 8
286 // CHECK1-NEXT:    [[ADD:%.*]] = fadd double 2.000000e+00, [[TMP3]]
287 // CHECK1-NEXT:    store double [[ADD]], double* [[DOTCAPTURE_EXPR_1]], align 8
288 // CHECK1-NEXT:    [[TMP4:%.*]] = load double, double* [[DOTCAPTURE_EXPR_1]], align 8
289 // CHECK1-NEXT:    [[SUB:%.*]] = fsub double [[TMP4]], 1.000000e+00
290 // CHECK1-NEXT:    [[DIV:%.*]] = fdiv double [[SUB]], 1.000000e+00
291 // CHECK1-NEXT:    [[CONV:%.*]] = fptoui double [[DIV]] to i64
292 // CHECK1-NEXT:    [[SUB3:%.*]] = sub i64 [[CONV]], 1
293 // CHECK1-NEXT:    store i64 [[SUB3]], i64* [[DOTCAPTURE_EXPR_2]], align 8
294 // CHECK1-NEXT:    store i64 1, i64* [[I]], align 8
295 // CHECK1-NEXT:    [[TMP5:%.*]] = load double, double* [[DOTCAPTURE_EXPR_1]], align 8
296 // CHECK1-NEXT:    [[CMP:%.*]] = fcmp olt double 1.000000e+00, [[TMP5]]
297 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
298 // CHECK1:       omp.precond.then:
299 // CHECK1-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
300 // CHECK1-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
301 // CHECK1-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_UB]], align 8
302 // CHECK1-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
303 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
304 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
305 // CHECK1-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP7]] to i64
306 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
307 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
308 // CHECK1-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 [[CONV5]])
309 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
310 // CHECK1:       omp.dispatch.cond:
311 // CHECK1-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
312 // CHECK1-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
313 // CHECK1-NEXT:    [[CMP6:%.*]] = icmp ugt i64 [[TMP10]], [[TMP11]]
314 // CHECK1-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
315 // CHECK1:       cond.true:
316 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
317 // CHECK1-NEXT:    br label [[COND_END:%.*]]
318 // CHECK1:       cond.false:
319 // CHECK1-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
320 // CHECK1-NEXT:    br label [[COND_END]]
321 // CHECK1:       cond.end:
322 // CHECK1-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
323 // CHECK1-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
324 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
325 // CHECK1-NEXT:    store i64 [[TMP14]], i64* [[DOTOMP_IV]], align 8
326 // CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
327 // CHECK1-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
328 // CHECK1-NEXT:    [[ADD7:%.*]] = add i64 [[TMP16]], 1
329 // CHECK1-NEXT:    [[CMP8:%.*]] = icmp ult i64 [[TMP15]], [[ADD7]]
330 // CHECK1-NEXT:    br i1 [[CMP8]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
331 // CHECK1:       omp.dispatch.body:
332 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
333 // CHECK1:       omp.inner.for.cond:
334 // CHECK1-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
335 // CHECK1-NEXT:    [[TMP18:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
336 // CHECK1-NEXT:    [[ADD9:%.*]] = add i64 [[TMP18]], 1
337 // CHECK1-NEXT:    [[CMP10:%.*]] = icmp ult i64 [[TMP17]], [[ADD9]]
338 // CHECK1-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
339 // CHECK1:       omp.inner.for.body:
340 // CHECK1-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
341 // CHECK1-NEXT:    [[MUL:%.*]] = mul i64 [[TMP19]], 1
342 // CHECK1-NEXT:    [[ADD11:%.*]] = add i64 1, [[MUL]]
343 // CHECK1-NEXT:    store i64 [[ADD11]], i64* [[I4]], align 8
344 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
345 // CHECK1:       omp.body.continue:
346 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
347 // CHECK1:       omp.inner.for.inc:
348 // CHECK1-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
349 // CHECK1-NEXT:    [[ADD12:%.*]] = add i64 [[TMP20]], 1
350 // CHECK1-NEXT:    store i64 [[ADD12]], i64* [[DOTOMP_IV]], align 8
351 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
352 // CHECK1:       omp.inner.for.end:
353 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
354 // CHECK1:       omp.dispatch.inc:
355 // CHECK1-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
356 // CHECK1-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
357 // CHECK1-NEXT:    [[ADD13:%.*]] = add i64 [[TMP21]], [[TMP22]]
358 // CHECK1-NEXT:    store i64 [[ADD13]], i64* [[DOTOMP_LB]], align 8
359 // CHECK1-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
360 // CHECK1-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
361 // CHECK1-NEXT:    [[ADD14:%.*]] = add i64 [[TMP23]], [[TMP24]]
362 // CHECK1-NEXT:    store i64 [[ADD14]], i64* [[DOTOMP_UB]], align 8
363 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
364 // CHECK1:       omp.dispatch.end:
365 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
366 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
367 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
368 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
369 // CHECK1:       omp.precond.end:
370 // CHECK1-NEXT:    ret void
371 //
372 //
373 // CHECK1-LABEL: define {{[^@]+}}@_Z23without_schedule_clausePfS_S_S_
374 // CHECK1-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
375 // CHECK1-NEXT:  entry:
376 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
377 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
378 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
379 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
380 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
381 // CHECK1-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
382 // CHECK1-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
383 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
384 // CHECK1-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
385 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
386 // CHECK1-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
387 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
388 // CHECK1-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
389 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
390 // CHECK1-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
391 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
392 // CHECK1-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
393 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
394 // CHECK1-NEXT:    ret void
395 //
396 //
397 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..1
398 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
399 // CHECK1-NEXT:  entry:
400 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
401 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
402 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
403 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
404 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
405 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
406 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
407 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
408 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
409 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
410 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
411 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
412 // CHECK1-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
413 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
414 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
415 // CHECK1-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
416 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1
417 // CHECK1-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
418 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 2
419 // CHECK1-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
420 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 3
421 // CHECK1-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
422 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
423 // CHECK1-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
424 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
425 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
426 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
427 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
428 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
429 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
430 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 4571423
431 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
432 // CHECK1:       cond.true:
433 // CHECK1-NEXT:    br label [[COND_END:%.*]]
434 // CHECK1:       cond.false:
435 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
436 // CHECK1-NEXT:    br label [[COND_END]]
437 // CHECK1:       cond.end:
438 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 4571423, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
439 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
440 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
441 // CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
442 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
443 // CHECK1:       omp.inner.for.cond:
444 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
445 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
446 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
447 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
448 // CHECK1:       omp.inner.for.body:
449 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
450 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 7
451 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 33, [[MUL]]
452 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
453 // CHECK1-NEXT:    [[TMP17:%.*]] = load float*, float** [[TMP4]], align 8
454 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
455 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
456 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM]]
457 // CHECK1-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
458 // CHECK1-NEXT:    [[TMP20:%.*]] = load float*, float** [[TMP6]], align 8
459 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4
460 // CHECK1-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP21]] to i64
461 // CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM2]]
462 // CHECK1-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX3]], align 4
463 // CHECK1-NEXT:    [[MUL4:%.*]] = fmul float [[TMP19]], [[TMP22]]
464 // CHECK1-NEXT:    [[TMP23:%.*]] = load float*, float** [[TMP8]], align 8
465 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4
466 // CHECK1-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP24]] to i64
467 // CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM5]]
468 // CHECK1-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX6]], align 4
469 // CHECK1-NEXT:    [[MUL7:%.*]] = fmul float [[MUL4]], [[TMP25]]
470 // CHECK1-NEXT:    [[TMP26:%.*]] = load float*, float** [[TMP2]], align 8
471 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I]], align 4
472 // CHECK1-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP27]] to i64
473 // CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[TMP26]], i64 [[IDXPROM8]]
474 // CHECK1-NEXT:    store float [[MUL7]], float* [[ARRAYIDX9]], align 4
475 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
476 // CHECK1:       omp.body.continue:
477 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
478 // CHECK1:       omp.inner.for.inc:
479 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
480 // CHECK1-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP28]], 1
481 // CHECK1-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
482 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
483 // CHECK1:       omp.inner.for.end:
484 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
485 // CHECK1:       omp.loop.exit:
486 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
487 // CHECK1-NEXT:    ret void
488 //
489 //
490 // CHECK1-LABEL: define {{[^@]+}}@_Z18static_not_chunkedPfS_S_S_
491 // CHECK1-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
492 // CHECK1-NEXT:  entry:
493 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
494 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
495 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
496 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
497 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 8
498 // CHECK1-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
499 // CHECK1-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
500 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
501 // CHECK1-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
502 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
503 // CHECK1-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
504 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
505 // CHECK1-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
506 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
507 // CHECK1-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
508 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
509 // CHECK1-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
510 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
511 // CHECK1-NEXT:    ret void
512 //
513 //
514 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..2
515 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
516 // CHECK1-NEXT:  entry:
517 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
518 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
519 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
520 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
521 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
522 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
523 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
524 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
525 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
526 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
527 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
528 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
529 // CHECK1-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
530 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
531 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
532 // CHECK1-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
533 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
534 // CHECK1-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
535 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 2
536 // CHECK1-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
537 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 3
538 // CHECK1-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
539 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
540 // CHECK1-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
541 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
542 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
543 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
544 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
545 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
546 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
547 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 4571423
548 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
549 // CHECK1:       cond.true:
550 // CHECK1-NEXT:    br label [[COND_END:%.*]]
551 // CHECK1:       cond.false:
552 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
553 // CHECK1-NEXT:    br label [[COND_END]]
554 // CHECK1:       cond.end:
555 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 4571423, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
556 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
557 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
558 // CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
559 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
560 // CHECK1:       omp.inner.for.cond:
561 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
562 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
563 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
564 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
565 // CHECK1:       omp.inner.for.body:
566 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
567 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 7
568 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
569 // CHECK1-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
570 // CHECK1-NEXT:    [[TMP17:%.*]] = load float*, float** [[TMP4]], align 8
571 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
572 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
573 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM]]
574 // CHECK1-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
575 // CHECK1-NEXT:    [[TMP20:%.*]] = load float*, float** [[TMP6]], align 8
576 // CHECK1-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4
577 // CHECK1-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP21]] to i64
578 // CHECK1-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM2]]
579 // CHECK1-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX3]], align 4
580 // CHECK1-NEXT:    [[MUL4:%.*]] = fmul float [[TMP19]], [[TMP22]]
581 // CHECK1-NEXT:    [[TMP23:%.*]] = load float*, float** [[TMP8]], align 8
582 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4
583 // CHECK1-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP24]] to i64
584 // CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM5]]
585 // CHECK1-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX6]], align 4
586 // CHECK1-NEXT:    [[MUL7:%.*]] = fmul float [[MUL4]], [[TMP25]]
587 // CHECK1-NEXT:    [[TMP26:%.*]] = load float*, float** [[TMP2]], align 8
588 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I]], align 4
589 // CHECK1-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP27]] to i64
590 // CHECK1-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[TMP26]], i64 [[IDXPROM8]]
591 // CHECK1-NEXT:    store float [[MUL7]], float* [[ARRAYIDX9]], align 4
592 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
593 // CHECK1:       omp.body.continue:
594 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
595 // CHECK1:       omp.inner.for.inc:
596 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
597 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
598 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
599 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
600 // CHECK1:       omp.inner.for.end:
601 // CHECK1-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
602 // CHECK1:       omp.loop.exit:
603 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
604 // CHECK1-NEXT:    ret void
605 //
606 //
607 // CHECK1-LABEL: define {{[^@]+}}@_Z14static_chunkedPfS_S_S_
608 // CHECK1-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
609 // CHECK1-NEXT:  entry:
610 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
611 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
612 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
613 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
614 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
615 // CHECK1-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
616 // CHECK1-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
617 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
618 // CHECK1-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
619 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
620 // CHECK1-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
621 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
622 // CHECK1-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
623 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
624 // CHECK1-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
625 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
626 // CHECK1-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
627 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
628 // CHECK1-NEXT:    ret void
629 //
630 //
631 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..3
632 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
633 // CHECK1-NEXT:  entry:
634 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
635 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
636 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
637 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
638 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
639 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
640 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
641 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
642 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
643 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
644 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
645 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
646 // CHECK1-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
647 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
648 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
649 // CHECK1-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
650 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
651 // CHECK1-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
652 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
653 // CHECK1-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
654 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
655 // CHECK1-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
656 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
657 // CHECK1-NEXT:    store i32 16908288, i32* [[DOTOMP_UB]], align 4
658 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
659 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
660 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
661 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
662 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 5)
663 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
664 // CHECK1:       omp.dispatch.cond:
665 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
666 // CHECK1-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP11]], 16908288
667 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
668 // CHECK1:       cond.true:
669 // CHECK1-NEXT:    br label [[COND_END:%.*]]
670 // CHECK1:       cond.false:
671 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
672 // CHECK1-NEXT:    br label [[COND_END]]
673 // CHECK1:       cond.end:
674 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 16908288, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
675 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
676 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
677 // CHECK1-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
678 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
679 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
680 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
681 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
682 // CHECK1:       omp.dispatch.body:
683 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
684 // CHECK1:       omp.inner.for.cond:
685 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
686 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
687 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[TMP16]], [[TMP17]]
688 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
689 // CHECK1:       omp.inner.for.body:
690 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
691 // CHECK1-NEXT:    [[MUL:%.*]] = mul i32 [[TMP18]], 127
692 // CHECK1-NEXT:    [[ADD:%.*]] = add i32 131071, [[MUL]]
693 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
694 // CHECK1-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP4]], align 8
695 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I]], align 4
696 // CHECK1-NEXT:    [[IDXPROM:%.*]] = zext i32 [[TMP20]] to i64
697 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[IDXPROM]]
698 // CHECK1-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX]], align 4
699 // CHECK1-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP6]], align 8
700 // CHECK1-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
701 // CHECK1-NEXT:    [[IDXPROM3:%.*]] = zext i32 [[TMP23]] to i64
702 // CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[IDXPROM3]]
703 // CHECK1-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4
704 // CHECK1-NEXT:    [[MUL5:%.*]] = fmul float [[TMP21]], [[TMP24]]
705 // CHECK1-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP8]], align 8
706 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[I]], align 4
707 // CHECK1-NEXT:    [[IDXPROM6:%.*]] = zext i32 [[TMP26]] to i64
708 // CHECK1-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[IDXPROM6]]
709 // CHECK1-NEXT:    [[TMP27:%.*]] = load float, float* [[ARRAYIDX7]], align 4
710 // CHECK1-NEXT:    [[MUL8:%.*]] = fmul float [[MUL5]], [[TMP27]]
711 // CHECK1-NEXT:    [[TMP28:%.*]] = load float*, float** [[TMP2]], align 8
712 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[I]], align 4
713 // CHECK1-NEXT:    [[IDXPROM9:%.*]] = zext i32 [[TMP29]] to i64
714 // CHECK1-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP28]], i64 [[IDXPROM9]]
715 // CHECK1-NEXT:    store float [[MUL8]], float* [[ARRAYIDX10]], align 4
716 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
717 // CHECK1:       omp.body.continue:
718 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
719 // CHECK1:       omp.inner.for.inc:
720 // CHECK1-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
721 // CHECK1-NEXT:    [[ADD11:%.*]] = add i32 [[TMP30]], 1
722 // CHECK1-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4
723 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
724 // CHECK1:       omp.inner.for.end:
725 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
726 // CHECK1:       omp.dispatch.inc:
727 // CHECK1-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
728 // CHECK1-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
729 // CHECK1-NEXT:    [[ADD12:%.*]] = add i32 [[TMP31]], [[TMP32]]
730 // CHECK1-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_LB]], align 4
731 // CHECK1-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
732 // CHECK1-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
733 // CHECK1-NEXT:    [[ADD13:%.*]] = add i32 [[TMP33]], [[TMP34]]
734 // CHECK1-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_UB]], align 4
735 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
736 // CHECK1:       omp.dispatch.end:
737 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
738 // CHECK1-NEXT:    ret void
739 //
740 //
741 // CHECK1-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
742 // CHECK1-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
743 // CHECK1-NEXT:  entry:
744 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
745 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
746 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
747 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
748 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 8
749 // CHECK1-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
750 // CHECK1-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
751 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
752 // CHECK1-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
753 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
754 // CHECK1-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
755 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
756 // CHECK1-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
757 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
758 // CHECK1-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
759 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
760 // CHECK1-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
761 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
762 // CHECK1-NEXT:    ret void
763 //
764 //
765 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..4
766 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
767 // CHECK1-NEXT:  entry:
768 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
769 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
770 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
771 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
772 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i64, align 8
773 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
774 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
775 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
776 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
777 // CHECK1-NEXT:    [[I:%.*]] = alloca i64, align 8
778 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
779 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
780 // CHECK1-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
781 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
782 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
783 // CHECK1-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
784 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
785 // CHECK1-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
786 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
787 // CHECK1-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
788 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
789 // CHECK1-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
790 // CHECK1-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
791 // CHECK1-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
792 // CHECK1-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
793 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
794 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
795 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
796 // CHECK1-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 1073741859, i64 0, i64 16908287, i64 1, i64 1)
797 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
798 // CHECK1:       omp.dispatch.cond:
799 // CHECK1-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
800 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
801 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
802 // CHECK1:       omp.dispatch.body:
803 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
804 // CHECK1-NEXT:    store i64 [[TMP12]], i64* [[DOTOMP_IV]], align 8
805 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
806 // CHECK1:       omp.inner.for.cond:
807 // CHECK1-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
808 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !5
809 // CHECK1-NEXT:    [[ADD:%.*]] = add i64 [[TMP14]], 1
810 // CHECK1-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP13]], [[ADD]]
811 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
812 // CHECK1:       omp.inner.for.body:
813 // CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
814 // CHECK1-NEXT:    [[MUL:%.*]] = mul i64 [[TMP15]], 127
815 // CHECK1-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
816 // CHECK1-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8, !llvm.access.group !5
817 // CHECK1-NEXT:    [[TMP16:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !5
818 // CHECK1-NEXT:    [[TMP17:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
819 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[TMP17]]
820 // CHECK1-NEXT:    [[TMP18:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !5
821 // CHECK1-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !5
822 // CHECK1-NEXT:    [[TMP20:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
823 // CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[TMP20]]
824 // CHECK1-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX2]], align 4, !llvm.access.group !5
825 // CHECK1-NEXT:    [[MUL3:%.*]] = fmul float [[TMP18]], [[TMP21]]
826 // CHECK1-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !5
827 // CHECK1-NEXT:    [[TMP23:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
828 // CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[TMP23]]
829 // CHECK1-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !llvm.access.group !5
830 // CHECK1-NEXT:    [[MUL5:%.*]] = fmul float [[MUL3]], [[TMP24]]
831 // CHECK1-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !5
832 // CHECK1-NEXT:    [[TMP26:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
833 // CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[TMP26]]
834 // CHECK1-NEXT:    store float [[MUL5]], float* [[ARRAYIDX6]], align 4, !llvm.access.group !5
835 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
836 // CHECK1:       omp.body.continue:
837 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
838 // CHECK1:       omp.inner.for.inc:
839 // CHECK1-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
840 // CHECK1-NEXT:    [[ADD7:%.*]] = add i64 [[TMP27]], 1
841 // CHECK1-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
842 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
843 // CHECK1:       omp.inner.for.end:
844 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
845 // CHECK1:       omp.dispatch.inc:
846 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
847 // CHECK1:       omp.dispatch.end:
848 // CHECK1-NEXT:    ret void
849 //
850 //
851 // CHECK1-LABEL: define {{[^@]+}}@_Z7guided7PfS_S_S_
852 // CHECK1-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
853 // CHECK1-NEXT:  entry:
854 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
855 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
856 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
857 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
858 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
859 // CHECK1-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
860 // CHECK1-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
861 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
862 // CHECK1-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
863 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
864 // CHECK1-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
865 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
866 // CHECK1-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
867 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
868 // CHECK1-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
869 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
870 // CHECK1-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
871 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
872 // CHECK1-NEXT:    ret void
873 //
874 //
875 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..5
876 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
877 // CHECK1-NEXT:  entry:
878 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
879 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
880 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
881 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
882 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i64, align 8
883 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
884 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
885 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
886 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
887 // CHECK1-NEXT:    [[I:%.*]] = alloca i64, align 8
888 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
889 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
890 // CHECK1-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
891 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
892 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
893 // CHECK1-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
894 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
895 // CHECK1-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
896 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
897 // CHECK1-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
898 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
899 // CHECK1-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
900 // CHECK1-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
901 // CHECK1-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
902 // CHECK1-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
903 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
904 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
905 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
906 // CHECK1-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 1073741860, i64 0, i64 16908287, i64 1, i64 7)
907 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
908 // CHECK1:       omp.dispatch.cond:
909 // CHECK1-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
910 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
911 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
912 // CHECK1:       omp.dispatch.body:
913 // CHECK1-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
914 // CHECK1-NEXT:    store i64 [[TMP12]], i64* [[DOTOMP_IV]], align 8
915 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
916 // CHECK1:       omp.inner.for.cond:
917 // CHECK1-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
918 // CHECK1-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !8
919 // CHECK1-NEXT:    [[ADD:%.*]] = add i64 [[TMP14]], 1
920 // CHECK1-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP13]], [[ADD]]
921 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
922 // CHECK1:       omp.inner.for.body:
923 // CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
924 // CHECK1-NEXT:    [[MUL:%.*]] = mul i64 [[TMP15]], 127
925 // CHECK1-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
926 // CHECK1-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8, !llvm.access.group !8
927 // CHECK1-NEXT:    [[TMP16:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !8
928 // CHECK1-NEXT:    [[TMP17:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
929 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[TMP17]]
930 // CHECK1-NEXT:    [[TMP18:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !8
931 // CHECK1-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !8
932 // CHECK1-NEXT:    [[TMP20:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
933 // CHECK1-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[TMP20]]
934 // CHECK1-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX2]], align 4, !llvm.access.group !8
935 // CHECK1-NEXT:    [[MUL3:%.*]] = fmul float [[TMP18]], [[TMP21]]
936 // CHECK1-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !8
937 // CHECK1-NEXT:    [[TMP23:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
938 // CHECK1-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[TMP23]]
939 // CHECK1-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !llvm.access.group !8
940 // CHECK1-NEXT:    [[MUL5:%.*]] = fmul float [[MUL3]], [[TMP24]]
941 // CHECK1-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !8
942 // CHECK1-NEXT:    [[TMP26:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
943 // CHECK1-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[TMP26]]
944 // CHECK1-NEXT:    store float [[MUL5]], float* [[ARRAYIDX6]], align 4, !llvm.access.group !8
945 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
946 // CHECK1:       omp.body.continue:
947 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
948 // CHECK1:       omp.inner.for.inc:
949 // CHECK1-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
950 // CHECK1-NEXT:    [[ADD7:%.*]] = add i64 [[TMP27]], 1
951 // CHECK1-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
952 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
953 // CHECK1:       omp.inner.for.end:
954 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
955 // CHECK1:       omp.dispatch.inc:
956 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
957 // CHECK1:       omp.dispatch.end:
958 // CHECK1-NEXT:    ret void
959 //
960 //
961 // CHECK1-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
962 // CHECK1-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
963 // CHECK1-NEXT:  entry:
964 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
965 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
966 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
967 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
968 // CHECK1-NEXT:    [[X:%.*]] = alloca i32, align 4
969 // CHECK1-NEXT:    [[Y:%.*]] = alloca i32, align 4
970 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
971 // CHECK1-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
972 // CHECK1-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
973 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
974 // CHECK1-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
975 // CHECK1-NEXT:    store i32 0, i32* [[X]], align 4
976 // CHECK1-NEXT:    store i32 0, i32* [[Y]], align 4
977 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
978 // CHECK1-NEXT:    store i32* [[Y]], i32** [[TMP0]], align 8
979 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
980 // CHECK1-NEXT:    store float** [[A_ADDR]], float*** [[TMP1]], align 8
981 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
982 // CHECK1-NEXT:    store float** [[B_ADDR]], float*** [[TMP2]], align 8
983 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
984 // CHECK1-NEXT:    store float** [[C_ADDR]], float*** [[TMP3]], align 8
985 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
986 // CHECK1-NEXT:    store float** [[D_ADDR]], float*** [[TMP4]], align 8
987 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
988 // CHECK1-NEXT:    ret void
989 //
990 //
991 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..6
992 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
993 // CHECK1-NEXT:  entry:
994 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
995 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
996 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
997 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
998 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i8, align 1
999 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1000 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
1001 // CHECK1-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
1002 // CHECK1-NEXT:    [[I:%.*]] = alloca i8, align 1
1003 // CHECK1-NEXT:    [[X:%.*]] = alloca i32, align 4
1004 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
1005 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
1006 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1007 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1008 // CHECK1-NEXT:    [[I7:%.*]] = alloca i8, align 1
1009 // CHECK1-NEXT:    [[X8:%.*]] = alloca i32, align 4
1010 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1011 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1012 // CHECK1-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
1013 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
1014 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
1015 // CHECK1-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8
1016 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
1017 // CHECK1-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
1018 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
1019 // CHECK1-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
1020 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
1021 // CHECK1-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
1022 // CHECK1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
1023 // CHECK1-NEXT:    [[TMP10:%.*]] = load float**, float*** [[TMP9]], align 8
1024 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP2]], align 4
1025 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP11]] to i8
1026 // CHECK1-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
1027 // CHECK1-NEXT:    [[TMP12:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
1028 // CHECK1-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP12]] to i32
1029 // CHECK1-NEXT:    [[SUB:%.*]] = sub i32 57, [[CONV3]]
1030 // CHECK1-NEXT:    [[ADD:%.*]] = add i32 [[SUB]], 1
1031 // CHECK1-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
1032 // CHECK1-NEXT:    [[CONV4:%.*]] = zext i32 [[DIV]] to i64
1033 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV4]], 11
1034 // CHECK1-NEXT:    [[SUB5:%.*]] = sub nsw i64 [[MUL]], 1
1035 // CHECK1-NEXT:    store i64 [[SUB5]], i64* [[DOTCAPTURE_EXPR_2]], align 8
1036 // CHECK1-NEXT:    [[TMP13:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
1037 // CHECK1-NEXT:    store i8 [[TMP13]], i8* [[I]], align 1
1038 // CHECK1-NEXT:    store i32 11, i32* [[X]], align 4
1039 // CHECK1-NEXT:    [[TMP14:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
1040 // CHECK1-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP14]] to i32
1041 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[CONV6]], 57
1042 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1043 // CHECK1:       omp.precond.then:
1044 // CHECK1-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
1045 // CHECK1-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
1046 // CHECK1-NEXT:    store i64 [[TMP15]], i64* [[DOTOMP_UB]], align 8
1047 // CHECK1-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
1048 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1049 // CHECK1-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
1050 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1051 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
1052 // CHECK1-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB2]], i32 [[TMP18]], i32 1073741862, i64 0, i64 [[TMP16]], i64 1, i64 1)
1053 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
1054 // CHECK1:       omp.dispatch.cond:
1055 // CHECK1-NEXT:    [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1056 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
1057 // CHECK1-NEXT:    [[TMP21:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB2]], i32 [[TMP20]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
1058 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP21]], 0
1059 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1060 // CHECK1:       omp.dispatch.body:
1061 // CHECK1-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
1062 // CHECK1-NEXT:    store i64 [[TMP22]], i64* [[DOTOMP_IV]], align 8
1063 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1064 // CHECK1:       omp.inner.for.cond:
1065 // CHECK1-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
1066 // CHECK1-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !11
1067 // CHECK1-NEXT:    [[CMP9:%.*]] = icmp sle i64 [[TMP23]], [[TMP24]]
1068 // CHECK1-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1069 // CHECK1:       omp.inner.for.body:
1070 // CHECK1-NEXT:    [[TMP25:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !11
1071 // CHECK1-NEXT:    [[CONV10:%.*]] = sext i8 [[TMP25]] to i64
1072 // CHECK1-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
1073 // CHECK1-NEXT:    [[DIV11:%.*]] = sdiv i64 [[TMP26]], 11
1074 // CHECK1-NEXT:    [[MUL12:%.*]] = mul nsw i64 [[DIV11]], 1
1075 // CHECK1-NEXT:    [[ADD13:%.*]] = add nsw i64 [[CONV10]], [[MUL12]]
1076 // CHECK1-NEXT:    [[CONV14:%.*]] = trunc i64 [[ADD13]] to i8
1077 // CHECK1-NEXT:    store i8 [[CONV14]], i8* [[I7]], align 1, !llvm.access.group !11
1078 // CHECK1-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
1079 // CHECK1-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
1080 // CHECK1-NEXT:    [[DIV15:%.*]] = sdiv i64 [[TMP28]], 11
1081 // CHECK1-NEXT:    [[MUL16:%.*]] = mul nsw i64 [[DIV15]], 11
1082 // CHECK1-NEXT:    [[SUB17:%.*]] = sub nsw i64 [[TMP27]], [[MUL16]]
1083 // CHECK1-NEXT:    [[MUL18:%.*]] = mul nsw i64 [[SUB17]], 1
1084 // CHECK1-NEXT:    [[SUB19:%.*]] = sub nsw i64 11, [[MUL18]]
1085 // CHECK1-NEXT:    [[CONV20:%.*]] = trunc i64 [[SUB19]] to i32
1086 // CHECK1-NEXT:    store i32 [[CONV20]], i32* [[X8]], align 4, !llvm.access.group !11
1087 // CHECK1-NEXT:    [[TMP29:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !11
1088 // CHECK1-NEXT:    [[TMP30:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
1089 // CHECK1-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP30]] to i64
1090 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP29]], i64 [[IDXPROM]]
1091 // CHECK1-NEXT:    [[TMP31:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !11
1092 // CHECK1-NEXT:    [[TMP32:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !11
1093 // CHECK1-NEXT:    [[TMP33:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
1094 // CHECK1-NEXT:    [[IDXPROM21:%.*]] = sext i8 [[TMP33]] to i64
1095 // CHECK1-NEXT:    [[ARRAYIDX22:%.*]] = getelementptr inbounds float, float* [[TMP32]], i64 [[IDXPROM21]]
1096 // CHECK1-NEXT:    [[TMP34:%.*]] = load float, float* [[ARRAYIDX22]], align 4, !llvm.access.group !11
1097 // CHECK1-NEXT:    [[MUL23:%.*]] = fmul float [[TMP31]], [[TMP34]]
1098 // CHECK1-NEXT:    [[TMP35:%.*]] = load float*, float** [[TMP10]], align 8, !llvm.access.group !11
1099 // CHECK1-NEXT:    [[TMP36:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
1100 // CHECK1-NEXT:    [[IDXPROM24:%.*]] = sext i8 [[TMP36]] to i64
1101 // CHECK1-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds float, float* [[TMP35]], i64 [[IDXPROM24]]
1102 // CHECK1-NEXT:    [[TMP37:%.*]] = load float, float* [[ARRAYIDX25]], align 4, !llvm.access.group !11
1103 // CHECK1-NEXT:    [[MUL26:%.*]] = fmul float [[MUL23]], [[TMP37]]
1104 // CHECK1-NEXT:    [[TMP38:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !11
1105 // CHECK1-NEXT:    [[TMP39:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
1106 // CHECK1-NEXT:    [[IDXPROM27:%.*]] = sext i8 [[TMP39]] to i64
1107 // CHECK1-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds float, float* [[TMP38]], i64 [[IDXPROM27]]
1108 // CHECK1-NEXT:    store float [[MUL26]], float* [[ARRAYIDX28]], align 4, !llvm.access.group !11
1109 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1110 // CHECK1:       omp.body.continue:
1111 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1112 // CHECK1:       omp.inner.for.inc:
1113 // CHECK1-NEXT:    [[TMP40:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
1114 // CHECK1-NEXT:    [[ADD29:%.*]] = add nsw i64 [[TMP40]], 1
1115 // CHECK1-NEXT:    store i64 [[ADD29]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
1116 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
1117 // CHECK1:       omp.inner.for.end:
1118 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
1119 // CHECK1:       omp.dispatch.inc:
1120 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
1121 // CHECK1:       omp.dispatch.end:
1122 // CHECK1-NEXT:    br label [[OMP_PRECOND_END]]
1123 // CHECK1:       omp.precond.end:
1124 // CHECK1-NEXT:    ret void
1125 //
1126 //
1127 // CHECK1-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
1128 // CHECK1-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
1129 // CHECK1-NEXT:  entry:
1130 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
1131 // CHECK1-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
1132 // CHECK1-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
1133 // CHECK1-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
1134 // CHECK1-NEXT:    [[X:%.*]] = alloca i32, align 4
1135 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 8
1136 // CHECK1-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
1137 // CHECK1-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
1138 // CHECK1-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
1139 // CHECK1-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
1140 // CHECK1-NEXT:    store i32 0, i32* [[X]], align 4
1141 // CHECK1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1142 // CHECK1-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
1143 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1144 // CHECK1-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
1145 // CHECK1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
1146 // CHECK1-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
1147 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
1148 // CHECK1-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
1149 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.6*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
1150 // CHECK1-NEXT:    ret void
1151 //
1152 //
1153 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..7
1154 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1155 // CHECK1-NEXT:  entry:
1156 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1157 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1158 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 8
1159 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1160 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i8, align 1
1161 // CHECK1-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
1162 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1163 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1164 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1165 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1166 // CHECK1-NEXT:    [[I:%.*]] = alloca i8, align 1
1167 // CHECK1-NEXT:    [[X:%.*]] = alloca i32, align 4
1168 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1169 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1170 // CHECK1-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 8
1171 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 8
1172 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
1173 // CHECK1-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
1174 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
1175 // CHECK1-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
1176 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
1177 // CHECK1-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
1178 // CHECK1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
1179 // CHECK1-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
1180 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1181 // CHECK1-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
1182 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1183 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1184 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1185 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
1186 // CHECK1-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 1073741861, i32 0, i32 199, i32 1, i32 1)
1187 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
1188 // CHECK1:       omp.dispatch.cond:
1189 // CHECK1-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
1190 // CHECK1-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
1191 // CHECK1-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1192 // CHECK1:       omp.dispatch.body:
1193 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1194 // CHECK1-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
1195 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1196 // CHECK1:       omp.inner.for.cond:
1197 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
1198 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
1199 // CHECK1-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
1200 // CHECK1-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1201 // CHECK1:       omp.inner.for.body:
1202 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
1203 // CHECK1-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP15]], 20
1204 // CHECK1-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
1205 // CHECK1-NEXT:    [[ADD:%.*]] = add nsw i32 48, [[MUL]]
1206 // CHECK1-NEXT:    [[CONV:%.*]] = trunc i32 [[ADD]] to i8
1207 // CHECK1-NEXT:    store i8 [[CONV]], i8* [[I]], align 1, !llvm.access.group !14
1208 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
1209 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
1210 // CHECK1-NEXT:    [[DIV2:%.*]] = sdiv i32 [[TMP17]], 20
1211 // CHECK1-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[DIV2]], 20
1212 // CHECK1-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP16]], [[MUL3]]
1213 // CHECK1-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[SUB]], 1
1214 // CHECK1-NEXT:    [[ADD5:%.*]] = add nsw i32 -10, [[MUL4]]
1215 // CHECK1-NEXT:    store i32 [[ADD5]], i32* [[X]], align 4, !llvm.access.group !14
1216 // CHECK1-NEXT:    [[TMP18:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !14
1217 // CHECK1-NEXT:    [[TMP19:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
1218 // CHECK1-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP19]] to i64
1219 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP18]], i64 [[IDXPROM]]
1220 // CHECK1-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !14
1221 // CHECK1-NEXT:    [[TMP21:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !14
1222 // CHECK1-NEXT:    [[TMP22:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
1223 // CHECK1-NEXT:    [[IDXPROM6:%.*]] = zext i8 [[TMP22]] to i64
1224 // CHECK1-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP21]], i64 [[IDXPROM6]]
1225 // CHECK1-NEXT:    [[TMP23:%.*]] = load float, float* [[ARRAYIDX7]], align 4, !llvm.access.group !14
1226 // CHECK1-NEXT:    [[MUL8:%.*]] = fmul float [[TMP20]], [[TMP23]]
1227 // CHECK1-NEXT:    [[TMP24:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !14
1228 // CHECK1-NEXT:    [[TMP25:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
1229 // CHECK1-NEXT:    [[IDXPROM9:%.*]] = zext i8 [[TMP25]] to i64
1230 // CHECK1-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP24]], i64 [[IDXPROM9]]
1231 // CHECK1-NEXT:    [[TMP26:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
1232 // CHECK1-NEXT:    [[MUL11:%.*]] = fmul float [[MUL8]], [[TMP26]]
1233 // CHECK1-NEXT:    [[TMP27:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !14
1234 // CHECK1-NEXT:    [[TMP28:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
1235 // CHECK1-NEXT:    [[IDXPROM12:%.*]] = zext i8 [[TMP28]] to i64
1236 // CHECK1-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds float, float* [[TMP27]], i64 [[IDXPROM12]]
1237 // CHECK1-NEXT:    store float [[MUL11]], float* [[ARRAYIDX13]], align 4, !llvm.access.group !14
1238 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1239 // CHECK1:       omp.body.continue:
1240 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1241 // CHECK1:       omp.inner.for.inc:
1242 // CHECK1-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
1243 // CHECK1-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP29]], 1
1244 // CHECK1-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
1245 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
1246 // CHECK1:       omp.inner.for.end:
1247 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
1248 // CHECK1:       omp.dispatch.inc:
1249 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
1250 // CHECK1:       omp.dispatch.end:
1251 // CHECK1-NEXT:    ret void
1252 //
1253 //
1254 // CHECK1-LABEL: define {{[^@]+}}@_Z3foov
1255 // CHECK1-SAME: () #[[ATTR3:[0-9]+]] {
1256 // CHECK1-NEXT:  entry:
1257 // CHECK1-NEXT:    call void @_Z8mayThrowv()
1258 // CHECK1-NEXT:    ret i32 0
1259 //
1260 //
1261 // CHECK1-LABEL: define {{[^@]+}}@_Z12parallel_forPfi
1262 // CHECK1-SAME: (float* [[A:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
1263 // CHECK1-NEXT:  entry:
1264 // CHECK1-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
1265 // CHECK1-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
1266 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
1267 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1268 // CHECK1-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 8
1269 // CHECK1-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
1270 // CHECK1-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
1271 // CHECK1-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
1272 // CHECK1-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
1273 // CHECK1-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
1274 // CHECK1-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
1275 // CHECK1-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP1]], align 16
1276 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
1277 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1278 // CHECK1-NEXT:    store float** [[A_ADDR]], float*** [[TMP3]], align 8
1279 // CHECK1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1280 // CHECK1-NEXT:    store i64 [[TMP1]], i64* [[TMP4]], align 8
1281 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
1282 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
1283 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
1284 // CHECK1-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.7*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
1285 // CHECK1-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1286 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP7]])
1287 // CHECK1-NEXT:    ret void
1288 //
1289 //
1290 // CHECK1-LABEL: define {{[^@]+}}@.omp_outlined..8
1291 // CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {
1292 // CHECK1-NEXT:  entry:
1293 // CHECK1-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1294 // CHECK1-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1295 // CHECK1-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 8
1296 // CHECK1-NEXT:    [[N:%.*]] = alloca i32, align 4
1297 // CHECK1-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1298 // CHECK1-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1299 // CHECK1-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1300 // CHECK1-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1301 // CHECK1-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1302 // CHECK1-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1303 // CHECK1-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
1304 // CHECK1-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
1305 // CHECK1-NEXT:    [[I:%.*]] = alloca i32, align 4
1306 // CHECK1-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1307 // CHECK1-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1308 // CHECK1-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 8
1309 // CHECK1-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 8
1310 // CHECK1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
1311 // CHECK1-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
1312 // CHECK1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
1313 // CHECK1-NEXT:    [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8
1314 // CHECK1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
1315 // CHECK1-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 8
1316 // CHECK1-NEXT:    store i32 [[TMP6]], i32* [[N]], align 4
1317 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1318 // CHECK1-NEXT:    store i32 16908288, i32* [[DOTOMP_UB]], align 4
1319 // CHECK1-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1320 // CHECK1-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1321 // CHECK1-NEXT:    [[TMP7:%.*]] = call i8* @llvm.stacksave()
1322 // CHECK1-NEXT:    store i8* [[TMP7]], i8** [[SAVED_STACK]], align 8
1323 // CHECK1-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP4]], align 16
1324 // CHECK1-NEXT:    store i64 [[TMP4]], i64* [[__VLA_EXPR0]], align 8
1325 // CHECK1-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1326 // CHECK1-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
1327 // CHECK1-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 5)
1328 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
1329 // CHECK1:       omp.dispatch.cond:
1330 // CHECK1-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1331 // CHECK1-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP10]], 16908288
1332 // CHECK1-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1333 // CHECK1:       cond.true:
1334 // CHECK1-NEXT:    br label [[COND_END:%.*]]
1335 // CHECK1:       cond.false:
1336 // CHECK1-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1337 // CHECK1-NEXT:    br label [[COND_END]]
1338 // CHECK1:       cond.end:
1339 // CHECK1-NEXT:    [[COND:%.*]] = phi i32 [ 16908288, [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
1340 // CHECK1-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1341 // CHECK1-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1342 // CHECK1-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
1343 // CHECK1-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1344 // CHECK1-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1345 // CHECK1-NEXT:    [[CMP1:%.*]] = icmp ule i32 [[TMP13]], [[TMP14]]
1346 // CHECK1-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_CLEANUP:%.*]]
1347 // CHECK1:       omp.dispatch.cleanup:
1348 // CHECK1-NEXT:    br label [[OMP_DISPATCH_END:%.*]]
1349 // CHECK1:       omp.dispatch.body:
1350 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1351 // CHECK1:       omp.inner.for.cond:
1352 // CHECK1-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1353 // CHECK1-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1354 // CHECK1-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[TMP15]], [[TMP16]]
1355 // CHECK1-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
1356 // CHECK1:       omp.inner.for.cond.cleanup:
1357 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
1358 // CHECK1:       omp.inner.for.body:
1359 // CHECK1-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1360 // CHECK1-NEXT:    [[MUL:%.*]] = mul i32 [[TMP17]], 127
1361 // CHECK1-NEXT:    [[ADD:%.*]] = add i32 131071, [[MUL]]
1362 // CHECK1-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
1363 // CHECK1-NEXT:    [[CALL:%.*]] = invoke i32 @_Z3foov()
1364 // CHECK1-NEXT:    to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
1365 // CHECK1:       invoke.cont:
1366 // CHECK1-NEXT:    [[CONV:%.*]] = sitofp i32 [[CALL]] to float
1367 // CHECK1-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
1368 // CHECK1-NEXT:    [[IDXPROM:%.*]] = zext i32 [[TMP18]] to i64
1369 // CHECK1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[VLA]], i64 [[IDXPROM]]
1370 // CHECK1-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
1371 // CHECK1-NEXT:    [[ADD3:%.*]] = fadd float [[CONV]], [[TMP19]]
1372 // CHECK1-NEXT:    [[TMP20:%.*]] = load i32, i32* [[N]], align 4
1373 // CHECK1-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP20]] to float
1374 // CHECK1-NEXT:    [[ADD5:%.*]] = fadd float [[ADD3]], [[CONV4]]
1375 // CHECK1-NEXT:    [[TMP21:%.*]] = load float*, float** [[TMP2]], align 8
1376 // CHECK1-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I]], align 4
1377 // CHECK1-NEXT:    [[IDXPROM6:%.*]] = zext i32 [[TMP22]] to i64
1378 // CHECK1-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP21]], i64 [[IDXPROM6]]
1379 // CHECK1-NEXT:    [[TMP23:%.*]] = load float, float* [[ARRAYIDX7]], align 4
1380 // CHECK1-NEXT:    [[ADD8:%.*]] = fadd float [[TMP23]], [[ADD5]]
1381 // CHECK1-NEXT:    store float [[ADD8]], float* [[ARRAYIDX7]], align 4
1382 // CHECK1-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1383 // CHECK1:       omp.body.continue:
1384 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1385 // CHECK1:       omp.inner.for.inc:
1386 // CHECK1-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1387 // CHECK1-NEXT:    [[ADD9:%.*]] = add i32 [[TMP24]], 1
1388 // CHECK1-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
1389 // CHECK1-NEXT:    br label [[OMP_INNER_FOR_COND]]
1390 // CHECK1:       omp.inner.for.end:
1391 // CHECK1-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
1392 // CHECK1:       omp.dispatch.inc:
1393 // CHECK1-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1394 // CHECK1-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1395 // CHECK1-NEXT:    [[ADD10:%.*]] = add i32 [[TMP25]], [[TMP26]]
1396 // CHECK1-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_LB]], align 4
1397 // CHECK1-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1398 // CHECK1-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1399 // CHECK1-NEXT:    [[ADD11:%.*]] = add i32 [[TMP27]], [[TMP28]]
1400 // CHECK1-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_UB]], align 4
1401 // CHECK1-NEXT:    br label [[OMP_DISPATCH_COND]]
1402 // CHECK1:       omp.dispatch.end:
1403 // CHECK1-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]])
1404 // CHECK1-NEXT:    [[TMP29:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
1405 // CHECK1-NEXT:    call void @llvm.stackrestore(i8* [[TMP29]])
1406 // CHECK1-NEXT:    ret void
1407 // CHECK1:       terminate.lpad:
1408 // CHECK1-NEXT:    [[TMP30:%.*]] = landingpad { i8*, i32 }
1409 // CHECK1-NEXT:    catch i8* null
1410 // CHECK1-NEXT:    [[TMP31:%.*]] = extractvalue { i8*, i32 } [[TMP30]], 0
1411 // CHECK1-NEXT:    call void @__clang_call_terminate(i8* [[TMP31]]) #[[ATTR7:[0-9]+]]
1412 // CHECK1-NEXT:    unreachable
1413 //
1414 //
1415 // CHECK1-LABEL: define {{[^@]+}}@__clang_call_terminate
1416 // CHECK1-SAME: (i8* [[TMP0:%.*]]) #[[ATTR6:[0-9]+]] comdat {
1417 // CHECK1-NEXT:    [[TMP2:%.*]] = call i8* @__cxa_begin_catch(i8* [[TMP0]]) #[[ATTR2:[0-9]+]]
1418 // CHECK1-NEXT:    call void @_ZSt9terminatev() #[[ATTR7]]
1419 // CHECK1-NEXT:    unreachable
1420 //
1421 //
1422 // CHECK2-LABEL: define {{[^@]+}}@_Z17with_var_schedulev
1423 // CHECK2-SAME: () #[[ATTR0:[0-9]+]] {
1424 // CHECK2-NEXT:  entry:
1425 // CHECK2-NEXT:    [[A:%.*]] = alloca double, align 8
1426 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
1427 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
1428 // CHECK2-NEXT:    store double 5.000000e+00, double* [[A]], align 8
1429 // CHECK2-NEXT:    [[TMP0:%.*]] = load double, double* [[A]], align 8
1430 // CHECK2-NEXT:    [[CONV:%.*]] = fptosi double [[TMP0]] to i8
1431 // CHECK2-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
1432 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1433 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
1434 // CHECK2-NEXT:    store i8 [[TMP2]], i8* [[TMP1]], align 1
1435 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
1436 // CHECK2-NEXT:    ret void
1437 //
1438 //
1439 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
1440 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
1441 // CHECK2-NEXT:  entry:
1442 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1443 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1444 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
1445 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
1446 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
1447 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i64, align 8
1448 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca double, align 8
1449 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
1450 // CHECK2-NEXT:    [[I:%.*]] = alloca i64, align 8
1451 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
1452 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
1453 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1454 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1455 // CHECK2-NEXT:    [[A:%.*]] = alloca double, align 8
1456 // CHECK2-NEXT:    [[I4:%.*]] = alloca i64, align 8
1457 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1458 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1459 // CHECK2-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
1460 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
1461 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
1462 // CHECK2-NEXT:    [[TMP2:%.*]] = load i8, i8* [[TMP1]], align 1
1463 // CHECK2-NEXT:    store i8 [[TMP2]], i8* [[DOTCAPTURE_EXPR_]], align 1
1464 // CHECK2-NEXT:    [[TMP3:%.*]] = load double, double* undef, align 8
1465 // CHECK2-NEXT:    [[ADD:%.*]] = fadd double 2.000000e+00, [[TMP3]]
1466 // CHECK2-NEXT:    store double [[ADD]], double* [[DOTCAPTURE_EXPR_1]], align 8
1467 // CHECK2-NEXT:    [[TMP4:%.*]] = load double, double* [[DOTCAPTURE_EXPR_1]], align 8
1468 // CHECK2-NEXT:    [[SUB:%.*]] = fsub double [[TMP4]], 1.000000e+00
1469 // CHECK2-NEXT:    [[DIV:%.*]] = fdiv double [[SUB]], 1.000000e+00
1470 // CHECK2-NEXT:    [[CONV:%.*]] = fptoui double [[DIV]] to i64
1471 // CHECK2-NEXT:    [[SUB3:%.*]] = sub i64 [[CONV]], 1
1472 // CHECK2-NEXT:    store i64 [[SUB3]], i64* [[DOTCAPTURE_EXPR_2]], align 8
1473 // CHECK2-NEXT:    store i64 1, i64* [[I]], align 8
1474 // CHECK2-NEXT:    [[TMP5:%.*]] = load double, double* [[DOTCAPTURE_EXPR_1]], align 8
1475 // CHECK2-NEXT:    [[CMP:%.*]] = fcmp olt double 1.000000e+00, [[TMP5]]
1476 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
1477 // CHECK2:       omp.precond.then:
1478 // CHECK2-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
1479 // CHECK2-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
1480 // CHECK2-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_UB]], align 8
1481 // CHECK2-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
1482 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1483 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
1484 // CHECK2-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP7]] to i64
1485 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1486 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
1487 // CHECK2-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 [[CONV5]])
1488 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
1489 // CHECK2:       omp.dispatch.cond:
1490 // CHECK2-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1491 // CHECK2-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
1492 // CHECK2-NEXT:    [[CMP6:%.*]] = icmp ugt i64 [[TMP10]], [[TMP11]]
1493 // CHECK2-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1494 // CHECK2:       cond.true:
1495 // CHECK2-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
1496 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1497 // CHECK2:       cond.false:
1498 // CHECK2-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1499 // CHECK2-NEXT:    br label [[COND_END]]
1500 // CHECK2:       cond.end:
1501 // CHECK2-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
1502 // CHECK2-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
1503 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
1504 // CHECK2-NEXT:    store i64 [[TMP14]], i64* [[DOTOMP_IV]], align 8
1505 // CHECK2-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
1506 // CHECK2-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1507 // CHECK2-NEXT:    [[ADD7:%.*]] = add i64 [[TMP16]], 1
1508 // CHECK2-NEXT:    [[CMP8:%.*]] = icmp ult i64 [[TMP15]], [[ADD7]]
1509 // CHECK2-NEXT:    br i1 [[CMP8]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1510 // CHECK2:       omp.dispatch.body:
1511 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1512 // CHECK2:       omp.inner.for.cond:
1513 // CHECK2-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
1514 // CHECK2-NEXT:    [[TMP18:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1515 // CHECK2-NEXT:    [[ADD9:%.*]] = add i64 [[TMP18]], 1
1516 // CHECK2-NEXT:    [[CMP10:%.*]] = icmp ult i64 [[TMP17]], [[ADD9]]
1517 // CHECK2-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1518 // CHECK2:       omp.inner.for.body:
1519 // CHECK2-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
1520 // CHECK2-NEXT:    [[MUL:%.*]] = mul i64 [[TMP19]], 1
1521 // CHECK2-NEXT:    [[ADD11:%.*]] = add i64 1, [[MUL]]
1522 // CHECK2-NEXT:    store i64 [[ADD11]], i64* [[I4]], align 8
1523 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1524 // CHECK2:       omp.body.continue:
1525 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1526 // CHECK2:       omp.inner.for.inc:
1527 // CHECK2-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
1528 // CHECK2-NEXT:    [[ADD12:%.*]] = add i64 [[TMP20]], 1
1529 // CHECK2-NEXT:    store i64 [[ADD12]], i64* [[DOTOMP_IV]], align 8
1530 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
1531 // CHECK2:       omp.inner.for.end:
1532 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
1533 // CHECK2:       omp.dispatch.inc:
1534 // CHECK2-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
1535 // CHECK2-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
1536 // CHECK2-NEXT:    [[ADD13:%.*]] = add i64 [[TMP21]], [[TMP22]]
1537 // CHECK2-NEXT:    store i64 [[ADD13]], i64* [[DOTOMP_LB]], align 8
1538 // CHECK2-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
1539 // CHECK2-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
1540 // CHECK2-NEXT:    [[ADD14:%.*]] = add i64 [[TMP23]], [[TMP24]]
1541 // CHECK2-NEXT:    store i64 [[ADD14]], i64* [[DOTOMP_UB]], align 8
1542 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
1543 // CHECK2:       omp.dispatch.end:
1544 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1545 // CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
1546 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
1547 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
1548 // CHECK2:       omp.precond.end:
1549 // CHECK2-NEXT:    ret void
1550 //
1551 //
1552 // CHECK2-LABEL: define {{[^@]+}}@_Z23without_schedule_clausePfS_S_S_
1553 // CHECK2-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
1554 // CHECK2-NEXT:  entry:
1555 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
1556 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
1557 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
1558 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
1559 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
1560 // CHECK2-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
1561 // CHECK2-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
1562 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
1563 // CHECK2-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
1564 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1565 // CHECK2-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
1566 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1567 // CHECK2-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
1568 // CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
1569 // CHECK2-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
1570 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
1571 // CHECK2-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
1572 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
1573 // CHECK2-NEXT:    ret void
1574 //
1575 //
1576 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..1
1577 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1578 // CHECK2-NEXT:  entry:
1579 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1580 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1581 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
1582 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1583 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1584 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1585 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1586 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1587 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1588 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
1589 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1590 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1591 // CHECK2-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
1592 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
1593 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
1594 // CHECK2-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
1595 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1
1596 // CHECK2-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
1597 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 2
1598 // CHECK2-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
1599 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 3
1600 // CHECK2-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
1601 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1602 // CHECK2-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
1603 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1604 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1605 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1606 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
1607 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1608 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1609 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 4571423
1610 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1611 // CHECK2:       cond.true:
1612 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1613 // CHECK2:       cond.false:
1614 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1615 // CHECK2-NEXT:    br label [[COND_END]]
1616 // CHECK2:       cond.end:
1617 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 4571423, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1618 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1619 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1620 // CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1621 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1622 // CHECK2:       omp.inner.for.cond:
1623 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1624 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1625 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1626 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1627 // CHECK2:       omp.inner.for.body:
1628 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1629 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 7
1630 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 33, [[MUL]]
1631 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
1632 // CHECK2-NEXT:    [[TMP17:%.*]] = load float*, float** [[TMP4]], align 8
1633 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
1634 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
1635 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM]]
1636 // CHECK2-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
1637 // CHECK2-NEXT:    [[TMP20:%.*]] = load float*, float** [[TMP6]], align 8
1638 // CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4
1639 // CHECK2-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP21]] to i64
1640 // CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM2]]
1641 // CHECK2-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX3]], align 4
1642 // CHECK2-NEXT:    [[MUL4:%.*]] = fmul float [[TMP19]], [[TMP22]]
1643 // CHECK2-NEXT:    [[TMP23:%.*]] = load float*, float** [[TMP8]], align 8
1644 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4
1645 // CHECK2-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP24]] to i64
1646 // CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM5]]
1647 // CHECK2-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX6]], align 4
1648 // CHECK2-NEXT:    [[MUL7:%.*]] = fmul float [[MUL4]], [[TMP25]]
1649 // CHECK2-NEXT:    [[TMP26:%.*]] = load float*, float** [[TMP2]], align 8
1650 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I]], align 4
1651 // CHECK2-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP27]] to i64
1652 // CHECK2-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[TMP26]], i64 [[IDXPROM8]]
1653 // CHECK2-NEXT:    store float [[MUL7]], float* [[ARRAYIDX9]], align 4
1654 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1655 // CHECK2:       omp.body.continue:
1656 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1657 // CHECK2:       omp.inner.for.inc:
1658 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1659 // CHECK2-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP28]], 1
1660 // CHECK2-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
1661 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
1662 // CHECK2:       omp.inner.for.end:
1663 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1664 // CHECK2:       omp.loop.exit:
1665 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
1666 // CHECK2-NEXT:    ret void
1667 //
1668 //
1669 // CHECK2-LABEL: define {{[^@]+}}@_Z18static_not_chunkedPfS_S_S_
1670 // CHECK2-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
1671 // CHECK2-NEXT:  entry:
1672 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
1673 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
1674 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
1675 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
1676 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 8
1677 // CHECK2-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
1678 // CHECK2-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
1679 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
1680 // CHECK2-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
1681 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1682 // CHECK2-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
1683 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1684 // CHECK2-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
1685 // CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
1686 // CHECK2-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
1687 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
1688 // CHECK2-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
1689 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
1690 // CHECK2-NEXT:    ret void
1691 //
1692 //
1693 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..2
1694 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1695 // CHECK2-NEXT:  entry:
1696 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1697 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1698 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
1699 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1700 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1701 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1702 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1703 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1704 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1705 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
1706 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1707 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1708 // CHECK2-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
1709 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
1710 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
1711 // CHECK2-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
1712 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
1713 // CHECK2-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
1714 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 2
1715 // CHECK2-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
1716 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 3
1717 // CHECK2-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
1718 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1719 // CHECK2-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
1720 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1721 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1722 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1723 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
1724 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
1725 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1726 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 4571423
1727 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1728 // CHECK2:       cond.true:
1729 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1730 // CHECK2:       cond.false:
1731 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1732 // CHECK2-NEXT:    br label [[COND_END]]
1733 // CHECK2:       cond.end:
1734 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 4571423, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1735 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1736 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1737 // CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1738 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1739 // CHECK2:       omp.inner.for.cond:
1740 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1741 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1742 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
1743 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1744 // CHECK2:       omp.inner.for.body:
1745 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1746 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 7
1747 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
1748 // CHECK2-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
1749 // CHECK2-NEXT:    [[TMP17:%.*]] = load float*, float** [[TMP4]], align 8
1750 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
1751 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
1752 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM]]
1753 // CHECK2-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
1754 // CHECK2-NEXT:    [[TMP20:%.*]] = load float*, float** [[TMP6]], align 8
1755 // CHECK2-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4
1756 // CHECK2-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP21]] to i64
1757 // CHECK2-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM2]]
1758 // CHECK2-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX3]], align 4
1759 // CHECK2-NEXT:    [[MUL4:%.*]] = fmul float [[TMP19]], [[TMP22]]
1760 // CHECK2-NEXT:    [[TMP23:%.*]] = load float*, float** [[TMP8]], align 8
1761 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4
1762 // CHECK2-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP24]] to i64
1763 // CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM5]]
1764 // CHECK2-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX6]], align 4
1765 // CHECK2-NEXT:    [[MUL7:%.*]] = fmul float [[MUL4]], [[TMP25]]
1766 // CHECK2-NEXT:    [[TMP26:%.*]] = load float*, float** [[TMP2]], align 8
1767 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I]], align 4
1768 // CHECK2-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP27]] to i64
1769 // CHECK2-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[TMP26]], i64 [[IDXPROM8]]
1770 // CHECK2-NEXT:    store float [[MUL7]], float* [[ARRAYIDX9]], align 4
1771 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1772 // CHECK2:       omp.body.continue:
1773 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1774 // CHECK2:       omp.inner.for.inc:
1775 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1776 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
1777 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
1778 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
1779 // CHECK2:       omp.inner.for.end:
1780 // CHECK2-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
1781 // CHECK2:       omp.loop.exit:
1782 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
1783 // CHECK2-NEXT:    ret void
1784 //
1785 //
1786 // CHECK2-LABEL: define {{[^@]+}}@_Z14static_chunkedPfS_S_S_
1787 // CHECK2-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
1788 // CHECK2-NEXT:  entry:
1789 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
1790 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
1791 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
1792 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
1793 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
1794 // CHECK2-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
1795 // CHECK2-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
1796 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
1797 // CHECK2-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
1798 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1799 // CHECK2-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
1800 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1801 // CHECK2-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
1802 // CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
1803 // CHECK2-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
1804 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
1805 // CHECK2-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
1806 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
1807 // CHECK2-NEXT:    ret void
1808 //
1809 //
1810 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..3
1811 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1812 // CHECK2-NEXT:  entry:
1813 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1814 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1815 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
1816 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
1817 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
1818 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
1819 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
1820 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
1821 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1822 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
1823 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1824 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1825 // CHECK2-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
1826 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
1827 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
1828 // CHECK2-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
1829 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
1830 // CHECK2-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
1831 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
1832 // CHECK2-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
1833 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
1834 // CHECK2-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
1835 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
1836 // CHECK2-NEXT:    store i32 16908288, i32* [[DOTOMP_UB]], align 4
1837 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
1838 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1839 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1840 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
1841 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 5)
1842 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
1843 // CHECK2:       omp.dispatch.cond:
1844 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1845 // CHECK2-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP11]], 16908288
1846 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
1847 // CHECK2:       cond.true:
1848 // CHECK2-NEXT:    br label [[COND_END:%.*]]
1849 // CHECK2:       cond.false:
1850 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1851 // CHECK2-NEXT:    br label [[COND_END]]
1852 // CHECK2:       cond.end:
1853 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 16908288, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
1854 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
1855 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1856 // CHECK2-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
1857 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1858 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1859 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
1860 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1861 // CHECK2:       omp.dispatch.body:
1862 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1863 // CHECK2:       omp.inner.for.cond:
1864 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1865 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1866 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[TMP16]], [[TMP17]]
1867 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1868 // CHECK2:       omp.inner.for.body:
1869 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1870 // CHECK2-NEXT:    [[MUL:%.*]] = mul i32 [[TMP18]], 127
1871 // CHECK2-NEXT:    [[ADD:%.*]] = add i32 131071, [[MUL]]
1872 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
1873 // CHECK2-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP4]], align 8
1874 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I]], align 4
1875 // CHECK2-NEXT:    [[IDXPROM:%.*]] = zext i32 [[TMP20]] to i64
1876 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[IDXPROM]]
1877 // CHECK2-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX]], align 4
1878 // CHECK2-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP6]], align 8
1879 // CHECK2-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
1880 // CHECK2-NEXT:    [[IDXPROM3:%.*]] = zext i32 [[TMP23]] to i64
1881 // CHECK2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[IDXPROM3]]
1882 // CHECK2-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4
1883 // CHECK2-NEXT:    [[MUL5:%.*]] = fmul float [[TMP21]], [[TMP24]]
1884 // CHECK2-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP8]], align 8
1885 // CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[I]], align 4
1886 // CHECK2-NEXT:    [[IDXPROM6:%.*]] = zext i32 [[TMP26]] to i64
1887 // CHECK2-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[IDXPROM6]]
1888 // CHECK2-NEXT:    [[TMP27:%.*]] = load float, float* [[ARRAYIDX7]], align 4
1889 // CHECK2-NEXT:    [[MUL8:%.*]] = fmul float [[MUL5]], [[TMP27]]
1890 // CHECK2-NEXT:    [[TMP28:%.*]] = load float*, float** [[TMP2]], align 8
1891 // CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[I]], align 4
1892 // CHECK2-NEXT:    [[IDXPROM9:%.*]] = zext i32 [[TMP29]] to i64
1893 // CHECK2-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP28]], i64 [[IDXPROM9]]
1894 // CHECK2-NEXT:    store float [[MUL8]], float* [[ARRAYIDX10]], align 4
1895 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
1896 // CHECK2:       omp.body.continue:
1897 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
1898 // CHECK2:       omp.inner.for.inc:
1899 // CHECK2-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
1900 // CHECK2-NEXT:    [[ADD11:%.*]] = add i32 [[TMP30]], 1
1901 // CHECK2-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4
1902 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
1903 // CHECK2:       omp.inner.for.end:
1904 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
1905 // CHECK2:       omp.dispatch.inc:
1906 // CHECK2-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
1907 // CHECK2-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1908 // CHECK2-NEXT:    [[ADD12:%.*]] = add i32 [[TMP31]], [[TMP32]]
1909 // CHECK2-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_LB]], align 4
1910 // CHECK2-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
1911 // CHECK2-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
1912 // CHECK2-NEXT:    [[ADD13:%.*]] = add i32 [[TMP33]], [[TMP34]]
1913 // CHECK2-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_UB]], align 4
1914 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
1915 // CHECK2:       omp.dispatch.end:
1916 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
1917 // CHECK2-NEXT:    ret void
1918 //
1919 //
1920 // CHECK2-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
1921 // CHECK2-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
1922 // CHECK2-NEXT:  entry:
1923 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
1924 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
1925 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
1926 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
1927 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 8
1928 // CHECK2-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
1929 // CHECK2-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
1930 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
1931 // CHECK2-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
1932 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
1933 // CHECK2-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
1934 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
1935 // CHECK2-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
1936 // CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
1937 // CHECK2-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
1938 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
1939 // CHECK2-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
1940 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
1941 // CHECK2-NEXT:    ret void
1942 //
1943 //
1944 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..4
1945 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
1946 // CHECK2-NEXT:  entry:
1947 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
1948 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
1949 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
1950 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
1951 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i64, align 8
1952 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
1953 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
1954 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
1955 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
1956 // CHECK2-NEXT:    [[I:%.*]] = alloca i64, align 8
1957 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
1958 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
1959 // CHECK2-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
1960 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
1961 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
1962 // CHECK2-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
1963 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
1964 // CHECK2-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
1965 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
1966 // CHECK2-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
1967 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
1968 // CHECK2-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
1969 // CHECK2-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
1970 // CHECK2-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
1971 // CHECK2-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
1972 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
1973 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
1974 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
1975 // CHECK2-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 35, i64 0, i64 16908287, i64 1, i64 1)
1976 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
1977 // CHECK2:       omp.dispatch.cond:
1978 // CHECK2-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
1979 // CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
1980 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
1981 // CHECK2:       omp.dispatch.body:
1982 // CHECK2-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
1983 // CHECK2-NEXT:    store i64 [[TMP12]], i64* [[DOTOMP_IV]], align 8
1984 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
1985 // CHECK2:       omp.inner.for.cond:
1986 // CHECK2-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
1987 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !5
1988 // CHECK2-NEXT:    [[ADD:%.*]] = add i64 [[TMP14]], 1
1989 // CHECK2-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP13]], [[ADD]]
1990 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
1991 // CHECK2:       omp.inner.for.body:
1992 // CHECK2-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
1993 // CHECK2-NEXT:    [[MUL:%.*]] = mul i64 [[TMP15]], 127
1994 // CHECK2-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
1995 // CHECK2-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8, !llvm.access.group !5
1996 // CHECK2-NEXT:    [[TMP16:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !5
1997 // CHECK2-NEXT:    [[TMP17:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
1998 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[TMP17]]
1999 // CHECK2-NEXT:    [[TMP18:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !5
2000 // CHECK2-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !5
2001 // CHECK2-NEXT:    [[TMP20:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
2002 // CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[TMP20]]
2003 // CHECK2-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX2]], align 4, !llvm.access.group !5
2004 // CHECK2-NEXT:    [[MUL3:%.*]] = fmul float [[TMP18]], [[TMP21]]
2005 // CHECK2-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !5
2006 // CHECK2-NEXT:    [[TMP23:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
2007 // CHECK2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[TMP23]]
2008 // CHECK2-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !llvm.access.group !5
2009 // CHECK2-NEXT:    [[MUL5:%.*]] = fmul float [[MUL3]], [[TMP24]]
2010 // CHECK2-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !5
2011 // CHECK2-NEXT:    [[TMP26:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
2012 // CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[TMP26]]
2013 // CHECK2-NEXT:    store float [[MUL5]], float* [[ARRAYIDX6]], align 4, !llvm.access.group !5
2014 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2015 // CHECK2:       omp.body.continue:
2016 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2017 // CHECK2:       omp.inner.for.inc:
2018 // CHECK2-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
2019 // CHECK2-NEXT:    [[ADD7:%.*]] = add i64 [[TMP27]], 1
2020 // CHECK2-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
2021 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
2022 // CHECK2:       omp.inner.for.end:
2023 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
2024 // CHECK2:       omp.dispatch.inc:
2025 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
2026 // CHECK2:       omp.dispatch.end:
2027 // CHECK2-NEXT:    ret void
2028 //
2029 //
2030 // CHECK2-LABEL: define {{[^@]+}}@_Z7guided7PfS_S_S_
2031 // CHECK2-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
2032 // CHECK2-NEXT:  entry:
2033 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
2034 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
2035 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
2036 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
2037 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
2038 // CHECK2-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
2039 // CHECK2-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
2040 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
2041 // CHECK2-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
2042 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2043 // CHECK2-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
2044 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2045 // CHECK2-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
2046 // CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
2047 // CHECK2-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
2048 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
2049 // CHECK2-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
2050 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
2051 // CHECK2-NEXT:    ret void
2052 //
2053 //
2054 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..5
2055 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
2056 // CHECK2-NEXT:  entry:
2057 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2058 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2059 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
2060 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
2061 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i64, align 8
2062 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
2063 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
2064 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
2065 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2066 // CHECK2-NEXT:    [[I:%.*]] = alloca i64, align 8
2067 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2068 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2069 // CHECK2-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
2070 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
2071 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
2072 // CHECK2-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
2073 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
2074 // CHECK2-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
2075 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
2076 // CHECK2-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
2077 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
2078 // CHECK2-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
2079 // CHECK2-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
2080 // CHECK2-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
2081 // CHECK2-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
2082 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2083 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2084 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
2085 // CHECK2-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 36, i64 0, i64 16908287, i64 1, i64 7)
2086 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
2087 // CHECK2:       omp.dispatch.cond:
2088 // CHECK2-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
2089 // CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
2090 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2091 // CHECK2:       omp.dispatch.body:
2092 // CHECK2-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
2093 // CHECK2-NEXT:    store i64 [[TMP12]], i64* [[DOTOMP_IV]], align 8
2094 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2095 // CHECK2:       omp.inner.for.cond:
2096 // CHECK2-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
2097 // CHECK2-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !8
2098 // CHECK2-NEXT:    [[ADD:%.*]] = add i64 [[TMP14]], 1
2099 // CHECK2-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP13]], [[ADD]]
2100 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2101 // CHECK2:       omp.inner.for.body:
2102 // CHECK2-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
2103 // CHECK2-NEXT:    [[MUL:%.*]] = mul i64 [[TMP15]], 127
2104 // CHECK2-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
2105 // CHECK2-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8, !llvm.access.group !8
2106 // CHECK2-NEXT:    [[TMP16:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !8
2107 // CHECK2-NEXT:    [[TMP17:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
2108 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[TMP17]]
2109 // CHECK2-NEXT:    [[TMP18:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !8
2110 // CHECK2-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !8
2111 // CHECK2-NEXT:    [[TMP20:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
2112 // CHECK2-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[TMP20]]
2113 // CHECK2-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX2]], align 4, !llvm.access.group !8
2114 // CHECK2-NEXT:    [[MUL3:%.*]] = fmul float [[TMP18]], [[TMP21]]
2115 // CHECK2-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !8
2116 // CHECK2-NEXT:    [[TMP23:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
2117 // CHECK2-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[TMP23]]
2118 // CHECK2-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !llvm.access.group !8
2119 // CHECK2-NEXT:    [[MUL5:%.*]] = fmul float [[MUL3]], [[TMP24]]
2120 // CHECK2-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !8
2121 // CHECK2-NEXT:    [[TMP26:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
2122 // CHECK2-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[TMP26]]
2123 // CHECK2-NEXT:    store float [[MUL5]], float* [[ARRAYIDX6]], align 4, !llvm.access.group !8
2124 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2125 // CHECK2:       omp.body.continue:
2126 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2127 // CHECK2:       omp.inner.for.inc:
2128 // CHECK2-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
2129 // CHECK2-NEXT:    [[ADD7:%.*]] = add i64 [[TMP27]], 1
2130 // CHECK2-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
2131 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
2132 // CHECK2:       omp.inner.for.end:
2133 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
2134 // CHECK2:       omp.dispatch.inc:
2135 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
2136 // CHECK2:       omp.dispatch.end:
2137 // CHECK2-NEXT:    ret void
2138 //
2139 //
2140 // CHECK2-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
2141 // CHECK2-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
2142 // CHECK2-NEXT:  entry:
2143 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
2144 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
2145 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
2146 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
2147 // CHECK2-NEXT:    [[X:%.*]] = alloca i32, align 4
2148 // CHECK2-NEXT:    [[Y:%.*]] = alloca i32, align 4
2149 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
2150 // CHECK2-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
2151 // CHECK2-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
2152 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
2153 // CHECK2-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
2154 // CHECK2-NEXT:    store i32 0, i32* [[X]], align 4
2155 // CHECK2-NEXT:    store i32 0, i32* [[Y]], align 4
2156 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2157 // CHECK2-NEXT:    store i32* [[Y]], i32** [[TMP0]], align 8
2158 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2159 // CHECK2-NEXT:    store float** [[A_ADDR]], float*** [[TMP1]], align 8
2160 // CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
2161 // CHECK2-NEXT:    store float** [[B_ADDR]], float*** [[TMP2]], align 8
2162 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
2163 // CHECK2-NEXT:    store float** [[C_ADDR]], float*** [[TMP3]], align 8
2164 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
2165 // CHECK2-NEXT:    store float** [[D_ADDR]], float*** [[TMP4]], align 8
2166 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
2167 // CHECK2-NEXT:    ret void
2168 //
2169 //
2170 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..6
2171 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
2172 // CHECK2-NEXT:  entry:
2173 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2174 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2175 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
2176 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
2177 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i8, align 1
2178 // CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
2179 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
2180 // CHECK2-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
2181 // CHECK2-NEXT:    [[I:%.*]] = alloca i8, align 1
2182 // CHECK2-NEXT:    [[X:%.*]] = alloca i32, align 4
2183 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
2184 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
2185 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
2186 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2187 // CHECK2-NEXT:    [[I7:%.*]] = alloca i8, align 1
2188 // CHECK2-NEXT:    [[X8:%.*]] = alloca i32, align 4
2189 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2190 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2191 // CHECK2-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
2192 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
2193 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
2194 // CHECK2-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8
2195 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
2196 // CHECK2-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
2197 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
2198 // CHECK2-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
2199 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
2200 // CHECK2-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
2201 // CHECK2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
2202 // CHECK2-NEXT:    [[TMP10:%.*]] = load float**, float*** [[TMP9]], align 8
2203 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP2]], align 4
2204 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP11]] to i8
2205 // CHECK2-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
2206 // CHECK2-NEXT:    [[TMP12:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
2207 // CHECK2-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP12]] to i32
2208 // CHECK2-NEXT:    [[SUB:%.*]] = sub i32 57, [[CONV3]]
2209 // CHECK2-NEXT:    [[ADD:%.*]] = add i32 [[SUB]], 1
2210 // CHECK2-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
2211 // CHECK2-NEXT:    [[CONV4:%.*]] = zext i32 [[DIV]] to i64
2212 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV4]], 11
2213 // CHECK2-NEXT:    [[SUB5:%.*]] = sub nsw i64 [[MUL]], 1
2214 // CHECK2-NEXT:    store i64 [[SUB5]], i64* [[DOTCAPTURE_EXPR_2]], align 8
2215 // CHECK2-NEXT:    [[TMP13:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
2216 // CHECK2-NEXT:    store i8 [[TMP13]], i8* [[I]], align 1
2217 // CHECK2-NEXT:    store i32 11, i32* [[X]], align 4
2218 // CHECK2-NEXT:    [[TMP14:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
2219 // CHECK2-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP14]] to i32
2220 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[CONV6]], 57
2221 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2222 // CHECK2:       omp.precond.then:
2223 // CHECK2-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
2224 // CHECK2-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
2225 // CHECK2-NEXT:    store i64 [[TMP15]], i64* [[DOTOMP_UB]], align 8
2226 // CHECK2-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
2227 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2228 // CHECK2-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
2229 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2230 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
2231 // CHECK2-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB2]], i32 [[TMP18]], i32 38, i64 0, i64 [[TMP16]], i64 1, i64 1)
2232 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
2233 // CHECK2:       omp.dispatch.cond:
2234 // CHECK2-NEXT:    [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2235 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
2236 // CHECK2-NEXT:    [[TMP21:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB2]], i32 [[TMP20]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
2237 // CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP21]], 0
2238 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2239 // CHECK2:       omp.dispatch.body:
2240 // CHECK2-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
2241 // CHECK2-NEXT:    store i64 [[TMP22]], i64* [[DOTOMP_IV]], align 8
2242 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2243 // CHECK2:       omp.inner.for.cond:
2244 // CHECK2-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
2245 // CHECK2-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !11
2246 // CHECK2-NEXT:    [[CMP9:%.*]] = icmp sle i64 [[TMP23]], [[TMP24]]
2247 // CHECK2-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2248 // CHECK2:       omp.inner.for.body:
2249 // CHECK2-NEXT:    [[TMP25:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !11
2250 // CHECK2-NEXT:    [[CONV10:%.*]] = sext i8 [[TMP25]] to i64
2251 // CHECK2-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
2252 // CHECK2-NEXT:    [[DIV11:%.*]] = sdiv i64 [[TMP26]], 11
2253 // CHECK2-NEXT:    [[MUL12:%.*]] = mul nsw i64 [[DIV11]], 1
2254 // CHECK2-NEXT:    [[ADD13:%.*]] = add nsw i64 [[CONV10]], [[MUL12]]
2255 // CHECK2-NEXT:    [[CONV14:%.*]] = trunc i64 [[ADD13]] to i8
2256 // CHECK2-NEXT:    store i8 [[CONV14]], i8* [[I7]], align 1, !llvm.access.group !11
2257 // CHECK2-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
2258 // CHECK2-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
2259 // CHECK2-NEXT:    [[DIV15:%.*]] = sdiv i64 [[TMP28]], 11
2260 // CHECK2-NEXT:    [[MUL16:%.*]] = mul nsw i64 [[DIV15]], 11
2261 // CHECK2-NEXT:    [[SUB17:%.*]] = sub nsw i64 [[TMP27]], [[MUL16]]
2262 // CHECK2-NEXT:    [[MUL18:%.*]] = mul nsw i64 [[SUB17]], 1
2263 // CHECK2-NEXT:    [[SUB19:%.*]] = sub nsw i64 11, [[MUL18]]
2264 // CHECK2-NEXT:    [[CONV20:%.*]] = trunc i64 [[SUB19]] to i32
2265 // CHECK2-NEXT:    store i32 [[CONV20]], i32* [[X8]], align 4, !llvm.access.group !11
2266 // CHECK2-NEXT:    [[TMP29:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !11
2267 // CHECK2-NEXT:    [[TMP30:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
2268 // CHECK2-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP30]] to i64
2269 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP29]], i64 [[IDXPROM]]
2270 // CHECK2-NEXT:    [[TMP31:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !11
2271 // CHECK2-NEXT:    [[TMP32:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !11
2272 // CHECK2-NEXT:    [[TMP33:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
2273 // CHECK2-NEXT:    [[IDXPROM21:%.*]] = sext i8 [[TMP33]] to i64
2274 // CHECK2-NEXT:    [[ARRAYIDX22:%.*]] = getelementptr inbounds float, float* [[TMP32]], i64 [[IDXPROM21]]
2275 // CHECK2-NEXT:    [[TMP34:%.*]] = load float, float* [[ARRAYIDX22]], align 4, !llvm.access.group !11
2276 // CHECK2-NEXT:    [[MUL23:%.*]] = fmul float [[TMP31]], [[TMP34]]
2277 // CHECK2-NEXT:    [[TMP35:%.*]] = load float*, float** [[TMP10]], align 8, !llvm.access.group !11
2278 // CHECK2-NEXT:    [[TMP36:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
2279 // CHECK2-NEXT:    [[IDXPROM24:%.*]] = sext i8 [[TMP36]] to i64
2280 // CHECK2-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds float, float* [[TMP35]], i64 [[IDXPROM24]]
2281 // CHECK2-NEXT:    [[TMP37:%.*]] = load float, float* [[ARRAYIDX25]], align 4, !llvm.access.group !11
2282 // CHECK2-NEXT:    [[MUL26:%.*]] = fmul float [[MUL23]], [[TMP37]]
2283 // CHECK2-NEXT:    [[TMP38:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !11
2284 // CHECK2-NEXT:    [[TMP39:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
2285 // CHECK2-NEXT:    [[IDXPROM27:%.*]] = sext i8 [[TMP39]] to i64
2286 // CHECK2-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds float, float* [[TMP38]], i64 [[IDXPROM27]]
2287 // CHECK2-NEXT:    store float [[MUL26]], float* [[ARRAYIDX28]], align 4, !llvm.access.group !11
2288 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2289 // CHECK2:       omp.body.continue:
2290 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2291 // CHECK2:       omp.inner.for.inc:
2292 // CHECK2-NEXT:    [[TMP40:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
2293 // CHECK2-NEXT:    [[ADD29:%.*]] = add nsw i64 [[TMP40]], 1
2294 // CHECK2-NEXT:    store i64 [[ADD29]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
2295 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
2296 // CHECK2:       omp.inner.for.end:
2297 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
2298 // CHECK2:       omp.dispatch.inc:
2299 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
2300 // CHECK2:       omp.dispatch.end:
2301 // CHECK2-NEXT:    br label [[OMP_PRECOND_END]]
2302 // CHECK2:       omp.precond.end:
2303 // CHECK2-NEXT:    ret void
2304 //
2305 //
2306 // CHECK2-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
2307 // CHECK2-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
2308 // CHECK2-NEXT:  entry:
2309 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
2310 // CHECK2-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
2311 // CHECK2-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
2312 // CHECK2-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
2313 // CHECK2-NEXT:    [[X:%.*]] = alloca i32, align 4
2314 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 8
2315 // CHECK2-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
2316 // CHECK2-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
2317 // CHECK2-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
2318 // CHECK2-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
2319 // CHECK2-NEXT:    store i32 0, i32* [[X]], align 4
2320 // CHECK2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2321 // CHECK2-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
2322 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2323 // CHECK2-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
2324 // CHECK2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
2325 // CHECK2-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
2326 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
2327 // CHECK2-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
2328 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.6*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
2329 // CHECK2-NEXT:    ret void
2330 //
2331 //
2332 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..7
2333 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
2334 // CHECK2-NEXT:  entry:
2335 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2336 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2337 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 8
2338 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2339 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i8, align 1
2340 // CHECK2-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
2341 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2342 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2343 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2344 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2345 // CHECK2-NEXT:    [[I:%.*]] = alloca i8, align 1
2346 // CHECK2-NEXT:    [[X:%.*]] = alloca i32, align 4
2347 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2348 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2349 // CHECK2-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 8
2350 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 8
2351 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
2352 // CHECK2-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
2353 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
2354 // CHECK2-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
2355 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
2356 // CHECK2-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
2357 // CHECK2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
2358 // CHECK2-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
2359 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2360 // CHECK2-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
2361 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2362 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2363 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2364 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
2365 // CHECK2-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 37, i32 0, i32 199, i32 1, i32 1)
2366 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
2367 // CHECK2:       omp.dispatch.cond:
2368 // CHECK2-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
2369 // CHECK2-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
2370 // CHECK2-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2371 // CHECK2:       omp.dispatch.body:
2372 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2373 // CHECK2-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
2374 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2375 // CHECK2:       omp.inner.for.cond:
2376 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
2377 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
2378 // CHECK2-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
2379 // CHECK2-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2380 // CHECK2:       omp.inner.for.body:
2381 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
2382 // CHECK2-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP15]], 20
2383 // CHECK2-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
2384 // CHECK2-NEXT:    [[ADD:%.*]] = add nsw i32 48, [[MUL]]
2385 // CHECK2-NEXT:    [[CONV:%.*]] = trunc i32 [[ADD]] to i8
2386 // CHECK2-NEXT:    store i8 [[CONV]], i8* [[I]], align 1, !llvm.access.group !14
2387 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
2388 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
2389 // CHECK2-NEXT:    [[DIV2:%.*]] = sdiv i32 [[TMP17]], 20
2390 // CHECK2-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[DIV2]], 20
2391 // CHECK2-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP16]], [[MUL3]]
2392 // CHECK2-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[SUB]], 1
2393 // CHECK2-NEXT:    [[ADD5:%.*]] = add nsw i32 -10, [[MUL4]]
2394 // CHECK2-NEXT:    store i32 [[ADD5]], i32* [[X]], align 4, !llvm.access.group !14
2395 // CHECK2-NEXT:    [[TMP18:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !14
2396 // CHECK2-NEXT:    [[TMP19:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
2397 // CHECK2-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP19]] to i64
2398 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP18]], i64 [[IDXPROM]]
2399 // CHECK2-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !14
2400 // CHECK2-NEXT:    [[TMP21:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !14
2401 // CHECK2-NEXT:    [[TMP22:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
2402 // CHECK2-NEXT:    [[IDXPROM6:%.*]] = zext i8 [[TMP22]] to i64
2403 // CHECK2-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP21]], i64 [[IDXPROM6]]
2404 // CHECK2-NEXT:    [[TMP23:%.*]] = load float, float* [[ARRAYIDX7]], align 4, !llvm.access.group !14
2405 // CHECK2-NEXT:    [[MUL8:%.*]] = fmul float [[TMP20]], [[TMP23]]
2406 // CHECK2-NEXT:    [[TMP24:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !14
2407 // CHECK2-NEXT:    [[TMP25:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
2408 // CHECK2-NEXT:    [[IDXPROM9:%.*]] = zext i8 [[TMP25]] to i64
2409 // CHECK2-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP24]], i64 [[IDXPROM9]]
2410 // CHECK2-NEXT:    [[TMP26:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
2411 // CHECK2-NEXT:    [[MUL11:%.*]] = fmul float [[MUL8]], [[TMP26]]
2412 // CHECK2-NEXT:    [[TMP27:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !14
2413 // CHECK2-NEXT:    [[TMP28:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
2414 // CHECK2-NEXT:    [[IDXPROM12:%.*]] = zext i8 [[TMP28]] to i64
2415 // CHECK2-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds float, float* [[TMP27]], i64 [[IDXPROM12]]
2416 // CHECK2-NEXT:    store float [[MUL11]], float* [[ARRAYIDX13]], align 4, !llvm.access.group !14
2417 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2418 // CHECK2:       omp.body.continue:
2419 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2420 // CHECK2:       omp.inner.for.inc:
2421 // CHECK2-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
2422 // CHECK2-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP29]], 1
2423 // CHECK2-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
2424 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
2425 // CHECK2:       omp.inner.for.end:
2426 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
2427 // CHECK2:       omp.dispatch.inc:
2428 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
2429 // CHECK2:       omp.dispatch.end:
2430 // CHECK2-NEXT:    ret void
2431 //
2432 //
2433 // CHECK2-LABEL: define {{[^@]+}}@_Z3foov
2434 // CHECK2-SAME: () #[[ATTR3:[0-9]+]] {
2435 // CHECK2-NEXT:  entry:
2436 // CHECK2-NEXT:    call void @_Z8mayThrowv()
2437 // CHECK2-NEXT:    ret i32 0
2438 //
2439 //
2440 // CHECK2-LABEL: define {{[^@]+}}@_Z12parallel_forPfi
2441 // CHECK2-SAME: (float* [[A:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
2442 // CHECK2-NEXT:  entry:
2443 // CHECK2-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
2444 // CHECK2-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
2445 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
2446 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
2447 // CHECK2-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 8
2448 // CHECK2-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
2449 // CHECK2-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
2450 // CHECK2-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
2451 // CHECK2-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
2452 // CHECK2-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
2453 // CHECK2-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
2454 // CHECK2-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP1]], align 16
2455 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
2456 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2457 // CHECK2-NEXT:    store float** [[A_ADDR]], float*** [[TMP3]], align 8
2458 // CHECK2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2459 // CHECK2-NEXT:    store i64 [[TMP1]], i64* [[TMP4]], align 8
2460 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
2461 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
2462 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
2463 // CHECK2-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.7*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
2464 // CHECK2-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
2465 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP7]])
2466 // CHECK2-NEXT:    ret void
2467 //
2468 //
2469 // CHECK2-LABEL: define {{[^@]+}}@.omp_outlined..8
2470 // CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {
2471 // CHECK2-NEXT:  entry:
2472 // CHECK2-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2473 // CHECK2-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2474 // CHECK2-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 8
2475 // CHECK2-NEXT:    [[N:%.*]] = alloca i32, align 4
2476 // CHECK2-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2477 // CHECK2-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2478 // CHECK2-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2479 // CHECK2-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2480 // CHECK2-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2481 // CHECK2-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2482 // CHECK2-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
2483 // CHECK2-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
2484 // CHECK2-NEXT:    [[I:%.*]] = alloca i32, align 4
2485 // CHECK2-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2486 // CHECK2-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2487 // CHECK2-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 8
2488 // CHECK2-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 8
2489 // CHECK2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
2490 // CHECK2-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
2491 // CHECK2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
2492 // CHECK2-NEXT:    [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8
2493 // CHECK2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
2494 // CHECK2-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 8
2495 // CHECK2-NEXT:    store i32 [[TMP6]], i32* [[N]], align 4
2496 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2497 // CHECK2-NEXT:    store i32 16908288, i32* [[DOTOMP_UB]], align 4
2498 // CHECK2-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2499 // CHECK2-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2500 // CHECK2-NEXT:    [[TMP7:%.*]] = call i8* @llvm.stacksave()
2501 // CHECK2-NEXT:    store i8* [[TMP7]], i8** [[SAVED_STACK]], align 8
2502 // CHECK2-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP4]], align 16
2503 // CHECK2-NEXT:    store i64 [[TMP4]], i64* [[__VLA_EXPR0]], align 8
2504 // CHECK2-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2505 // CHECK2-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
2506 // CHECK2-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 5)
2507 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
2508 // CHECK2:       omp.dispatch.cond:
2509 // CHECK2-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2510 // CHECK2-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP10]], 16908288
2511 // CHECK2-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2512 // CHECK2:       cond.true:
2513 // CHECK2-NEXT:    br label [[COND_END:%.*]]
2514 // CHECK2:       cond.false:
2515 // CHECK2-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2516 // CHECK2-NEXT:    br label [[COND_END]]
2517 // CHECK2:       cond.end:
2518 // CHECK2-NEXT:    [[COND:%.*]] = phi i32 [ 16908288, [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
2519 // CHECK2-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2520 // CHECK2-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2521 // CHECK2-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
2522 // CHECK2-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2523 // CHECK2-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2524 // CHECK2-NEXT:    [[CMP1:%.*]] = icmp ule i32 [[TMP13]], [[TMP14]]
2525 // CHECK2-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_CLEANUP:%.*]]
2526 // CHECK2:       omp.dispatch.cleanup:
2527 // CHECK2-NEXT:    br label [[OMP_DISPATCH_END:%.*]]
2528 // CHECK2:       omp.dispatch.body:
2529 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2530 // CHECK2:       omp.inner.for.cond:
2531 // CHECK2-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2532 // CHECK2-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2533 // CHECK2-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[TMP15]], [[TMP16]]
2534 // CHECK2-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
2535 // CHECK2:       omp.inner.for.cond.cleanup:
2536 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
2537 // CHECK2:       omp.inner.for.body:
2538 // CHECK2-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2539 // CHECK2-NEXT:    [[MUL:%.*]] = mul i32 [[TMP17]], 127
2540 // CHECK2-NEXT:    [[ADD:%.*]] = add i32 131071, [[MUL]]
2541 // CHECK2-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
2542 // CHECK2-NEXT:    [[CALL:%.*]] = invoke i32 @_Z3foov()
2543 // CHECK2-NEXT:    to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
2544 // CHECK2:       invoke.cont:
2545 // CHECK2-NEXT:    [[CONV:%.*]] = sitofp i32 [[CALL]] to float
2546 // CHECK2-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
2547 // CHECK2-NEXT:    [[IDXPROM:%.*]] = zext i32 [[TMP18]] to i64
2548 // CHECK2-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[VLA]], i64 [[IDXPROM]]
2549 // CHECK2-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
2550 // CHECK2-NEXT:    [[ADD3:%.*]] = fadd float [[CONV]], [[TMP19]]
2551 // CHECK2-NEXT:    [[TMP20:%.*]] = load i32, i32* [[N]], align 4
2552 // CHECK2-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP20]] to float
2553 // CHECK2-NEXT:    [[ADD5:%.*]] = fadd float [[ADD3]], [[CONV4]]
2554 // CHECK2-NEXT:    [[TMP21:%.*]] = load float*, float** [[TMP2]], align 8
2555 // CHECK2-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I]], align 4
2556 // CHECK2-NEXT:    [[IDXPROM6:%.*]] = zext i32 [[TMP22]] to i64
2557 // CHECK2-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP21]], i64 [[IDXPROM6]]
2558 // CHECK2-NEXT:    [[TMP23:%.*]] = load float, float* [[ARRAYIDX7]], align 4
2559 // CHECK2-NEXT:    [[ADD8:%.*]] = fadd float [[TMP23]], [[ADD5]]
2560 // CHECK2-NEXT:    store float [[ADD8]], float* [[ARRAYIDX7]], align 4
2561 // CHECK2-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2562 // CHECK2:       omp.body.continue:
2563 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2564 // CHECK2:       omp.inner.for.inc:
2565 // CHECK2-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2566 // CHECK2-NEXT:    [[ADD9:%.*]] = add i32 [[TMP24]], 1
2567 // CHECK2-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
2568 // CHECK2-NEXT:    br label [[OMP_INNER_FOR_COND]]
2569 // CHECK2:       omp.inner.for.end:
2570 // CHECK2-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
2571 // CHECK2:       omp.dispatch.inc:
2572 // CHECK2-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2573 // CHECK2-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2574 // CHECK2-NEXT:    [[ADD10:%.*]] = add i32 [[TMP25]], [[TMP26]]
2575 // CHECK2-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_LB]], align 4
2576 // CHECK2-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2577 // CHECK2-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
2578 // CHECK2-NEXT:    [[ADD11:%.*]] = add i32 [[TMP27]], [[TMP28]]
2579 // CHECK2-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_UB]], align 4
2580 // CHECK2-NEXT:    br label [[OMP_DISPATCH_COND]]
2581 // CHECK2:       omp.dispatch.end:
2582 // CHECK2-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]])
2583 // CHECK2-NEXT:    [[TMP29:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
2584 // CHECK2-NEXT:    call void @llvm.stackrestore(i8* [[TMP29]])
2585 // CHECK2-NEXT:    ret void
2586 // CHECK2:       terminate.lpad:
2587 // CHECK2-NEXT:    [[TMP30:%.*]] = landingpad { i8*, i32 }
2588 // CHECK2-NEXT:    catch i8* null
2589 // CHECK2-NEXT:    [[TMP31:%.*]] = extractvalue { i8*, i32 } [[TMP30]], 0
2590 // CHECK2-NEXT:    call void @__clang_call_terminate(i8* [[TMP31]]) #[[ATTR7:[0-9]+]]
2591 // CHECK2-NEXT:    unreachable
2592 //
2593 //
2594 // CHECK2-LABEL: define {{[^@]+}}@__clang_call_terminate
2595 // CHECK2-SAME: (i8* [[TMP0:%.*]]) #[[ATTR6:[0-9]+]] comdat {
2596 // CHECK2-NEXT:    [[TMP2:%.*]] = call i8* @__cxa_begin_catch(i8* [[TMP0]]) #[[ATTR2:[0-9]+]]
2597 // CHECK2-NEXT:    call void @_ZSt9terminatev() #[[ATTR7]]
2598 // CHECK2-NEXT:    unreachable
2599 //
2600 //
2601 // CHECK3-LABEL: define {{[^@]+}}@_Z17with_var_schedulev
2602 // CHECK3-SAME: () #[[ATTR0:[0-9]+]] {
2603 // CHECK3-NEXT:  entry:
2604 // CHECK3-NEXT:    [[A:%.*]] = alloca double, align 8
2605 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
2606 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
2607 // CHECK3-NEXT:    store double 5.000000e+00, double* [[A]], align 8
2608 // CHECK3-NEXT:    [[TMP0:%.*]] = load double, double* [[A]], align 8
2609 // CHECK3-NEXT:    [[CONV:%.*]] = fptosi double [[TMP0]] to i8
2610 // CHECK3-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
2611 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2612 // CHECK3-NEXT:    [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
2613 // CHECK3-NEXT:    store i8 [[TMP2]], i8* [[TMP1]], align 1
2614 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
2615 // CHECK3-NEXT:    ret void
2616 //
2617 //
2618 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
2619 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
2620 // CHECK3-NEXT:  entry:
2621 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2622 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2623 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
2624 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
2625 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
2626 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i64, align 8
2627 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca double, align 8
2628 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
2629 // CHECK3-NEXT:    [[I:%.*]] = alloca i64, align 8
2630 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
2631 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
2632 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
2633 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2634 // CHECK3-NEXT:    [[A:%.*]] = alloca double, align 8
2635 // CHECK3-NEXT:    [[I4:%.*]] = alloca i64, align 8
2636 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2637 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2638 // CHECK3-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
2639 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
2640 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
2641 // CHECK3-NEXT:    [[TMP2:%.*]] = load i8, i8* [[TMP1]], align 1
2642 // CHECK3-NEXT:    store i8 [[TMP2]], i8* [[DOTCAPTURE_EXPR_]], align 1
2643 // CHECK3-NEXT:    [[TMP3:%.*]] = load double, double* undef, align 8
2644 // CHECK3-NEXT:    [[ADD:%.*]] = fadd double 2.000000e+00, [[TMP3]]
2645 // CHECK3-NEXT:    store double [[ADD]], double* [[DOTCAPTURE_EXPR_1]], align 8
2646 // CHECK3-NEXT:    [[TMP4:%.*]] = load double, double* [[DOTCAPTURE_EXPR_1]], align 8
2647 // CHECK3-NEXT:    [[SUB:%.*]] = fsub double [[TMP4]], 1.000000e+00
2648 // CHECK3-NEXT:    [[DIV:%.*]] = fdiv double [[SUB]], 1.000000e+00
2649 // CHECK3-NEXT:    [[CONV:%.*]] = fptoui double [[DIV]] to i64
2650 // CHECK3-NEXT:    [[SUB3:%.*]] = sub i64 [[CONV]], 1
2651 // CHECK3-NEXT:    store i64 [[SUB3]], i64* [[DOTCAPTURE_EXPR_2]], align 8
2652 // CHECK3-NEXT:    store i64 1, i64* [[I]], align 8
2653 // CHECK3-NEXT:    [[TMP5:%.*]] = load double, double* [[DOTCAPTURE_EXPR_1]], align 8
2654 // CHECK3-NEXT:    [[CMP:%.*]] = fcmp olt double 1.000000e+00, [[TMP5]]
2655 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
2656 // CHECK3:       omp.precond.then:
2657 // CHECK3-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
2658 // CHECK3-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
2659 // CHECK3-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_UB]], align 8
2660 // CHECK3-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
2661 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2662 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
2663 // CHECK3-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP7]] to i64
2664 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2665 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
2666 // CHECK3-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 [[CONV5]])
2667 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
2668 // CHECK3:       omp.dispatch.cond:
2669 // CHECK3-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
2670 // CHECK3-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
2671 // CHECK3-NEXT:    [[CMP6:%.*]] = icmp ugt i64 [[TMP10]], [[TMP11]]
2672 // CHECK3-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2673 // CHECK3:       cond.true:
2674 // CHECK3-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
2675 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2676 // CHECK3:       cond.false:
2677 // CHECK3-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
2678 // CHECK3-NEXT:    br label [[COND_END]]
2679 // CHECK3:       cond.end:
2680 // CHECK3-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
2681 // CHECK3-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
2682 // CHECK3-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
2683 // CHECK3-NEXT:    store i64 [[TMP14]], i64* [[DOTOMP_IV]], align 8
2684 // CHECK3-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
2685 // CHECK3-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
2686 // CHECK3-NEXT:    [[ADD7:%.*]] = add i64 [[TMP16]], 1
2687 // CHECK3-NEXT:    [[CMP8:%.*]] = icmp ult i64 [[TMP15]], [[ADD7]]
2688 // CHECK3-NEXT:    br i1 [[CMP8]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
2689 // CHECK3:       omp.dispatch.body:
2690 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2691 // CHECK3:       omp.inner.for.cond:
2692 // CHECK3-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
2693 // CHECK3-NEXT:    [[TMP18:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
2694 // CHECK3-NEXT:    [[ADD9:%.*]] = add i64 [[TMP18]], 1
2695 // CHECK3-NEXT:    [[CMP10:%.*]] = icmp ult i64 [[TMP17]], [[ADD9]]
2696 // CHECK3-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2697 // CHECK3:       omp.inner.for.body:
2698 // CHECK3-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
2699 // CHECK3-NEXT:    [[MUL:%.*]] = mul i64 [[TMP19]], 1
2700 // CHECK3-NEXT:    [[ADD11:%.*]] = add i64 1, [[MUL]]
2701 // CHECK3-NEXT:    store i64 [[ADD11]], i64* [[I4]], align 8
2702 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2703 // CHECK3:       omp.body.continue:
2704 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2705 // CHECK3:       omp.inner.for.inc:
2706 // CHECK3-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
2707 // CHECK3-NEXT:    [[ADD12:%.*]] = add i64 [[TMP20]], 1
2708 // CHECK3-NEXT:    store i64 [[ADD12]], i64* [[DOTOMP_IV]], align 8
2709 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
2710 // CHECK3:       omp.inner.for.end:
2711 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
2712 // CHECK3:       omp.dispatch.inc:
2713 // CHECK3-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
2714 // CHECK3-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
2715 // CHECK3-NEXT:    [[ADD13:%.*]] = add i64 [[TMP21]], [[TMP22]]
2716 // CHECK3-NEXT:    store i64 [[ADD13]], i64* [[DOTOMP_LB]], align 8
2717 // CHECK3-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
2718 // CHECK3-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
2719 // CHECK3-NEXT:    [[ADD14:%.*]] = add i64 [[TMP23]], [[TMP24]]
2720 // CHECK3-NEXT:    store i64 [[ADD14]], i64* [[DOTOMP_UB]], align 8
2721 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
2722 // CHECK3:       omp.dispatch.end:
2723 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2724 // CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
2725 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
2726 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
2727 // CHECK3:       omp.precond.end:
2728 // CHECK3-NEXT:    ret void
2729 //
2730 //
2731 // CHECK3-LABEL: define {{[^@]+}}@_Z23without_schedule_clausePfS_S_S_
2732 // CHECK3-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
2733 // CHECK3-NEXT:  entry:
2734 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
2735 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
2736 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
2737 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
2738 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
2739 // CHECK3-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
2740 // CHECK3-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
2741 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
2742 // CHECK3-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
2743 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2744 // CHECK3-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
2745 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2746 // CHECK3-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
2747 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
2748 // CHECK3-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
2749 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
2750 // CHECK3-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
2751 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
2752 // CHECK3-NEXT:    ret void
2753 //
2754 //
2755 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..1
2756 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
2757 // CHECK3-NEXT:  entry:
2758 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2759 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2760 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
2761 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2762 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2763 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2764 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2765 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2766 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2767 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2768 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2769 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2770 // CHECK3-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
2771 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
2772 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
2773 // CHECK3-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
2774 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1
2775 // CHECK3-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
2776 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 2
2777 // CHECK3-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
2778 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 3
2779 // CHECK3-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
2780 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2781 // CHECK3-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
2782 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2783 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2784 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2785 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
2786 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2787 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2788 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 4571423
2789 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2790 // CHECK3:       cond.true:
2791 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2792 // CHECK3:       cond.false:
2793 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2794 // CHECK3-NEXT:    br label [[COND_END]]
2795 // CHECK3:       cond.end:
2796 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 4571423, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2797 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2798 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2799 // CHECK3-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
2800 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2801 // CHECK3:       omp.inner.for.cond:
2802 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2803 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2804 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2805 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2806 // CHECK3:       omp.inner.for.body:
2807 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2808 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 7
2809 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 33, [[MUL]]
2810 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
2811 // CHECK3-NEXT:    [[TMP17:%.*]] = load float*, float** [[TMP4]], align 8
2812 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
2813 // CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
2814 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM]]
2815 // CHECK3-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
2816 // CHECK3-NEXT:    [[TMP20:%.*]] = load float*, float** [[TMP6]], align 8
2817 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4
2818 // CHECK3-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP21]] to i64
2819 // CHECK3-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM2]]
2820 // CHECK3-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX3]], align 4
2821 // CHECK3-NEXT:    [[MUL4:%.*]] = fmul float [[TMP19]], [[TMP22]]
2822 // CHECK3-NEXT:    [[TMP23:%.*]] = load float*, float** [[TMP8]], align 8
2823 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4
2824 // CHECK3-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP24]] to i64
2825 // CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM5]]
2826 // CHECK3-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX6]], align 4
2827 // CHECK3-NEXT:    [[MUL7:%.*]] = fmul float [[MUL4]], [[TMP25]]
2828 // CHECK3-NEXT:    [[TMP26:%.*]] = load float*, float** [[TMP2]], align 8
2829 // CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I]], align 4
2830 // CHECK3-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP27]] to i64
2831 // CHECK3-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[TMP26]], i64 [[IDXPROM8]]
2832 // CHECK3-NEXT:    store float [[MUL7]], float* [[ARRAYIDX9]], align 4
2833 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2834 // CHECK3:       omp.body.continue:
2835 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2836 // CHECK3:       omp.inner.for.inc:
2837 // CHECK3-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2838 // CHECK3-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP28]], 1
2839 // CHECK3-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
2840 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
2841 // CHECK3:       omp.inner.for.end:
2842 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2843 // CHECK3:       omp.loop.exit:
2844 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
2845 // CHECK3-NEXT:    ret void
2846 //
2847 //
2848 // CHECK3-LABEL: define {{[^@]+}}@_Z18static_not_chunkedPfS_S_S_
2849 // CHECK3-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
2850 // CHECK3-NEXT:  entry:
2851 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
2852 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
2853 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
2854 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
2855 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 8
2856 // CHECK3-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
2857 // CHECK3-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
2858 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
2859 // CHECK3-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
2860 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2861 // CHECK3-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
2862 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2863 // CHECK3-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
2864 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
2865 // CHECK3-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
2866 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
2867 // CHECK3-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
2868 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
2869 // CHECK3-NEXT:    ret void
2870 //
2871 //
2872 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..2
2873 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
2874 // CHECK3-NEXT:  entry:
2875 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2876 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2877 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
2878 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2879 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2880 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2881 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2882 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
2883 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
2884 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
2885 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
2886 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
2887 // CHECK3-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
2888 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
2889 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
2890 // CHECK3-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
2891 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
2892 // CHECK3-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
2893 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 2
2894 // CHECK3-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
2895 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 3
2896 // CHECK3-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
2897 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
2898 // CHECK3-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
2899 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
2900 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
2901 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
2902 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
2903 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
2904 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2905 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 4571423
2906 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
2907 // CHECK3:       cond.true:
2908 // CHECK3-NEXT:    br label [[COND_END:%.*]]
2909 // CHECK3:       cond.false:
2910 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2911 // CHECK3-NEXT:    br label [[COND_END]]
2912 // CHECK3:       cond.end:
2913 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 4571423, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
2914 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
2915 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
2916 // CHECK3-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
2917 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
2918 // CHECK3:       omp.inner.for.cond:
2919 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2920 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
2921 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
2922 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
2923 // CHECK3:       omp.inner.for.body:
2924 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2925 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 7
2926 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
2927 // CHECK3-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
2928 // CHECK3-NEXT:    [[TMP17:%.*]] = load float*, float** [[TMP4]], align 8
2929 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
2930 // CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
2931 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM]]
2932 // CHECK3-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
2933 // CHECK3-NEXT:    [[TMP20:%.*]] = load float*, float** [[TMP6]], align 8
2934 // CHECK3-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4
2935 // CHECK3-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP21]] to i64
2936 // CHECK3-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM2]]
2937 // CHECK3-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX3]], align 4
2938 // CHECK3-NEXT:    [[MUL4:%.*]] = fmul float [[TMP19]], [[TMP22]]
2939 // CHECK3-NEXT:    [[TMP23:%.*]] = load float*, float** [[TMP8]], align 8
2940 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4
2941 // CHECK3-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP24]] to i64
2942 // CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM5]]
2943 // CHECK3-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX6]], align 4
2944 // CHECK3-NEXT:    [[MUL7:%.*]] = fmul float [[MUL4]], [[TMP25]]
2945 // CHECK3-NEXT:    [[TMP26:%.*]] = load float*, float** [[TMP2]], align 8
2946 // CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I]], align 4
2947 // CHECK3-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP27]] to i64
2948 // CHECK3-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[TMP26]], i64 [[IDXPROM8]]
2949 // CHECK3-NEXT:    store float [[MUL7]], float* [[ARRAYIDX9]], align 4
2950 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
2951 // CHECK3:       omp.body.continue:
2952 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
2953 // CHECK3:       omp.inner.for.inc:
2954 // CHECK3-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
2955 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
2956 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
2957 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
2958 // CHECK3:       omp.inner.for.end:
2959 // CHECK3-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
2960 // CHECK3:       omp.loop.exit:
2961 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
2962 // CHECK3-NEXT:    ret void
2963 //
2964 //
2965 // CHECK3-LABEL: define {{[^@]+}}@_Z14static_chunkedPfS_S_S_
2966 // CHECK3-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
2967 // CHECK3-NEXT:  entry:
2968 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
2969 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
2970 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
2971 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
2972 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
2973 // CHECK3-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
2974 // CHECK3-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
2975 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
2976 // CHECK3-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
2977 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
2978 // CHECK3-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
2979 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
2980 // CHECK3-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
2981 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
2982 // CHECK3-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
2983 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
2984 // CHECK3-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
2985 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
2986 // CHECK3-NEXT:    ret void
2987 //
2988 //
2989 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..3
2990 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
2991 // CHECK3-NEXT:  entry:
2992 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
2993 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
2994 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
2995 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
2996 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
2997 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
2998 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
2999 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3000 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3001 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3002 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3003 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3004 // CHECK3-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
3005 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
3006 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
3007 // CHECK3-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
3008 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
3009 // CHECK3-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
3010 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
3011 // CHECK3-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
3012 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
3013 // CHECK3-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
3014 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3015 // CHECK3-NEXT:    store i32 16908288, i32* [[DOTOMP_UB]], align 4
3016 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3017 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3018 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3019 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3020 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 5)
3021 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
3022 // CHECK3:       omp.dispatch.cond:
3023 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3024 // CHECK3-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP11]], 16908288
3025 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3026 // CHECK3:       cond.true:
3027 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3028 // CHECK3:       cond.false:
3029 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3030 // CHECK3-NEXT:    br label [[COND_END]]
3031 // CHECK3:       cond.end:
3032 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 16908288, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3033 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3034 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3035 // CHECK3-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
3036 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3037 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3038 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
3039 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
3040 // CHECK3:       omp.dispatch.body:
3041 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3042 // CHECK3:       omp.inner.for.cond:
3043 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3044 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3045 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[TMP16]], [[TMP17]]
3046 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3047 // CHECK3:       omp.inner.for.body:
3048 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3049 // CHECK3-NEXT:    [[MUL:%.*]] = mul i32 [[TMP18]], 127
3050 // CHECK3-NEXT:    [[ADD:%.*]] = add i32 131071, [[MUL]]
3051 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
3052 // CHECK3-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP4]], align 8
3053 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I]], align 4
3054 // CHECK3-NEXT:    [[IDXPROM:%.*]] = zext i32 [[TMP20]] to i64
3055 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[IDXPROM]]
3056 // CHECK3-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX]], align 4
3057 // CHECK3-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP6]], align 8
3058 // CHECK3-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
3059 // CHECK3-NEXT:    [[IDXPROM3:%.*]] = zext i32 [[TMP23]] to i64
3060 // CHECK3-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[IDXPROM3]]
3061 // CHECK3-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4
3062 // CHECK3-NEXT:    [[MUL5:%.*]] = fmul float [[TMP21]], [[TMP24]]
3063 // CHECK3-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP8]], align 8
3064 // CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[I]], align 4
3065 // CHECK3-NEXT:    [[IDXPROM6:%.*]] = zext i32 [[TMP26]] to i64
3066 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[IDXPROM6]]
3067 // CHECK3-NEXT:    [[TMP27:%.*]] = load float, float* [[ARRAYIDX7]], align 4
3068 // CHECK3-NEXT:    [[MUL8:%.*]] = fmul float [[MUL5]], [[TMP27]]
3069 // CHECK3-NEXT:    [[TMP28:%.*]] = load float*, float** [[TMP2]], align 8
3070 // CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[I]], align 4
3071 // CHECK3-NEXT:    [[IDXPROM9:%.*]] = zext i32 [[TMP29]] to i64
3072 // CHECK3-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP28]], i64 [[IDXPROM9]]
3073 // CHECK3-NEXT:    store float [[MUL8]], float* [[ARRAYIDX10]], align 4
3074 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3075 // CHECK3:       omp.body.continue:
3076 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3077 // CHECK3:       omp.inner.for.inc:
3078 // CHECK3-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3079 // CHECK3-NEXT:    [[ADD11:%.*]] = add i32 [[TMP30]], 1
3080 // CHECK3-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4
3081 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
3082 // CHECK3:       omp.inner.for.end:
3083 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
3084 // CHECK3:       omp.dispatch.inc:
3085 // CHECK3-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3086 // CHECK3-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3087 // CHECK3-NEXT:    [[ADD12:%.*]] = add i32 [[TMP31]], [[TMP32]]
3088 // CHECK3-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_LB]], align 4
3089 // CHECK3-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3090 // CHECK3-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3091 // CHECK3-NEXT:    [[ADD13:%.*]] = add i32 [[TMP33]], [[TMP34]]
3092 // CHECK3-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_UB]], align 4
3093 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
3094 // CHECK3:       omp.dispatch.end:
3095 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
3096 // CHECK3-NEXT:    ret void
3097 //
3098 //
3099 // CHECK3-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
3100 // CHECK3-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
3101 // CHECK3-NEXT:  entry:
3102 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
3103 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
3104 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
3105 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
3106 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 8
3107 // CHECK3-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
3108 // CHECK3-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
3109 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
3110 // CHECK3-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
3111 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3112 // CHECK3-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
3113 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3114 // CHECK3-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
3115 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
3116 // CHECK3-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
3117 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
3118 // CHECK3-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
3119 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
3120 // CHECK3-NEXT:    ret void
3121 //
3122 //
3123 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..4
3124 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3125 // CHECK3-NEXT:  entry:
3126 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3127 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3128 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
3129 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
3130 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i64, align 8
3131 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
3132 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
3133 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
3134 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3135 // CHECK3-NEXT:    [[I:%.*]] = alloca i64, align 8
3136 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3137 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3138 // CHECK3-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
3139 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
3140 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
3141 // CHECK3-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
3142 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
3143 // CHECK3-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
3144 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
3145 // CHECK3-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
3146 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
3147 // CHECK3-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
3148 // CHECK3-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
3149 // CHECK3-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
3150 // CHECK3-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
3151 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3152 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3153 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3154 // CHECK3-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 1073741859, i64 0, i64 16908287, i64 1, i64 1)
3155 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
3156 // CHECK3:       omp.dispatch.cond:
3157 // CHECK3-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
3158 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
3159 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
3160 // CHECK3:       omp.dispatch.body:
3161 // CHECK3-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
3162 // CHECK3-NEXT:    store i64 [[TMP12]], i64* [[DOTOMP_IV]], align 8
3163 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3164 // CHECK3:       omp.inner.for.cond:
3165 // CHECK3-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
3166 // CHECK3-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !5
3167 // CHECK3-NEXT:    [[ADD:%.*]] = add i64 [[TMP14]], 1
3168 // CHECK3-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP13]], [[ADD]]
3169 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3170 // CHECK3:       omp.inner.for.body:
3171 // CHECK3-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
3172 // CHECK3-NEXT:    [[MUL:%.*]] = mul i64 [[TMP15]], 127
3173 // CHECK3-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
3174 // CHECK3-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8, !llvm.access.group !5
3175 // CHECK3-NEXT:    [[TMP16:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !5
3176 // CHECK3-NEXT:    [[TMP17:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
3177 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[TMP17]]
3178 // CHECK3-NEXT:    [[TMP18:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !5
3179 // CHECK3-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !5
3180 // CHECK3-NEXT:    [[TMP20:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
3181 // CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[TMP20]]
3182 // CHECK3-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX2]], align 4, !llvm.access.group !5
3183 // CHECK3-NEXT:    [[MUL3:%.*]] = fmul float [[TMP18]], [[TMP21]]
3184 // CHECK3-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !5
3185 // CHECK3-NEXT:    [[TMP23:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
3186 // CHECK3-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[TMP23]]
3187 // CHECK3-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !llvm.access.group !5
3188 // CHECK3-NEXT:    [[MUL5:%.*]] = fmul float [[MUL3]], [[TMP24]]
3189 // CHECK3-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !5
3190 // CHECK3-NEXT:    [[TMP26:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
3191 // CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[TMP26]]
3192 // CHECK3-NEXT:    store float [[MUL5]], float* [[ARRAYIDX6]], align 4, !llvm.access.group !5
3193 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3194 // CHECK3:       omp.body.continue:
3195 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3196 // CHECK3:       omp.inner.for.inc:
3197 // CHECK3-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
3198 // CHECK3-NEXT:    [[ADD7:%.*]] = add i64 [[TMP27]], 1
3199 // CHECK3-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
3200 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
3201 // CHECK3:       omp.inner.for.end:
3202 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
3203 // CHECK3:       omp.dispatch.inc:
3204 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
3205 // CHECK3:       omp.dispatch.end:
3206 // CHECK3-NEXT:    ret void
3207 //
3208 //
3209 // CHECK3-LABEL: define {{[^@]+}}@_Z7guided7PfS_S_S_
3210 // CHECK3-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
3211 // CHECK3-NEXT:  entry:
3212 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
3213 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
3214 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
3215 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
3216 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
3217 // CHECK3-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
3218 // CHECK3-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
3219 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
3220 // CHECK3-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
3221 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3222 // CHECK3-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
3223 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3224 // CHECK3-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
3225 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
3226 // CHECK3-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
3227 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
3228 // CHECK3-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
3229 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
3230 // CHECK3-NEXT:    ret void
3231 //
3232 //
3233 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..5
3234 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3235 // CHECK3-NEXT:  entry:
3236 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3237 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3238 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
3239 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
3240 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i64, align 8
3241 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
3242 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
3243 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
3244 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3245 // CHECK3-NEXT:    [[I:%.*]] = alloca i64, align 8
3246 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3247 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3248 // CHECK3-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
3249 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
3250 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
3251 // CHECK3-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
3252 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
3253 // CHECK3-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
3254 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
3255 // CHECK3-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
3256 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
3257 // CHECK3-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
3258 // CHECK3-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
3259 // CHECK3-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
3260 // CHECK3-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
3261 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3262 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3263 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3264 // CHECK3-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 1073741860, i64 0, i64 16908287, i64 1, i64 7)
3265 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
3266 // CHECK3:       omp.dispatch.cond:
3267 // CHECK3-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
3268 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
3269 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
3270 // CHECK3:       omp.dispatch.body:
3271 // CHECK3-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
3272 // CHECK3-NEXT:    store i64 [[TMP12]], i64* [[DOTOMP_IV]], align 8
3273 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3274 // CHECK3:       omp.inner.for.cond:
3275 // CHECK3-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
3276 // CHECK3-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !8
3277 // CHECK3-NEXT:    [[ADD:%.*]] = add i64 [[TMP14]], 1
3278 // CHECK3-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP13]], [[ADD]]
3279 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3280 // CHECK3:       omp.inner.for.body:
3281 // CHECK3-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
3282 // CHECK3-NEXT:    [[MUL:%.*]] = mul i64 [[TMP15]], 127
3283 // CHECK3-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
3284 // CHECK3-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8, !llvm.access.group !8
3285 // CHECK3-NEXT:    [[TMP16:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !8
3286 // CHECK3-NEXT:    [[TMP17:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
3287 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[TMP17]]
3288 // CHECK3-NEXT:    [[TMP18:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !8
3289 // CHECK3-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !8
3290 // CHECK3-NEXT:    [[TMP20:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
3291 // CHECK3-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[TMP20]]
3292 // CHECK3-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX2]], align 4, !llvm.access.group !8
3293 // CHECK3-NEXT:    [[MUL3:%.*]] = fmul float [[TMP18]], [[TMP21]]
3294 // CHECK3-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !8
3295 // CHECK3-NEXT:    [[TMP23:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
3296 // CHECK3-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[TMP23]]
3297 // CHECK3-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !llvm.access.group !8
3298 // CHECK3-NEXT:    [[MUL5:%.*]] = fmul float [[MUL3]], [[TMP24]]
3299 // CHECK3-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !8
3300 // CHECK3-NEXT:    [[TMP26:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
3301 // CHECK3-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[TMP26]]
3302 // CHECK3-NEXT:    store float [[MUL5]], float* [[ARRAYIDX6]], align 4, !llvm.access.group !8
3303 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3304 // CHECK3:       omp.body.continue:
3305 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3306 // CHECK3:       omp.inner.for.inc:
3307 // CHECK3-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
3308 // CHECK3-NEXT:    [[ADD7:%.*]] = add i64 [[TMP27]], 1
3309 // CHECK3-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
3310 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
3311 // CHECK3:       omp.inner.for.end:
3312 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
3313 // CHECK3:       omp.dispatch.inc:
3314 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
3315 // CHECK3:       omp.dispatch.end:
3316 // CHECK3-NEXT:    ret void
3317 //
3318 //
3319 // CHECK3-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
3320 // CHECK3-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
3321 // CHECK3-NEXT:  entry:
3322 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
3323 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
3324 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
3325 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
3326 // CHECK3-NEXT:    [[X:%.*]] = alloca i32, align 4
3327 // CHECK3-NEXT:    [[Y:%.*]] = alloca i32, align 4
3328 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
3329 // CHECK3-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
3330 // CHECK3-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
3331 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
3332 // CHECK3-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
3333 // CHECK3-NEXT:    store i32 0, i32* [[X]], align 4
3334 // CHECK3-NEXT:    store i32 0, i32* [[Y]], align 4
3335 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3336 // CHECK3-NEXT:    store i32* [[Y]], i32** [[TMP0]], align 8
3337 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3338 // CHECK3-NEXT:    store float** [[A_ADDR]], float*** [[TMP1]], align 8
3339 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
3340 // CHECK3-NEXT:    store float** [[B_ADDR]], float*** [[TMP2]], align 8
3341 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
3342 // CHECK3-NEXT:    store float** [[C_ADDR]], float*** [[TMP3]], align 8
3343 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
3344 // CHECK3-NEXT:    store float** [[D_ADDR]], float*** [[TMP4]], align 8
3345 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
3346 // CHECK3-NEXT:    ret void
3347 //
3348 //
3349 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..6
3350 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3351 // CHECK3-NEXT:  entry:
3352 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3353 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3354 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
3355 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
3356 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i8, align 1
3357 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
3358 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
3359 // CHECK3-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
3360 // CHECK3-NEXT:    [[I:%.*]] = alloca i8, align 1
3361 // CHECK3-NEXT:    [[X:%.*]] = alloca i32, align 4
3362 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
3363 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
3364 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
3365 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3366 // CHECK3-NEXT:    [[I7:%.*]] = alloca i8, align 1
3367 // CHECK3-NEXT:    [[X8:%.*]] = alloca i32, align 4
3368 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3369 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3370 // CHECK3-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
3371 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
3372 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
3373 // CHECK3-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8
3374 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
3375 // CHECK3-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
3376 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
3377 // CHECK3-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
3378 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
3379 // CHECK3-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
3380 // CHECK3-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
3381 // CHECK3-NEXT:    [[TMP10:%.*]] = load float**, float*** [[TMP9]], align 8
3382 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP2]], align 4
3383 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP11]] to i8
3384 // CHECK3-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
3385 // CHECK3-NEXT:    [[TMP12:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
3386 // CHECK3-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP12]] to i32
3387 // CHECK3-NEXT:    [[SUB:%.*]] = sub i32 57, [[CONV3]]
3388 // CHECK3-NEXT:    [[ADD:%.*]] = add i32 [[SUB]], 1
3389 // CHECK3-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
3390 // CHECK3-NEXT:    [[CONV4:%.*]] = zext i32 [[DIV]] to i64
3391 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV4]], 11
3392 // CHECK3-NEXT:    [[SUB5:%.*]] = sub nsw i64 [[MUL]], 1
3393 // CHECK3-NEXT:    store i64 [[SUB5]], i64* [[DOTCAPTURE_EXPR_2]], align 8
3394 // CHECK3-NEXT:    [[TMP13:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
3395 // CHECK3-NEXT:    store i8 [[TMP13]], i8* [[I]], align 1
3396 // CHECK3-NEXT:    store i32 11, i32* [[X]], align 4
3397 // CHECK3-NEXT:    [[TMP14:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
3398 // CHECK3-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP14]] to i32
3399 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[CONV6]], 57
3400 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3401 // CHECK3:       omp.precond.then:
3402 // CHECK3-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
3403 // CHECK3-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
3404 // CHECK3-NEXT:    store i64 [[TMP15]], i64* [[DOTOMP_UB]], align 8
3405 // CHECK3-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
3406 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3407 // CHECK3-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
3408 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3409 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
3410 // CHECK3-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB2]], i32 [[TMP18]], i32 1073741862, i64 0, i64 [[TMP16]], i64 1, i64 1)
3411 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
3412 // CHECK3:       omp.dispatch.cond:
3413 // CHECK3-NEXT:    [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3414 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
3415 // CHECK3-NEXT:    [[TMP21:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB2]], i32 [[TMP20]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
3416 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP21]], 0
3417 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
3418 // CHECK3:       omp.dispatch.body:
3419 // CHECK3-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
3420 // CHECK3-NEXT:    store i64 [[TMP22]], i64* [[DOTOMP_IV]], align 8
3421 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3422 // CHECK3:       omp.inner.for.cond:
3423 // CHECK3-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
3424 // CHECK3-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !11
3425 // CHECK3-NEXT:    [[CMP9:%.*]] = icmp sle i64 [[TMP23]], [[TMP24]]
3426 // CHECK3-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3427 // CHECK3:       omp.inner.for.body:
3428 // CHECK3-NEXT:    [[TMP25:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !11
3429 // CHECK3-NEXT:    [[CONV10:%.*]] = sext i8 [[TMP25]] to i64
3430 // CHECK3-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
3431 // CHECK3-NEXT:    [[DIV11:%.*]] = sdiv i64 [[TMP26]], 11
3432 // CHECK3-NEXT:    [[MUL12:%.*]] = mul nsw i64 [[DIV11]], 1
3433 // CHECK3-NEXT:    [[ADD13:%.*]] = add nsw i64 [[CONV10]], [[MUL12]]
3434 // CHECK3-NEXT:    [[CONV14:%.*]] = trunc i64 [[ADD13]] to i8
3435 // CHECK3-NEXT:    store i8 [[CONV14]], i8* [[I7]], align 1, !llvm.access.group !11
3436 // CHECK3-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
3437 // CHECK3-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
3438 // CHECK3-NEXT:    [[DIV15:%.*]] = sdiv i64 [[TMP28]], 11
3439 // CHECK3-NEXT:    [[MUL16:%.*]] = mul nsw i64 [[DIV15]], 11
3440 // CHECK3-NEXT:    [[SUB17:%.*]] = sub nsw i64 [[TMP27]], [[MUL16]]
3441 // CHECK3-NEXT:    [[MUL18:%.*]] = mul nsw i64 [[SUB17]], 1
3442 // CHECK3-NEXT:    [[SUB19:%.*]] = sub nsw i64 11, [[MUL18]]
3443 // CHECK3-NEXT:    [[CONV20:%.*]] = trunc i64 [[SUB19]] to i32
3444 // CHECK3-NEXT:    store i32 [[CONV20]], i32* [[X8]], align 4, !llvm.access.group !11
3445 // CHECK3-NEXT:    [[TMP29:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !11
3446 // CHECK3-NEXT:    [[TMP30:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
3447 // CHECK3-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP30]] to i64
3448 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP29]], i64 [[IDXPROM]]
3449 // CHECK3-NEXT:    [[TMP31:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !11
3450 // CHECK3-NEXT:    [[TMP32:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !11
3451 // CHECK3-NEXT:    [[TMP33:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
3452 // CHECK3-NEXT:    [[IDXPROM21:%.*]] = sext i8 [[TMP33]] to i64
3453 // CHECK3-NEXT:    [[ARRAYIDX22:%.*]] = getelementptr inbounds float, float* [[TMP32]], i64 [[IDXPROM21]]
3454 // CHECK3-NEXT:    [[TMP34:%.*]] = load float, float* [[ARRAYIDX22]], align 4, !llvm.access.group !11
3455 // CHECK3-NEXT:    [[MUL23:%.*]] = fmul float [[TMP31]], [[TMP34]]
3456 // CHECK3-NEXT:    [[TMP35:%.*]] = load float*, float** [[TMP10]], align 8, !llvm.access.group !11
3457 // CHECK3-NEXT:    [[TMP36:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
3458 // CHECK3-NEXT:    [[IDXPROM24:%.*]] = sext i8 [[TMP36]] to i64
3459 // CHECK3-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds float, float* [[TMP35]], i64 [[IDXPROM24]]
3460 // CHECK3-NEXT:    [[TMP37:%.*]] = load float, float* [[ARRAYIDX25]], align 4, !llvm.access.group !11
3461 // CHECK3-NEXT:    [[MUL26:%.*]] = fmul float [[MUL23]], [[TMP37]]
3462 // CHECK3-NEXT:    [[TMP38:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !11
3463 // CHECK3-NEXT:    [[TMP39:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
3464 // CHECK3-NEXT:    [[IDXPROM27:%.*]] = sext i8 [[TMP39]] to i64
3465 // CHECK3-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds float, float* [[TMP38]], i64 [[IDXPROM27]]
3466 // CHECK3-NEXT:    store float [[MUL26]], float* [[ARRAYIDX28]], align 4, !llvm.access.group !11
3467 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3468 // CHECK3:       omp.body.continue:
3469 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3470 // CHECK3:       omp.inner.for.inc:
3471 // CHECK3-NEXT:    [[TMP40:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
3472 // CHECK3-NEXT:    [[ADD29:%.*]] = add nsw i64 [[TMP40]], 1
3473 // CHECK3-NEXT:    store i64 [[ADD29]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
3474 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
3475 // CHECK3:       omp.inner.for.end:
3476 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
3477 // CHECK3:       omp.dispatch.inc:
3478 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
3479 // CHECK3:       omp.dispatch.end:
3480 // CHECK3-NEXT:    br label [[OMP_PRECOND_END]]
3481 // CHECK3:       omp.precond.end:
3482 // CHECK3-NEXT:    ret void
3483 //
3484 //
3485 // CHECK3-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
3486 // CHECK3-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
3487 // CHECK3-NEXT:  entry:
3488 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
3489 // CHECK3-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
3490 // CHECK3-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
3491 // CHECK3-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
3492 // CHECK3-NEXT:    [[X:%.*]] = alloca i32, align 4
3493 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 8
3494 // CHECK3-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
3495 // CHECK3-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
3496 // CHECK3-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
3497 // CHECK3-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
3498 // CHECK3-NEXT:    store i32 0, i32* [[X]], align 4
3499 // CHECK3-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3500 // CHECK3-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
3501 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3502 // CHECK3-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
3503 // CHECK3-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
3504 // CHECK3-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
3505 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
3506 // CHECK3-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
3507 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.6*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
3508 // CHECK3-NEXT:    ret void
3509 //
3510 //
3511 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..7
3512 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3513 // CHECK3-NEXT:  entry:
3514 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3515 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3516 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 8
3517 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3518 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i8, align 1
3519 // CHECK3-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
3520 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3521 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3522 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3523 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3524 // CHECK3-NEXT:    [[I:%.*]] = alloca i8, align 1
3525 // CHECK3-NEXT:    [[X:%.*]] = alloca i32, align 4
3526 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3527 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3528 // CHECK3-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 8
3529 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 8
3530 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
3531 // CHECK3-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
3532 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
3533 // CHECK3-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
3534 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
3535 // CHECK3-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
3536 // CHECK3-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
3537 // CHECK3-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
3538 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3539 // CHECK3-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
3540 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3541 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3542 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3543 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3544 // CHECK3-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 1073741861, i32 0, i32 199, i32 1, i32 1)
3545 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
3546 // CHECK3:       omp.dispatch.cond:
3547 // CHECK3-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
3548 // CHECK3-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
3549 // CHECK3-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
3550 // CHECK3:       omp.dispatch.body:
3551 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3552 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
3553 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3554 // CHECK3:       omp.inner.for.cond:
3555 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3556 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
3557 // CHECK3-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
3558 // CHECK3-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3559 // CHECK3:       omp.inner.for.body:
3560 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3561 // CHECK3-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP15]], 20
3562 // CHECK3-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
3563 // CHECK3-NEXT:    [[ADD:%.*]] = add nsw i32 48, [[MUL]]
3564 // CHECK3-NEXT:    [[CONV:%.*]] = trunc i32 [[ADD]] to i8
3565 // CHECK3-NEXT:    store i8 [[CONV]], i8* [[I]], align 1, !llvm.access.group !14
3566 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3567 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3568 // CHECK3-NEXT:    [[DIV2:%.*]] = sdiv i32 [[TMP17]], 20
3569 // CHECK3-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[DIV2]], 20
3570 // CHECK3-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP16]], [[MUL3]]
3571 // CHECK3-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[SUB]], 1
3572 // CHECK3-NEXT:    [[ADD5:%.*]] = add nsw i32 -10, [[MUL4]]
3573 // CHECK3-NEXT:    store i32 [[ADD5]], i32* [[X]], align 4, !llvm.access.group !14
3574 // CHECK3-NEXT:    [[TMP18:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !14
3575 // CHECK3-NEXT:    [[TMP19:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
3576 // CHECK3-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP19]] to i64
3577 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP18]], i64 [[IDXPROM]]
3578 // CHECK3-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !14
3579 // CHECK3-NEXT:    [[TMP21:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !14
3580 // CHECK3-NEXT:    [[TMP22:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
3581 // CHECK3-NEXT:    [[IDXPROM6:%.*]] = zext i8 [[TMP22]] to i64
3582 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP21]], i64 [[IDXPROM6]]
3583 // CHECK3-NEXT:    [[TMP23:%.*]] = load float, float* [[ARRAYIDX7]], align 4, !llvm.access.group !14
3584 // CHECK3-NEXT:    [[MUL8:%.*]] = fmul float [[TMP20]], [[TMP23]]
3585 // CHECK3-NEXT:    [[TMP24:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !14
3586 // CHECK3-NEXT:    [[TMP25:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
3587 // CHECK3-NEXT:    [[IDXPROM9:%.*]] = zext i8 [[TMP25]] to i64
3588 // CHECK3-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP24]], i64 [[IDXPROM9]]
3589 // CHECK3-NEXT:    [[TMP26:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
3590 // CHECK3-NEXT:    [[MUL11:%.*]] = fmul float [[MUL8]], [[TMP26]]
3591 // CHECK3-NEXT:    [[TMP27:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !14
3592 // CHECK3-NEXT:    [[TMP28:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
3593 // CHECK3-NEXT:    [[IDXPROM12:%.*]] = zext i8 [[TMP28]] to i64
3594 // CHECK3-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds float, float* [[TMP27]], i64 [[IDXPROM12]]
3595 // CHECK3-NEXT:    store float [[MUL11]], float* [[ARRAYIDX13]], align 4, !llvm.access.group !14
3596 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3597 // CHECK3:       omp.body.continue:
3598 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3599 // CHECK3:       omp.inner.for.inc:
3600 // CHECK3-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3601 // CHECK3-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP29]], 1
3602 // CHECK3-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
3603 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
3604 // CHECK3:       omp.inner.for.end:
3605 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
3606 // CHECK3:       omp.dispatch.inc:
3607 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
3608 // CHECK3:       omp.dispatch.end:
3609 // CHECK3-NEXT:    ret void
3610 //
3611 //
3612 // CHECK3-LABEL: define {{[^@]+}}@_Z3foov
3613 // CHECK3-SAME: () #[[ATTR3:[0-9]+]] {
3614 // CHECK3-NEXT:  entry:
3615 // CHECK3-NEXT:    call void @_Z8mayThrowv()
3616 // CHECK3-NEXT:    ret i32 0
3617 //
3618 //
3619 // CHECK3-LABEL: define {{[^@]+}}@_Z12parallel_forPfi
3620 // CHECK3-SAME: (float* [[A:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
3621 // CHECK3-NEXT:  entry:
3622 // CHECK3-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
3623 // CHECK3-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
3624 // CHECK3-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
3625 // CHECK3-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
3626 // CHECK3-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 8
3627 // CHECK3-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
3628 // CHECK3-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
3629 // CHECK3-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
3630 // CHECK3-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
3631 // CHECK3-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
3632 // CHECK3-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
3633 // CHECK3-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP1]], align 16
3634 // CHECK3-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
3635 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3636 // CHECK3-NEXT:    store float** [[A_ADDR]], float*** [[TMP3]], align 8
3637 // CHECK3-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3638 // CHECK3-NEXT:    store i64 [[TMP1]], i64* [[TMP4]], align 8
3639 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
3640 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
3641 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
3642 // CHECK3-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.7*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
3643 // CHECK3-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
3644 // CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP7]])
3645 // CHECK3-NEXT:    ret void
3646 //
3647 //
3648 // CHECK3-LABEL: define {{[^@]+}}@.omp_outlined..8
3649 // CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {
3650 // CHECK3-NEXT:  entry:
3651 // CHECK3-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3652 // CHECK3-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3653 // CHECK3-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 8
3654 // CHECK3-NEXT:    [[N:%.*]] = alloca i32, align 4
3655 // CHECK3-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3656 // CHECK3-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3657 // CHECK3-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3658 // CHECK3-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3659 // CHECK3-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3660 // CHECK3-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3661 // CHECK3-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
3662 // CHECK3-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
3663 // CHECK3-NEXT:    [[I:%.*]] = alloca i32, align 4
3664 // CHECK3-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3665 // CHECK3-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3666 // CHECK3-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 8
3667 // CHECK3-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 8
3668 // CHECK3-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
3669 // CHECK3-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
3670 // CHECK3-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
3671 // CHECK3-NEXT:    [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8
3672 // CHECK3-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
3673 // CHECK3-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 8
3674 // CHECK3-NEXT:    store i32 [[TMP6]], i32* [[N]], align 4
3675 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3676 // CHECK3-NEXT:    store i32 16908288, i32* [[DOTOMP_UB]], align 4
3677 // CHECK3-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3678 // CHECK3-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3679 // CHECK3-NEXT:    [[TMP7:%.*]] = call i8* @llvm.stacksave()
3680 // CHECK3-NEXT:    store i8* [[TMP7]], i8** [[SAVED_STACK]], align 8
3681 // CHECK3-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP4]], align 16
3682 // CHECK3-NEXT:    store i64 [[TMP4]], i64* [[__VLA_EXPR0]], align 8
3683 // CHECK3-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3684 // CHECK3-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
3685 // CHECK3-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 5)
3686 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
3687 // CHECK3:       omp.dispatch.cond:
3688 // CHECK3-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3689 // CHECK3-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP10]], 16908288
3690 // CHECK3-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3691 // CHECK3:       cond.true:
3692 // CHECK3-NEXT:    br label [[COND_END:%.*]]
3693 // CHECK3:       cond.false:
3694 // CHECK3-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3695 // CHECK3-NEXT:    br label [[COND_END]]
3696 // CHECK3:       cond.end:
3697 // CHECK3-NEXT:    [[COND:%.*]] = phi i32 [ 16908288, [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
3698 // CHECK3-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3699 // CHECK3-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3700 // CHECK3-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
3701 // CHECK3-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3702 // CHECK3-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3703 // CHECK3-NEXT:    [[CMP1:%.*]] = icmp ule i32 [[TMP13]], [[TMP14]]
3704 // CHECK3-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_CLEANUP:%.*]]
3705 // CHECK3:       omp.dispatch.cleanup:
3706 // CHECK3-NEXT:    br label [[OMP_DISPATCH_END:%.*]]
3707 // CHECK3:       omp.dispatch.body:
3708 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3709 // CHECK3:       omp.inner.for.cond:
3710 // CHECK3-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3711 // CHECK3-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3712 // CHECK3-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[TMP15]], [[TMP16]]
3713 // CHECK3-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
3714 // CHECK3:       omp.inner.for.cond.cleanup:
3715 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
3716 // CHECK3:       omp.inner.for.body:
3717 // CHECK3-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3718 // CHECK3-NEXT:    [[MUL:%.*]] = mul i32 [[TMP17]], 127
3719 // CHECK3-NEXT:    [[ADD:%.*]] = add i32 131071, [[MUL]]
3720 // CHECK3-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
3721 // CHECK3-NEXT:    [[CALL:%.*]] = invoke i32 @_Z3foov()
3722 // CHECK3-NEXT:    to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
3723 // CHECK3:       invoke.cont:
3724 // CHECK3-NEXT:    [[CONV:%.*]] = sitofp i32 [[CALL]] to float
3725 // CHECK3-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
3726 // CHECK3-NEXT:    [[IDXPROM:%.*]] = zext i32 [[TMP18]] to i64
3727 // CHECK3-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[VLA]], i64 [[IDXPROM]]
3728 // CHECK3-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
3729 // CHECK3-NEXT:    [[ADD3:%.*]] = fadd float [[CONV]], [[TMP19]]
3730 // CHECK3-NEXT:    [[TMP20:%.*]] = load i32, i32* [[N]], align 4
3731 // CHECK3-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP20]] to float
3732 // CHECK3-NEXT:    [[ADD5:%.*]] = fadd float [[ADD3]], [[CONV4]]
3733 // CHECK3-NEXT:    [[TMP21:%.*]] = load float*, float** [[TMP2]], align 8
3734 // CHECK3-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I]], align 4
3735 // CHECK3-NEXT:    [[IDXPROM6:%.*]] = zext i32 [[TMP22]] to i64
3736 // CHECK3-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP21]], i64 [[IDXPROM6]]
3737 // CHECK3-NEXT:    [[TMP23:%.*]] = load float, float* [[ARRAYIDX7]], align 4
3738 // CHECK3-NEXT:    [[ADD8:%.*]] = fadd float [[TMP23]], [[ADD5]]
3739 // CHECK3-NEXT:    store float [[ADD8]], float* [[ARRAYIDX7]], align 4
3740 // CHECK3-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3741 // CHECK3:       omp.body.continue:
3742 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3743 // CHECK3:       omp.inner.for.inc:
3744 // CHECK3-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3745 // CHECK3-NEXT:    [[ADD9:%.*]] = add i32 [[TMP24]], 1
3746 // CHECK3-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
3747 // CHECK3-NEXT:    br label [[OMP_INNER_FOR_COND]]
3748 // CHECK3:       omp.inner.for.end:
3749 // CHECK3-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
3750 // CHECK3:       omp.dispatch.inc:
3751 // CHECK3-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3752 // CHECK3-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3753 // CHECK3-NEXT:    [[ADD10:%.*]] = add i32 [[TMP25]], [[TMP26]]
3754 // CHECK3-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_LB]], align 4
3755 // CHECK3-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3756 // CHECK3-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
3757 // CHECK3-NEXT:    [[ADD11:%.*]] = add i32 [[TMP27]], [[TMP28]]
3758 // CHECK3-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_UB]], align 4
3759 // CHECK3-NEXT:    br label [[OMP_DISPATCH_COND]]
3760 // CHECK3:       omp.dispatch.end:
3761 // CHECK3-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]])
3762 // CHECK3-NEXT:    [[TMP29:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
3763 // CHECK3-NEXT:    call void @llvm.stackrestore(i8* [[TMP29]])
3764 // CHECK3-NEXT:    ret void
3765 // CHECK3:       terminate.lpad:
3766 // CHECK3-NEXT:    [[TMP30:%.*]] = landingpad { i8*, i32 }
3767 // CHECK3-NEXT:    catch i8* null
3768 // CHECK3-NEXT:    [[TMP31:%.*]] = extractvalue { i8*, i32 } [[TMP30]], 0
3769 // CHECK3-NEXT:    call void @__clang_call_terminate(i8* [[TMP31]]) #[[ATTR7:[0-9]+]]
3770 // CHECK3-NEXT:    unreachable
3771 //
3772 //
3773 // CHECK3-LABEL: define {{[^@]+}}@__clang_call_terminate
3774 // CHECK3-SAME: (i8* [[TMP0:%.*]]) #[[ATTR6:[0-9]+]] comdat {
3775 // CHECK3-NEXT:    [[TMP2:%.*]] = call i8* @__cxa_begin_catch(i8* [[TMP0]]) #[[ATTR2:[0-9]+]]
3776 // CHECK3-NEXT:    call void @_ZSt9terminatev() #[[ATTR7]]
3777 // CHECK3-NEXT:    unreachable
3778 //
3779 //
3780 // CHECK4-LABEL: define {{[^@]+}}@_Z17with_var_schedulev
3781 // CHECK4-SAME: () #[[ATTR0:[0-9]+]] {
3782 // CHECK4-NEXT:  entry:
3783 // CHECK4-NEXT:    [[A:%.*]] = alloca double, align 8
3784 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
3785 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
3786 // CHECK4-NEXT:    store double 5.000000e+00, double* [[A]], align 8
3787 // CHECK4-NEXT:    [[TMP0:%.*]] = load double, double* [[A]], align 8
3788 // CHECK4-NEXT:    [[CONV:%.*]] = fptosi double [[TMP0]] to i8
3789 // CHECK4-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
3790 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3791 // CHECK4-NEXT:    [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
3792 // CHECK4-NEXT:    store i8 [[TMP2]], i8* [[TMP1]], align 1
3793 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
3794 // CHECK4-NEXT:    ret void
3795 //
3796 //
3797 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
3798 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
3799 // CHECK4-NEXT:  entry:
3800 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3801 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3802 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
3803 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
3804 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
3805 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i64, align 8
3806 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca double, align 8
3807 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
3808 // CHECK4-NEXT:    [[I:%.*]] = alloca i64, align 8
3809 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
3810 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
3811 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
3812 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3813 // CHECK4-NEXT:    [[A:%.*]] = alloca double, align 8
3814 // CHECK4-NEXT:    [[I4:%.*]] = alloca i64, align 8
3815 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3816 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3817 // CHECK4-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
3818 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
3819 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
3820 // CHECK4-NEXT:    [[TMP2:%.*]] = load i8, i8* [[TMP1]], align 1
3821 // CHECK4-NEXT:    store i8 [[TMP2]], i8* [[DOTCAPTURE_EXPR_]], align 1
3822 // CHECK4-NEXT:    [[TMP3:%.*]] = load double, double* undef, align 8
3823 // CHECK4-NEXT:    [[ADD:%.*]] = fadd double 2.000000e+00, [[TMP3]]
3824 // CHECK4-NEXT:    store double [[ADD]], double* [[DOTCAPTURE_EXPR_1]], align 8
3825 // CHECK4-NEXT:    [[TMP4:%.*]] = load double, double* [[DOTCAPTURE_EXPR_1]], align 8
3826 // CHECK4-NEXT:    [[SUB:%.*]] = fsub double [[TMP4]], 1.000000e+00
3827 // CHECK4-NEXT:    [[DIV:%.*]] = fdiv double [[SUB]], 1.000000e+00
3828 // CHECK4-NEXT:    [[CONV:%.*]] = fptoui double [[DIV]] to i64
3829 // CHECK4-NEXT:    [[SUB3:%.*]] = sub i64 [[CONV]], 1
3830 // CHECK4-NEXT:    store i64 [[SUB3]], i64* [[DOTCAPTURE_EXPR_2]], align 8
3831 // CHECK4-NEXT:    store i64 1, i64* [[I]], align 8
3832 // CHECK4-NEXT:    [[TMP5:%.*]] = load double, double* [[DOTCAPTURE_EXPR_1]], align 8
3833 // CHECK4-NEXT:    [[CMP:%.*]] = fcmp olt double 1.000000e+00, [[TMP5]]
3834 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
3835 // CHECK4:       omp.precond.then:
3836 // CHECK4-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
3837 // CHECK4-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
3838 // CHECK4-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_UB]], align 8
3839 // CHECK4-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
3840 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3841 // CHECK4-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
3842 // CHECK4-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP7]] to i64
3843 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3844 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
3845 // CHECK4-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 [[CONV5]])
3846 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
3847 // CHECK4:       omp.dispatch.cond:
3848 // CHECK4-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
3849 // CHECK4-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
3850 // CHECK4-NEXT:    [[CMP6:%.*]] = icmp ugt i64 [[TMP10]], [[TMP11]]
3851 // CHECK4-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3852 // CHECK4:       cond.true:
3853 // CHECK4-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
3854 // CHECK4-NEXT:    br label [[COND_END:%.*]]
3855 // CHECK4:       cond.false:
3856 // CHECK4-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
3857 // CHECK4-NEXT:    br label [[COND_END]]
3858 // CHECK4:       cond.end:
3859 // CHECK4-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
3860 // CHECK4-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
3861 // CHECK4-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
3862 // CHECK4-NEXT:    store i64 [[TMP14]], i64* [[DOTOMP_IV]], align 8
3863 // CHECK4-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
3864 // CHECK4-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
3865 // CHECK4-NEXT:    [[ADD7:%.*]] = add i64 [[TMP16]], 1
3866 // CHECK4-NEXT:    [[CMP8:%.*]] = icmp ult i64 [[TMP15]], [[ADD7]]
3867 // CHECK4-NEXT:    br i1 [[CMP8]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
3868 // CHECK4:       omp.dispatch.body:
3869 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3870 // CHECK4:       omp.inner.for.cond:
3871 // CHECK4-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
3872 // CHECK4-NEXT:    [[TMP18:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
3873 // CHECK4-NEXT:    [[ADD9:%.*]] = add i64 [[TMP18]], 1
3874 // CHECK4-NEXT:    [[CMP10:%.*]] = icmp ult i64 [[TMP17]], [[ADD9]]
3875 // CHECK4-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3876 // CHECK4:       omp.inner.for.body:
3877 // CHECK4-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
3878 // CHECK4-NEXT:    [[MUL:%.*]] = mul i64 [[TMP19]], 1
3879 // CHECK4-NEXT:    [[ADD11:%.*]] = add i64 1, [[MUL]]
3880 // CHECK4-NEXT:    store i64 [[ADD11]], i64* [[I4]], align 8
3881 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
3882 // CHECK4:       omp.body.continue:
3883 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
3884 // CHECK4:       omp.inner.for.inc:
3885 // CHECK4-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
3886 // CHECK4-NEXT:    [[ADD12:%.*]] = add i64 [[TMP20]], 1
3887 // CHECK4-NEXT:    store i64 [[ADD12]], i64* [[DOTOMP_IV]], align 8
3888 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
3889 // CHECK4:       omp.inner.for.end:
3890 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
3891 // CHECK4:       omp.dispatch.inc:
3892 // CHECK4-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
3893 // CHECK4-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
3894 // CHECK4-NEXT:    [[ADD13:%.*]] = add i64 [[TMP21]], [[TMP22]]
3895 // CHECK4-NEXT:    store i64 [[ADD13]], i64* [[DOTOMP_LB]], align 8
3896 // CHECK4-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
3897 // CHECK4-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
3898 // CHECK4-NEXT:    [[ADD14:%.*]] = add i64 [[TMP23]], [[TMP24]]
3899 // CHECK4-NEXT:    store i64 [[ADD14]], i64* [[DOTOMP_UB]], align 8
3900 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
3901 // CHECK4:       omp.dispatch.end:
3902 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3903 // CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
3904 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
3905 // CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
3906 // CHECK4:       omp.precond.end:
3907 // CHECK4-NEXT:    ret void
3908 //
3909 //
3910 // CHECK4-LABEL: define {{[^@]+}}@_Z23without_schedule_clausePfS_S_S_
3911 // CHECK4-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
3912 // CHECK4-NEXT:  entry:
3913 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
3914 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
3915 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
3916 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
3917 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
3918 // CHECK4-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
3919 // CHECK4-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
3920 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
3921 // CHECK4-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
3922 // CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
3923 // CHECK4-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
3924 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
3925 // CHECK4-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
3926 // CHECK4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
3927 // CHECK4-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
3928 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
3929 // CHECK4-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
3930 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
3931 // CHECK4-NEXT:    ret void
3932 //
3933 //
3934 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..1
3935 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
3936 // CHECK4-NEXT:  entry:
3937 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
3938 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
3939 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
3940 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
3941 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
3942 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
3943 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
3944 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
3945 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
3946 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
3947 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
3948 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
3949 // CHECK4-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
3950 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
3951 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
3952 // CHECK4-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
3953 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1
3954 // CHECK4-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
3955 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 2
3956 // CHECK4-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
3957 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 3
3958 // CHECK4-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
3959 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
3960 // CHECK4-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
3961 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
3962 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
3963 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
3964 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
3965 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
3966 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3967 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 4571423
3968 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
3969 // CHECK4:       cond.true:
3970 // CHECK4-NEXT:    br label [[COND_END:%.*]]
3971 // CHECK4:       cond.false:
3972 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3973 // CHECK4-NEXT:    br label [[COND_END]]
3974 // CHECK4:       cond.end:
3975 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 4571423, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
3976 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
3977 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
3978 // CHECK4-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
3979 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
3980 // CHECK4:       omp.inner.for.cond:
3981 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3982 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
3983 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
3984 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
3985 // CHECK4:       omp.inner.for.body:
3986 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
3987 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 7
3988 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 33, [[MUL]]
3989 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
3990 // CHECK4-NEXT:    [[TMP17:%.*]] = load float*, float** [[TMP4]], align 8
3991 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
3992 // CHECK4-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
3993 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM]]
3994 // CHECK4-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
3995 // CHECK4-NEXT:    [[TMP20:%.*]] = load float*, float** [[TMP6]], align 8
3996 // CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4
3997 // CHECK4-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP21]] to i64
3998 // CHECK4-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM2]]
3999 // CHECK4-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX3]], align 4
4000 // CHECK4-NEXT:    [[MUL4:%.*]] = fmul float [[TMP19]], [[TMP22]]
4001 // CHECK4-NEXT:    [[TMP23:%.*]] = load float*, float** [[TMP8]], align 8
4002 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4
4003 // CHECK4-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP24]] to i64
4004 // CHECK4-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM5]]
4005 // CHECK4-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX6]], align 4
4006 // CHECK4-NEXT:    [[MUL7:%.*]] = fmul float [[MUL4]], [[TMP25]]
4007 // CHECK4-NEXT:    [[TMP26:%.*]] = load float*, float** [[TMP2]], align 8
4008 // CHECK4-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I]], align 4
4009 // CHECK4-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP27]] to i64
4010 // CHECK4-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[TMP26]], i64 [[IDXPROM8]]
4011 // CHECK4-NEXT:    store float [[MUL7]], float* [[ARRAYIDX9]], align 4
4012 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4013 // CHECK4:       omp.body.continue:
4014 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4015 // CHECK4:       omp.inner.for.inc:
4016 // CHECK4-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4017 // CHECK4-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP28]], 1
4018 // CHECK4-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
4019 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
4020 // CHECK4:       omp.inner.for.end:
4021 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4022 // CHECK4:       omp.loop.exit:
4023 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
4024 // CHECK4-NEXT:    ret void
4025 //
4026 //
4027 // CHECK4-LABEL: define {{[^@]+}}@_Z18static_not_chunkedPfS_S_S_
4028 // CHECK4-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
4029 // CHECK4-NEXT:  entry:
4030 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
4031 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
4032 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
4033 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
4034 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 8
4035 // CHECK4-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
4036 // CHECK4-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
4037 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
4038 // CHECK4-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
4039 // CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4040 // CHECK4-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
4041 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4042 // CHECK4-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
4043 // CHECK4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4044 // CHECK4-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
4045 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
4046 // CHECK4-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
4047 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
4048 // CHECK4-NEXT:    ret void
4049 //
4050 //
4051 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..2
4052 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4053 // CHECK4-NEXT:  entry:
4054 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4055 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4056 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
4057 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4058 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4059 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4060 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4061 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4062 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4063 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
4064 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4065 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4066 // CHECK4-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
4067 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
4068 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
4069 // CHECK4-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
4070 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
4071 // CHECK4-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
4072 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 2
4073 // CHECK4-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
4074 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 3
4075 // CHECK4-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
4076 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4077 // CHECK4-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
4078 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4079 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4080 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4081 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
4082 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
4083 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4084 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 4571423
4085 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4086 // CHECK4:       cond.true:
4087 // CHECK4-NEXT:    br label [[COND_END:%.*]]
4088 // CHECK4:       cond.false:
4089 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4090 // CHECK4-NEXT:    br label [[COND_END]]
4091 // CHECK4:       cond.end:
4092 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 4571423, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
4093 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4094 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4095 // CHECK4-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
4096 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4097 // CHECK4:       omp.inner.for.cond:
4098 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4099 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4100 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
4101 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4102 // CHECK4:       omp.inner.for.body:
4103 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4104 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 7
4105 // CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
4106 // CHECK4-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
4107 // CHECK4-NEXT:    [[TMP17:%.*]] = load float*, float** [[TMP4]], align 8
4108 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
4109 // CHECK4-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
4110 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM]]
4111 // CHECK4-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
4112 // CHECK4-NEXT:    [[TMP20:%.*]] = load float*, float** [[TMP6]], align 8
4113 // CHECK4-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4
4114 // CHECK4-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP21]] to i64
4115 // CHECK4-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM2]]
4116 // CHECK4-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX3]], align 4
4117 // CHECK4-NEXT:    [[MUL4:%.*]] = fmul float [[TMP19]], [[TMP22]]
4118 // CHECK4-NEXT:    [[TMP23:%.*]] = load float*, float** [[TMP8]], align 8
4119 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4
4120 // CHECK4-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP24]] to i64
4121 // CHECK4-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM5]]
4122 // CHECK4-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX6]], align 4
4123 // CHECK4-NEXT:    [[MUL7:%.*]] = fmul float [[MUL4]], [[TMP25]]
4124 // CHECK4-NEXT:    [[TMP26:%.*]] = load float*, float** [[TMP2]], align 8
4125 // CHECK4-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I]], align 4
4126 // CHECK4-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP27]] to i64
4127 // CHECK4-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[TMP26]], i64 [[IDXPROM8]]
4128 // CHECK4-NEXT:    store float [[MUL7]], float* [[ARRAYIDX9]], align 4
4129 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4130 // CHECK4:       omp.body.continue:
4131 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4132 // CHECK4:       omp.inner.for.inc:
4133 // CHECK4-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4134 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
4135 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
4136 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
4137 // CHECK4:       omp.inner.for.end:
4138 // CHECK4-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
4139 // CHECK4:       omp.loop.exit:
4140 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
4141 // CHECK4-NEXT:    ret void
4142 //
4143 //
4144 // CHECK4-LABEL: define {{[^@]+}}@_Z14static_chunkedPfS_S_S_
4145 // CHECK4-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
4146 // CHECK4-NEXT:  entry:
4147 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
4148 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
4149 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
4150 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
4151 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
4152 // CHECK4-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
4153 // CHECK4-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
4154 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
4155 // CHECK4-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
4156 // CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4157 // CHECK4-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
4158 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4159 // CHECK4-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
4160 // CHECK4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4161 // CHECK4-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
4162 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
4163 // CHECK4-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
4164 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
4165 // CHECK4-NEXT:    ret void
4166 //
4167 //
4168 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..3
4169 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4170 // CHECK4-NEXT:  entry:
4171 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4172 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4173 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
4174 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4175 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4176 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4177 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4178 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4179 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4180 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
4181 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4182 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4183 // CHECK4-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
4184 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
4185 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
4186 // CHECK4-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
4187 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
4188 // CHECK4-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
4189 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
4190 // CHECK4-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
4191 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
4192 // CHECK4-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
4193 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4194 // CHECK4-NEXT:    store i32 16908288, i32* [[DOTOMP_UB]], align 4
4195 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4196 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4197 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4198 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
4199 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 5)
4200 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
4201 // CHECK4:       omp.dispatch.cond:
4202 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4203 // CHECK4-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP11]], 16908288
4204 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4205 // CHECK4:       cond.true:
4206 // CHECK4-NEXT:    br label [[COND_END:%.*]]
4207 // CHECK4:       cond.false:
4208 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4209 // CHECK4-NEXT:    br label [[COND_END]]
4210 // CHECK4:       cond.end:
4211 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 16908288, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
4212 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4213 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4214 // CHECK4-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
4215 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4216 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4217 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
4218 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
4219 // CHECK4:       omp.dispatch.body:
4220 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4221 // CHECK4:       omp.inner.for.cond:
4222 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4223 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4224 // CHECK4-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[TMP16]], [[TMP17]]
4225 // CHECK4-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4226 // CHECK4:       omp.inner.for.body:
4227 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4228 // CHECK4-NEXT:    [[MUL:%.*]] = mul i32 [[TMP18]], 127
4229 // CHECK4-NEXT:    [[ADD:%.*]] = add i32 131071, [[MUL]]
4230 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
4231 // CHECK4-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP4]], align 8
4232 // CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I]], align 4
4233 // CHECK4-NEXT:    [[IDXPROM:%.*]] = zext i32 [[TMP20]] to i64
4234 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[IDXPROM]]
4235 // CHECK4-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX]], align 4
4236 // CHECK4-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP6]], align 8
4237 // CHECK4-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
4238 // CHECK4-NEXT:    [[IDXPROM3:%.*]] = zext i32 [[TMP23]] to i64
4239 // CHECK4-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[IDXPROM3]]
4240 // CHECK4-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4
4241 // CHECK4-NEXT:    [[MUL5:%.*]] = fmul float [[TMP21]], [[TMP24]]
4242 // CHECK4-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP8]], align 8
4243 // CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[I]], align 4
4244 // CHECK4-NEXT:    [[IDXPROM6:%.*]] = zext i32 [[TMP26]] to i64
4245 // CHECK4-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[IDXPROM6]]
4246 // CHECK4-NEXT:    [[TMP27:%.*]] = load float, float* [[ARRAYIDX7]], align 4
4247 // CHECK4-NEXT:    [[MUL8:%.*]] = fmul float [[MUL5]], [[TMP27]]
4248 // CHECK4-NEXT:    [[TMP28:%.*]] = load float*, float** [[TMP2]], align 8
4249 // CHECK4-NEXT:    [[TMP29:%.*]] = load i32, i32* [[I]], align 4
4250 // CHECK4-NEXT:    [[IDXPROM9:%.*]] = zext i32 [[TMP29]] to i64
4251 // CHECK4-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP28]], i64 [[IDXPROM9]]
4252 // CHECK4-NEXT:    store float [[MUL8]], float* [[ARRAYIDX10]], align 4
4253 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4254 // CHECK4:       omp.body.continue:
4255 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4256 // CHECK4:       omp.inner.for.inc:
4257 // CHECK4-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4258 // CHECK4-NEXT:    [[ADD11:%.*]] = add i32 [[TMP30]], 1
4259 // CHECK4-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4
4260 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
4261 // CHECK4:       omp.inner.for.end:
4262 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
4263 // CHECK4:       omp.dispatch.inc:
4264 // CHECK4-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4265 // CHECK4-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4266 // CHECK4-NEXT:    [[ADD12:%.*]] = add i32 [[TMP31]], [[TMP32]]
4267 // CHECK4-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_LB]], align 4
4268 // CHECK4-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4269 // CHECK4-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4270 // CHECK4-NEXT:    [[ADD13:%.*]] = add i32 [[TMP33]], [[TMP34]]
4271 // CHECK4-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_UB]], align 4
4272 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
4273 // CHECK4:       omp.dispatch.end:
4274 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
4275 // CHECK4-NEXT:    ret void
4276 //
4277 //
4278 // CHECK4-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
4279 // CHECK4-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
4280 // CHECK4-NEXT:  entry:
4281 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
4282 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
4283 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
4284 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
4285 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 8
4286 // CHECK4-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
4287 // CHECK4-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
4288 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
4289 // CHECK4-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
4290 // CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4291 // CHECK4-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
4292 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4293 // CHECK4-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
4294 // CHECK4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4295 // CHECK4-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
4296 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
4297 // CHECK4-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
4298 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
4299 // CHECK4-NEXT:    ret void
4300 //
4301 //
4302 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..4
4303 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4304 // CHECK4-NEXT:  entry:
4305 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4306 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4307 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
4308 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
4309 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i64, align 8
4310 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
4311 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
4312 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
4313 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4314 // CHECK4-NEXT:    [[I:%.*]] = alloca i64, align 8
4315 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4316 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4317 // CHECK4-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
4318 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
4319 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
4320 // CHECK4-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
4321 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
4322 // CHECK4-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
4323 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
4324 // CHECK4-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
4325 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
4326 // CHECK4-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
4327 // CHECK4-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
4328 // CHECK4-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
4329 // CHECK4-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
4330 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4331 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4332 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
4333 // CHECK4-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 35, i64 0, i64 16908287, i64 1, i64 1)
4334 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
4335 // CHECK4:       omp.dispatch.cond:
4336 // CHECK4-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
4337 // CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
4338 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
4339 // CHECK4:       omp.dispatch.body:
4340 // CHECK4-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
4341 // CHECK4-NEXT:    store i64 [[TMP12]], i64* [[DOTOMP_IV]], align 8
4342 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4343 // CHECK4:       omp.inner.for.cond:
4344 // CHECK4-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
4345 // CHECK4-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !5
4346 // CHECK4-NEXT:    [[ADD:%.*]] = add i64 [[TMP14]], 1
4347 // CHECK4-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP13]], [[ADD]]
4348 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4349 // CHECK4:       omp.inner.for.body:
4350 // CHECK4-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
4351 // CHECK4-NEXT:    [[MUL:%.*]] = mul i64 [[TMP15]], 127
4352 // CHECK4-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
4353 // CHECK4-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8, !llvm.access.group !5
4354 // CHECK4-NEXT:    [[TMP16:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !5
4355 // CHECK4-NEXT:    [[TMP17:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
4356 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[TMP17]]
4357 // CHECK4-NEXT:    [[TMP18:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !5
4358 // CHECK4-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !5
4359 // CHECK4-NEXT:    [[TMP20:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
4360 // CHECK4-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[TMP20]]
4361 // CHECK4-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX2]], align 4, !llvm.access.group !5
4362 // CHECK4-NEXT:    [[MUL3:%.*]] = fmul float [[TMP18]], [[TMP21]]
4363 // CHECK4-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !5
4364 // CHECK4-NEXT:    [[TMP23:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
4365 // CHECK4-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[TMP23]]
4366 // CHECK4-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !llvm.access.group !5
4367 // CHECK4-NEXT:    [[MUL5:%.*]] = fmul float [[MUL3]], [[TMP24]]
4368 // CHECK4-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !5
4369 // CHECK4-NEXT:    [[TMP26:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
4370 // CHECK4-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[TMP26]]
4371 // CHECK4-NEXT:    store float [[MUL5]], float* [[ARRAYIDX6]], align 4, !llvm.access.group !5
4372 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4373 // CHECK4:       omp.body.continue:
4374 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4375 // CHECK4:       omp.inner.for.inc:
4376 // CHECK4-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
4377 // CHECK4-NEXT:    [[ADD7:%.*]] = add i64 [[TMP27]], 1
4378 // CHECK4-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
4379 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
4380 // CHECK4:       omp.inner.for.end:
4381 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
4382 // CHECK4:       omp.dispatch.inc:
4383 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
4384 // CHECK4:       omp.dispatch.end:
4385 // CHECK4-NEXT:    ret void
4386 //
4387 //
4388 // CHECK4-LABEL: define {{[^@]+}}@_Z7guided7PfS_S_S_
4389 // CHECK4-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
4390 // CHECK4-NEXT:  entry:
4391 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
4392 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
4393 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
4394 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
4395 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
4396 // CHECK4-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
4397 // CHECK4-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
4398 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
4399 // CHECK4-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
4400 // CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4401 // CHECK4-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
4402 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4403 // CHECK4-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
4404 // CHECK4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4405 // CHECK4-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
4406 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
4407 // CHECK4-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
4408 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
4409 // CHECK4-NEXT:    ret void
4410 //
4411 //
4412 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..5
4413 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4414 // CHECK4-NEXT:  entry:
4415 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4416 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4417 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
4418 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
4419 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i64, align 8
4420 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
4421 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
4422 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
4423 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4424 // CHECK4-NEXT:    [[I:%.*]] = alloca i64, align 8
4425 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4426 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4427 // CHECK4-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
4428 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
4429 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
4430 // CHECK4-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
4431 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
4432 // CHECK4-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
4433 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
4434 // CHECK4-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
4435 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
4436 // CHECK4-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
4437 // CHECK4-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
4438 // CHECK4-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
4439 // CHECK4-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
4440 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4441 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4442 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
4443 // CHECK4-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 36, i64 0, i64 16908287, i64 1, i64 7)
4444 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
4445 // CHECK4:       omp.dispatch.cond:
4446 // CHECK4-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
4447 // CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
4448 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
4449 // CHECK4:       omp.dispatch.body:
4450 // CHECK4-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
4451 // CHECK4-NEXT:    store i64 [[TMP12]], i64* [[DOTOMP_IV]], align 8
4452 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4453 // CHECK4:       omp.inner.for.cond:
4454 // CHECK4-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
4455 // CHECK4-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !8
4456 // CHECK4-NEXT:    [[ADD:%.*]] = add i64 [[TMP14]], 1
4457 // CHECK4-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP13]], [[ADD]]
4458 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4459 // CHECK4:       omp.inner.for.body:
4460 // CHECK4-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
4461 // CHECK4-NEXT:    [[MUL:%.*]] = mul i64 [[TMP15]], 127
4462 // CHECK4-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
4463 // CHECK4-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8, !llvm.access.group !8
4464 // CHECK4-NEXT:    [[TMP16:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !8
4465 // CHECK4-NEXT:    [[TMP17:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
4466 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[TMP17]]
4467 // CHECK4-NEXT:    [[TMP18:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !8
4468 // CHECK4-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !8
4469 // CHECK4-NEXT:    [[TMP20:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
4470 // CHECK4-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[TMP20]]
4471 // CHECK4-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX2]], align 4, !llvm.access.group !8
4472 // CHECK4-NEXT:    [[MUL3:%.*]] = fmul float [[TMP18]], [[TMP21]]
4473 // CHECK4-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !8
4474 // CHECK4-NEXT:    [[TMP23:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
4475 // CHECK4-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[TMP23]]
4476 // CHECK4-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !llvm.access.group !8
4477 // CHECK4-NEXT:    [[MUL5:%.*]] = fmul float [[MUL3]], [[TMP24]]
4478 // CHECK4-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !8
4479 // CHECK4-NEXT:    [[TMP26:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
4480 // CHECK4-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[TMP26]]
4481 // CHECK4-NEXT:    store float [[MUL5]], float* [[ARRAYIDX6]], align 4, !llvm.access.group !8
4482 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4483 // CHECK4:       omp.body.continue:
4484 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4485 // CHECK4:       omp.inner.for.inc:
4486 // CHECK4-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
4487 // CHECK4-NEXT:    [[ADD7:%.*]] = add i64 [[TMP27]], 1
4488 // CHECK4-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
4489 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
4490 // CHECK4:       omp.inner.for.end:
4491 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
4492 // CHECK4:       omp.dispatch.inc:
4493 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
4494 // CHECK4:       omp.dispatch.end:
4495 // CHECK4-NEXT:    ret void
4496 //
4497 //
4498 // CHECK4-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
4499 // CHECK4-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
4500 // CHECK4-NEXT:  entry:
4501 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
4502 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
4503 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
4504 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
4505 // CHECK4-NEXT:    [[X:%.*]] = alloca i32, align 4
4506 // CHECK4-NEXT:    [[Y:%.*]] = alloca i32, align 4
4507 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
4508 // CHECK4-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
4509 // CHECK4-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
4510 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
4511 // CHECK4-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
4512 // CHECK4-NEXT:    store i32 0, i32* [[X]], align 4
4513 // CHECK4-NEXT:    store i32 0, i32* [[Y]], align 4
4514 // CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4515 // CHECK4-NEXT:    store i32* [[Y]], i32** [[TMP0]], align 8
4516 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4517 // CHECK4-NEXT:    store float** [[A_ADDR]], float*** [[TMP1]], align 8
4518 // CHECK4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4519 // CHECK4-NEXT:    store float** [[B_ADDR]], float*** [[TMP2]], align 8
4520 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
4521 // CHECK4-NEXT:    store float** [[C_ADDR]], float*** [[TMP3]], align 8
4522 // CHECK4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
4523 // CHECK4-NEXT:    store float** [[D_ADDR]], float*** [[TMP4]], align 8
4524 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
4525 // CHECK4-NEXT:    ret void
4526 //
4527 //
4528 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..6
4529 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4530 // CHECK4-NEXT:  entry:
4531 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4532 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4533 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
4534 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
4535 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i8, align 1
4536 // CHECK4-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
4537 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
4538 // CHECK4-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
4539 // CHECK4-NEXT:    [[I:%.*]] = alloca i8, align 1
4540 // CHECK4-NEXT:    [[X:%.*]] = alloca i32, align 4
4541 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
4542 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
4543 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
4544 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4545 // CHECK4-NEXT:    [[I7:%.*]] = alloca i8, align 1
4546 // CHECK4-NEXT:    [[X8:%.*]] = alloca i32, align 4
4547 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4548 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4549 // CHECK4-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
4550 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
4551 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
4552 // CHECK4-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8
4553 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
4554 // CHECK4-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
4555 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
4556 // CHECK4-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
4557 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
4558 // CHECK4-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
4559 // CHECK4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
4560 // CHECK4-NEXT:    [[TMP10:%.*]] = load float**, float*** [[TMP9]], align 8
4561 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP2]], align 4
4562 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP11]] to i8
4563 // CHECK4-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
4564 // CHECK4-NEXT:    [[TMP12:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
4565 // CHECK4-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP12]] to i32
4566 // CHECK4-NEXT:    [[SUB:%.*]] = sub i32 57, [[CONV3]]
4567 // CHECK4-NEXT:    [[ADD:%.*]] = add i32 [[SUB]], 1
4568 // CHECK4-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
4569 // CHECK4-NEXT:    [[CONV4:%.*]] = zext i32 [[DIV]] to i64
4570 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV4]], 11
4571 // CHECK4-NEXT:    [[SUB5:%.*]] = sub nsw i64 [[MUL]], 1
4572 // CHECK4-NEXT:    store i64 [[SUB5]], i64* [[DOTCAPTURE_EXPR_2]], align 8
4573 // CHECK4-NEXT:    [[TMP13:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
4574 // CHECK4-NEXT:    store i8 [[TMP13]], i8* [[I]], align 1
4575 // CHECK4-NEXT:    store i32 11, i32* [[X]], align 4
4576 // CHECK4-NEXT:    [[TMP14:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
4577 // CHECK4-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP14]] to i32
4578 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[CONV6]], 57
4579 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
4580 // CHECK4:       omp.precond.then:
4581 // CHECK4-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
4582 // CHECK4-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
4583 // CHECK4-NEXT:    store i64 [[TMP15]], i64* [[DOTOMP_UB]], align 8
4584 // CHECK4-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
4585 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4586 // CHECK4-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
4587 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4588 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
4589 // CHECK4-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB2]], i32 [[TMP18]], i32 38, i64 0, i64 [[TMP16]], i64 1, i64 1)
4590 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
4591 // CHECK4:       omp.dispatch.cond:
4592 // CHECK4-NEXT:    [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4593 // CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
4594 // CHECK4-NEXT:    [[TMP21:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB2]], i32 [[TMP20]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
4595 // CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP21]], 0
4596 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
4597 // CHECK4:       omp.dispatch.body:
4598 // CHECK4-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
4599 // CHECK4-NEXT:    store i64 [[TMP22]], i64* [[DOTOMP_IV]], align 8
4600 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4601 // CHECK4:       omp.inner.for.cond:
4602 // CHECK4-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
4603 // CHECK4-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !11
4604 // CHECK4-NEXT:    [[CMP9:%.*]] = icmp sle i64 [[TMP23]], [[TMP24]]
4605 // CHECK4-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4606 // CHECK4:       omp.inner.for.body:
4607 // CHECK4-NEXT:    [[TMP25:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !11
4608 // CHECK4-NEXT:    [[CONV10:%.*]] = sext i8 [[TMP25]] to i64
4609 // CHECK4-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
4610 // CHECK4-NEXT:    [[DIV11:%.*]] = sdiv i64 [[TMP26]], 11
4611 // CHECK4-NEXT:    [[MUL12:%.*]] = mul nsw i64 [[DIV11]], 1
4612 // CHECK4-NEXT:    [[ADD13:%.*]] = add nsw i64 [[CONV10]], [[MUL12]]
4613 // CHECK4-NEXT:    [[CONV14:%.*]] = trunc i64 [[ADD13]] to i8
4614 // CHECK4-NEXT:    store i8 [[CONV14]], i8* [[I7]], align 1, !llvm.access.group !11
4615 // CHECK4-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
4616 // CHECK4-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
4617 // CHECK4-NEXT:    [[DIV15:%.*]] = sdiv i64 [[TMP28]], 11
4618 // CHECK4-NEXT:    [[MUL16:%.*]] = mul nsw i64 [[DIV15]], 11
4619 // CHECK4-NEXT:    [[SUB17:%.*]] = sub nsw i64 [[TMP27]], [[MUL16]]
4620 // CHECK4-NEXT:    [[MUL18:%.*]] = mul nsw i64 [[SUB17]], 1
4621 // CHECK4-NEXT:    [[SUB19:%.*]] = sub nsw i64 11, [[MUL18]]
4622 // CHECK4-NEXT:    [[CONV20:%.*]] = trunc i64 [[SUB19]] to i32
4623 // CHECK4-NEXT:    store i32 [[CONV20]], i32* [[X8]], align 4, !llvm.access.group !11
4624 // CHECK4-NEXT:    [[TMP29:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !11
4625 // CHECK4-NEXT:    [[TMP30:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
4626 // CHECK4-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP30]] to i64
4627 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP29]], i64 [[IDXPROM]]
4628 // CHECK4-NEXT:    [[TMP31:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !11
4629 // CHECK4-NEXT:    [[TMP32:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !11
4630 // CHECK4-NEXT:    [[TMP33:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
4631 // CHECK4-NEXT:    [[IDXPROM21:%.*]] = sext i8 [[TMP33]] to i64
4632 // CHECK4-NEXT:    [[ARRAYIDX22:%.*]] = getelementptr inbounds float, float* [[TMP32]], i64 [[IDXPROM21]]
4633 // CHECK4-NEXT:    [[TMP34:%.*]] = load float, float* [[ARRAYIDX22]], align 4, !llvm.access.group !11
4634 // CHECK4-NEXT:    [[MUL23:%.*]] = fmul float [[TMP31]], [[TMP34]]
4635 // CHECK4-NEXT:    [[TMP35:%.*]] = load float*, float** [[TMP10]], align 8, !llvm.access.group !11
4636 // CHECK4-NEXT:    [[TMP36:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
4637 // CHECK4-NEXT:    [[IDXPROM24:%.*]] = sext i8 [[TMP36]] to i64
4638 // CHECK4-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds float, float* [[TMP35]], i64 [[IDXPROM24]]
4639 // CHECK4-NEXT:    [[TMP37:%.*]] = load float, float* [[ARRAYIDX25]], align 4, !llvm.access.group !11
4640 // CHECK4-NEXT:    [[MUL26:%.*]] = fmul float [[MUL23]], [[TMP37]]
4641 // CHECK4-NEXT:    [[TMP38:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !11
4642 // CHECK4-NEXT:    [[TMP39:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
4643 // CHECK4-NEXT:    [[IDXPROM27:%.*]] = sext i8 [[TMP39]] to i64
4644 // CHECK4-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds float, float* [[TMP38]], i64 [[IDXPROM27]]
4645 // CHECK4-NEXT:    store float [[MUL26]], float* [[ARRAYIDX28]], align 4, !llvm.access.group !11
4646 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4647 // CHECK4:       omp.body.continue:
4648 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4649 // CHECK4:       omp.inner.for.inc:
4650 // CHECK4-NEXT:    [[TMP40:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
4651 // CHECK4-NEXT:    [[ADD29:%.*]] = add nsw i64 [[TMP40]], 1
4652 // CHECK4-NEXT:    store i64 [[ADD29]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
4653 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
4654 // CHECK4:       omp.inner.for.end:
4655 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
4656 // CHECK4:       omp.dispatch.inc:
4657 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
4658 // CHECK4:       omp.dispatch.end:
4659 // CHECK4-NEXT:    br label [[OMP_PRECOND_END]]
4660 // CHECK4:       omp.precond.end:
4661 // CHECK4-NEXT:    ret void
4662 //
4663 //
4664 // CHECK4-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
4665 // CHECK4-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
4666 // CHECK4-NEXT:  entry:
4667 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
4668 // CHECK4-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
4669 // CHECK4-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
4670 // CHECK4-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
4671 // CHECK4-NEXT:    [[X:%.*]] = alloca i32, align 4
4672 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 8
4673 // CHECK4-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
4674 // CHECK4-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
4675 // CHECK4-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
4676 // CHECK4-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
4677 // CHECK4-NEXT:    store i32 0, i32* [[X]], align 4
4678 // CHECK4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4679 // CHECK4-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
4680 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4681 // CHECK4-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
4682 // CHECK4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4683 // CHECK4-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
4684 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
4685 // CHECK4-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
4686 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.6*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
4687 // CHECK4-NEXT:    ret void
4688 //
4689 //
4690 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..7
4691 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
4692 // CHECK4-NEXT:  entry:
4693 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4694 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4695 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 8
4696 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4697 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i8, align 1
4698 // CHECK4-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
4699 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4700 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4701 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4702 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4703 // CHECK4-NEXT:    [[I:%.*]] = alloca i8, align 1
4704 // CHECK4-NEXT:    [[X:%.*]] = alloca i32, align 4
4705 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4706 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4707 // CHECK4-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 8
4708 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 8
4709 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
4710 // CHECK4-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
4711 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
4712 // CHECK4-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
4713 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
4714 // CHECK4-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
4715 // CHECK4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
4716 // CHECK4-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
4717 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4718 // CHECK4-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
4719 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4720 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4721 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4722 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
4723 // CHECK4-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 37, i32 0, i32 199, i32 1, i32 1)
4724 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
4725 // CHECK4:       omp.dispatch.cond:
4726 // CHECK4-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
4727 // CHECK4-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
4728 // CHECK4-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
4729 // CHECK4:       omp.dispatch.body:
4730 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4731 // CHECK4-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
4732 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4733 // CHECK4:       omp.inner.for.cond:
4734 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
4735 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
4736 // CHECK4-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
4737 // CHECK4-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
4738 // CHECK4:       omp.inner.for.body:
4739 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
4740 // CHECK4-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP15]], 20
4741 // CHECK4-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
4742 // CHECK4-NEXT:    [[ADD:%.*]] = add nsw i32 48, [[MUL]]
4743 // CHECK4-NEXT:    [[CONV:%.*]] = trunc i32 [[ADD]] to i8
4744 // CHECK4-NEXT:    store i8 [[CONV]], i8* [[I]], align 1, !llvm.access.group !14
4745 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
4746 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
4747 // CHECK4-NEXT:    [[DIV2:%.*]] = sdiv i32 [[TMP17]], 20
4748 // CHECK4-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[DIV2]], 20
4749 // CHECK4-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP16]], [[MUL3]]
4750 // CHECK4-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[SUB]], 1
4751 // CHECK4-NEXT:    [[ADD5:%.*]] = add nsw i32 -10, [[MUL4]]
4752 // CHECK4-NEXT:    store i32 [[ADD5]], i32* [[X]], align 4, !llvm.access.group !14
4753 // CHECK4-NEXT:    [[TMP18:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !14
4754 // CHECK4-NEXT:    [[TMP19:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
4755 // CHECK4-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP19]] to i64
4756 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP18]], i64 [[IDXPROM]]
4757 // CHECK4-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !14
4758 // CHECK4-NEXT:    [[TMP21:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !14
4759 // CHECK4-NEXT:    [[TMP22:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
4760 // CHECK4-NEXT:    [[IDXPROM6:%.*]] = zext i8 [[TMP22]] to i64
4761 // CHECK4-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP21]], i64 [[IDXPROM6]]
4762 // CHECK4-NEXT:    [[TMP23:%.*]] = load float, float* [[ARRAYIDX7]], align 4, !llvm.access.group !14
4763 // CHECK4-NEXT:    [[MUL8:%.*]] = fmul float [[TMP20]], [[TMP23]]
4764 // CHECK4-NEXT:    [[TMP24:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !14
4765 // CHECK4-NEXT:    [[TMP25:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
4766 // CHECK4-NEXT:    [[IDXPROM9:%.*]] = zext i8 [[TMP25]] to i64
4767 // CHECK4-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP24]], i64 [[IDXPROM9]]
4768 // CHECK4-NEXT:    [[TMP26:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
4769 // CHECK4-NEXT:    [[MUL11:%.*]] = fmul float [[MUL8]], [[TMP26]]
4770 // CHECK4-NEXT:    [[TMP27:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !14
4771 // CHECK4-NEXT:    [[TMP28:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
4772 // CHECK4-NEXT:    [[IDXPROM12:%.*]] = zext i8 [[TMP28]] to i64
4773 // CHECK4-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds float, float* [[TMP27]], i64 [[IDXPROM12]]
4774 // CHECK4-NEXT:    store float [[MUL11]], float* [[ARRAYIDX13]], align 4, !llvm.access.group !14
4775 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4776 // CHECK4:       omp.body.continue:
4777 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4778 // CHECK4:       omp.inner.for.inc:
4779 // CHECK4-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
4780 // CHECK4-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP29]], 1
4781 // CHECK4-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
4782 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
4783 // CHECK4:       omp.inner.for.end:
4784 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
4785 // CHECK4:       omp.dispatch.inc:
4786 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
4787 // CHECK4:       omp.dispatch.end:
4788 // CHECK4-NEXT:    ret void
4789 //
4790 //
4791 // CHECK4-LABEL: define {{[^@]+}}@_Z3foov
4792 // CHECK4-SAME: () #[[ATTR3:[0-9]+]] {
4793 // CHECK4-NEXT:  entry:
4794 // CHECK4-NEXT:    call void @_Z8mayThrowv()
4795 // CHECK4-NEXT:    ret i32 0
4796 //
4797 //
4798 // CHECK4-LABEL: define {{[^@]+}}@_Z12parallel_forPfi
4799 // CHECK4-SAME: (float* [[A:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
4800 // CHECK4-NEXT:  entry:
4801 // CHECK4-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
4802 // CHECK4-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
4803 // CHECK4-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
4804 // CHECK4-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
4805 // CHECK4-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 8
4806 // CHECK4-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
4807 // CHECK4-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
4808 // CHECK4-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
4809 // CHECK4-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
4810 // CHECK4-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
4811 // CHECK4-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
4812 // CHECK4-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP1]], align 16
4813 // CHECK4-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
4814 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
4815 // CHECK4-NEXT:    store float** [[A_ADDR]], float*** [[TMP3]], align 8
4816 // CHECK4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
4817 // CHECK4-NEXT:    store i64 [[TMP1]], i64* [[TMP4]], align 8
4818 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
4819 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
4820 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
4821 // CHECK4-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.7*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
4822 // CHECK4-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
4823 // CHECK4-NEXT:    call void @llvm.stackrestore(i8* [[TMP7]])
4824 // CHECK4-NEXT:    ret void
4825 //
4826 //
4827 // CHECK4-LABEL: define {{[^@]+}}@.omp_outlined..8
4828 // CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) {
4829 // CHECK4-NEXT:  entry:
4830 // CHECK4-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4831 // CHECK4-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4832 // CHECK4-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 8
4833 // CHECK4-NEXT:    [[N:%.*]] = alloca i32, align 4
4834 // CHECK4-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
4835 // CHECK4-NEXT:    [[TMP:%.*]] = alloca i32, align 4
4836 // CHECK4-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
4837 // CHECK4-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
4838 // CHECK4-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
4839 // CHECK4-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4840 // CHECK4-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
4841 // CHECK4-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
4842 // CHECK4-NEXT:    [[I:%.*]] = alloca i32, align 4
4843 // CHECK4-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4844 // CHECK4-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4845 // CHECK4-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 8
4846 // CHECK4-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 8
4847 // CHECK4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
4848 // CHECK4-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
4849 // CHECK4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
4850 // CHECK4-NEXT:    [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8
4851 // CHECK4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
4852 // CHECK4-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 8
4853 // CHECK4-NEXT:    store i32 [[TMP6]], i32* [[N]], align 4
4854 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
4855 // CHECK4-NEXT:    store i32 16908288, i32* [[DOTOMP_UB]], align 4
4856 // CHECK4-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
4857 // CHECK4-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
4858 // CHECK4-NEXT:    [[TMP7:%.*]] = call i8* @llvm.stacksave()
4859 // CHECK4-NEXT:    store i8* [[TMP7]], i8** [[SAVED_STACK]], align 8
4860 // CHECK4-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP4]], align 16
4861 // CHECK4-NEXT:    store i64 [[TMP4]], i64* [[__VLA_EXPR0]], align 8
4862 // CHECK4-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
4863 // CHECK4-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
4864 // CHECK4-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 5)
4865 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
4866 // CHECK4:       omp.dispatch.cond:
4867 // CHECK4-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4868 // CHECK4-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP10]], 16908288
4869 // CHECK4-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
4870 // CHECK4:       cond.true:
4871 // CHECK4-NEXT:    br label [[COND_END:%.*]]
4872 // CHECK4:       cond.false:
4873 // CHECK4-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4874 // CHECK4-NEXT:    br label [[COND_END]]
4875 // CHECK4:       cond.end:
4876 // CHECK4-NEXT:    [[COND:%.*]] = phi i32 [ 16908288, [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
4877 // CHECK4-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
4878 // CHECK4-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4879 // CHECK4-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
4880 // CHECK4-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4881 // CHECK4-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4882 // CHECK4-NEXT:    [[CMP1:%.*]] = icmp ule i32 [[TMP13]], [[TMP14]]
4883 // CHECK4-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_CLEANUP:%.*]]
4884 // CHECK4:       omp.dispatch.cleanup:
4885 // CHECK4-NEXT:    br label [[OMP_DISPATCH_END:%.*]]
4886 // CHECK4:       omp.dispatch.body:
4887 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
4888 // CHECK4:       omp.inner.for.cond:
4889 // CHECK4-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4890 // CHECK4-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4891 // CHECK4-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[TMP15]], [[TMP16]]
4892 // CHECK4-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
4893 // CHECK4:       omp.inner.for.cond.cleanup:
4894 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
4895 // CHECK4:       omp.inner.for.body:
4896 // CHECK4-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4897 // CHECK4-NEXT:    [[MUL:%.*]] = mul i32 [[TMP17]], 127
4898 // CHECK4-NEXT:    [[ADD:%.*]] = add i32 131071, [[MUL]]
4899 // CHECK4-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
4900 // CHECK4-NEXT:    [[CALL:%.*]] = invoke i32 @_Z3foov()
4901 // CHECK4-NEXT:    to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]]
4902 // CHECK4:       invoke.cont:
4903 // CHECK4-NEXT:    [[CONV:%.*]] = sitofp i32 [[CALL]] to float
4904 // CHECK4-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
4905 // CHECK4-NEXT:    [[IDXPROM:%.*]] = zext i32 [[TMP18]] to i64
4906 // CHECK4-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[VLA]], i64 [[IDXPROM]]
4907 // CHECK4-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
4908 // CHECK4-NEXT:    [[ADD3:%.*]] = fadd float [[CONV]], [[TMP19]]
4909 // CHECK4-NEXT:    [[TMP20:%.*]] = load i32, i32* [[N]], align 4
4910 // CHECK4-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP20]] to float
4911 // CHECK4-NEXT:    [[ADD5:%.*]] = fadd float [[ADD3]], [[CONV4]]
4912 // CHECK4-NEXT:    [[TMP21:%.*]] = load float*, float** [[TMP2]], align 8
4913 // CHECK4-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I]], align 4
4914 // CHECK4-NEXT:    [[IDXPROM6:%.*]] = zext i32 [[TMP22]] to i64
4915 // CHECK4-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP21]], i64 [[IDXPROM6]]
4916 // CHECK4-NEXT:    [[TMP23:%.*]] = load float, float* [[ARRAYIDX7]], align 4
4917 // CHECK4-NEXT:    [[ADD8:%.*]] = fadd float [[TMP23]], [[ADD5]]
4918 // CHECK4-NEXT:    store float [[ADD8]], float* [[ARRAYIDX7]], align 4
4919 // CHECK4-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
4920 // CHECK4:       omp.body.continue:
4921 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
4922 // CHECK4:       omp.inner.for.inc:
4923 // CHECK4-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
4924 // CHECK4-NEXT:    [[ADD9:%.*]] = add i32 [[TMP24]], 1
4925 // CHECK4-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
4926 // CHECK4-NEXT:    br label [[OMP_INNER_FOR_COND]]
4927 // CHECK4:       omp.inner.for.end:
4928 // CHECK4-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
4929 // CHECK4:       omp.dispatch.inc:
4930 // CHECK4-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
4931 // CHECK4-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4932 // CHECK4-NEXT:    [[ADD10:%.*]] = add i32 [[TMP25]], [[TMP26]]
4933 // CHECK4-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_LB]], align 4
4934 // CHECK4-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
4935 // CHECK4-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
4936 // CHECK4-NEXT:    [[ADD11:%.*]] = add i32 [[TMP27]], [[TMP28]]
4937 // CHECK4-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_UB]], align 4
4938 // CHECK4-NEXT:    br label [[OMP_DISPATCH_COND]]
4939 // CHECK4:       omp.dispatch.end:
4940 // CHECK4-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]])
4941 // CHECK4-NEXT:    [[TMP29:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
4942 // CHECK4-NEXT:    call void @llvm.stackrestore(i8* [[TMP29]])
4943 // CHECK4-NEXT:    ret void
4944 // CHECK4:       terminate.lpad:
4945 // CHECK4-NEXT:    [[TMP30:%.*]] = landingpad { i8*, i32 }
4946 // CHECK4-NEXT:    catch i8* null
4947 // CHECK4-NEXT:    [[TMP31:%.*]] = extractvalue { i8*, i32 } [[TMP30]], 0
4948 // CHECK4-NEXT:    call void @__clang_call_terminate(i8* [[TMP31]]) #[[ATTR7:[0-9]+]]
4949 // CHECK4-NEXT:    unreachable
4950 //
4951 //
4952 // CHECK4-LABEL: define {{[^@]+}}@__clang_call_terminate
4953 // CHECK4-SAME: (i8* [[TMP0:%.*]]) #[[ATTR6:[0-9]+]] comdat {
4954 // CHECK4-NEXT:    [[TMP2:%.*]] = call i8* @__cxa_begin_catch(i8* [[TMP0]]) #[[ATTR2:[0-9]+]]
4955 // CHECK4-NEXT:    call void @_ZSt9terminatev() #[[ATTR7]]
4956 // CHECK4-NEXT:    unreachable
4957 //
4958 //
4959 // CHECK5-LABEL: define {{[^@]+}}@_Z17with_var_schedulev
4960 // CHECK5-SAME: () #[[ATTR0:[0-9]+]] !dbg [[DBG7:![0-9]+]] {
4961 // CHECK5-NEXT:  entry:
4962 // CHECK5-NEXT:    [[A:%.*]] = alloca double, align 8
4963 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
4964 // CHECK5-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
4965 // CHECK5-NEXT:    store double 5.000000e+00, double* [[A]], align 8, !dbg [[DBG10:![0-9]+]]
4966 // CHECK5-NEXT:    [[TMP0:%.*]] = load double, double* [[A]], align 8, !dbg [[DBG11:![0-9]+]]
4967 // CHECK5-NEXT:    [[CONV:%.*]] = fptosi double [[TMP0]] to i8, !dbg [[DBG11]]
4968 // CHECK5-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1, !dbg [[DBG11]]
4969 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0, !dbg [[DBG11]]
4970 // CHECK5-NEXT:    [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !dbg [[DBG11]]
4971 // CHECK5-NEXT:    store i8 [[TMP2]], i8* [[TMP1]], align 1, !dbg [[DBG11]]
4972 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB4:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]]), !dbg [[DBG11]]
4973 // CHECK5-NEXT:    ret void, !dbg [[DBG12:![0-9]+]]
4974 //
4975 //
4976 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined.
4977 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] !dbg [[DBG13:![0-9]+]] {
4978 // CHECK5-NEXT:  entry:
4979 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
4980 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
4981 // CHECK5-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
4982 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
4983 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
4984 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i64, align 8
4985 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca double, align 8
4986 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
4987 // CHECK5-NEXT:    [[I:%.*]] = alloca i64, align 8
4988 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
4989 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
4990 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
4991 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
4992 // CHECK5-NEXT:    [[A:%.*]] = alloca double, align 8
4993 // CHECK5-NEXT:    [[I4:%.*]] = alloca i64, align 8
4994 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
4995 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
4996 // CHECK5-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
4997 // CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8, !dbg [[DBG14:![0-9]+]]
4998 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0, !dbg [[DBG14]]
4999 // CHECK5-NEXT:    [[TMP2:%.*]] = load i8, i8* [[TMP1]], align 1, !dbg [[DBG14]]
5000 // CHECK5-NEXT:    store i8 [[TMP2]], i8* [[DOTCAPTURE_EXPR_]], align 1, !dbg [[DBG14]]
5001 // CHECK5-NEXT:    [[TMP3:%.*]] = load double, double* undef, align 8, !dbg [[DBG15:![0-9]+]]
5002 // CHECK5-NEXT:    [[ADD:%.*]] = fadd double 2.000000e+00, [[TMP3]], !dbg [[DBG15]]
5003 // CHECK5-NEXT:    store double [[ADD]], double* [[DOTCAPTURE_EXPR_1]], align 8, !dbg [[DBG15]]
5004 // CHECK5-NEXT:    [[TMP4:%.*]] = load double, double* [[DOTCAPTURE_EXPR_1]], align 8, !dbg [[DBG15]]
5005 // CHECK5-NEXT:    [[SUB:%.*]] = fsub double [[TMP4]], 1.000000e+00, !dbg [[DBG15]]
5006 // CHECK5-NEXT:    [[DIV:%.*]] = fdiv double [[SUB]], 1.000000e+00, !dbg [[DBG15]]
5007 // CHECK5-NEXT:    [[CONV:%.*]] = fptoui double [[DIV]] to i64, !dbg [[DBG15]]
5008 // CHECK5-NEXT:    [[SUB3:%.*]] = sub i64 [[CONV]], 1, !dbg [[DBG15]]
5009 // CHECK5-NEXT:    store i64 [[SUB3]], i64* [[DOTCAPTURE_EXPR_2]], align 8, !dbg [[DBG15]]
5010 // CHECK5-NEXT:    store i64 1, i64* [[I]], align 8, !dbg [[DBG15]]
5011 // CHECK5-NEXT:    [[TMP5:%.*]] = load double, double* [[DOTCAPTURE_EXPR_1]], align 8, !dbg [[DBG15]]
5012 // CHECK5-NEXT:    [[CMP:%.*]] = fcmp olt double 1.000000e+00, [[TMP5]], !dbg [[DBG15]]
5013 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]], !dbg [[DBG14]]
5014 // CHECK5:       omp.precond.then:
5015 // CHECK5-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8, !dbg [[DBG15]]
5016 // CHECK5-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8, !dbg [[DBG15]]
5017 // CHECK5-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_UB]], align 8, !dbg [[DBG15]]
5018 // CHECK5-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8, !dbg [[DBG15]]
5019 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4, !dbg [[DBG15]]
5020 // CHECK5-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !dbg [[DBG14]]
5021 // CHECK5-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP7]] to i64, !dbg [[DBG14]]
5022 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !dbg [[DBG14]]
5023 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4, !dbg [[DBG14]]
5024 // CHECK5-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 [[CONV5]]), !dbg [[DBG14]]
5025 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]], !dbg [[DBG14]]
5026 // CHECK5:       omp.dispatch.cond:
5027 // CHECK5-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !dbg [[DBG15]]
5028 // CHECK5-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8, !dbg [[DBG15]]
5029 // CHECK5-NEXT:    [[CMP6:%.*]] = icmp ugt i64 [[TMP10]], [[TMP11]], !dbg [[DBG15]]
5030 // CHECK5-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]], !dbg [[DBG15]]
5031 // CHECK5:       cond.true:
5032 // CHECK5-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8, !dbg [[DBG15]]
5033 // CHECK5-NEXT:    br label [[COND_END:%.*]], !dbg [[DBG15]]
5034 // CHECK5:       cond.false:
5035 // CHECK5-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !dbg [[DBG15]]
5036 // CHECK5-NEXT:    br label [[COND_END]], !dbg [[DBG15]]
5037 // CHECK5:       cond.end:
5038 // CHECK5-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ], !dbg [[DBG15]]
5039 // CHECK5-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8, !dbg [[DBG15]]
5040 // CHECK5-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8, !dbg [[DBG15]]
5041 // CHECK5-NEXT:    store i64 [[TMP14]], i64* [[DOTOMP_IV]], align 8, !dbg [[DBG15]]
5042 // CHECK5-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG15]]
5043 // CHECK5-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !dbg [[DBG15]]
5044 // CHECK5-NEXT:    [[ADD7:%.*]] = add i64 [[TMP16]], 1, !dbg [[DBG15]]
5045 // CHECK5-NEXT:    [[CMP8:%.*]] = icmp ult i64 [[TMP15]], [[ADD7]], !dbg [[DBG15]]
5046 // CHECK5-NEXT:    br i1 [[CMP8]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]], !dbg [[DBG14]]
5047 // CHECK5:       omp.dispatch.body:
5048 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]], !dbg [[DBG14]]
5049 // CHECK5:       omp.inner.for.cond:
5050 // CHECK5-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG15]]
5051 // CHECK5-NEXT:    [[TMP18:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !dbg [[DBG15]]
5052 // CHECK5-NEXT:    [[ADD9:%.*]] = add i64 [[TMP18]], 1, !dbg [[DBG15]]
5053 // CHECK5-NEXT:    [[CMP10:%.*]] = icmp ult i64 [[TMP17]], [[ADD9]], !dbg [[DBG15]]
5054 // CHECK5-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]], !dbg [[DBG14]]
5055 // CHECK5:       omp.inner.for.body:
5056 // CHECK5-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG15]]
5057 // CHECK5-NEXT:    [[MUL:%.*]] = mul i64 [[TMP19]], 1, !dbg [[DBG15]]
5058 // CHECK5-NEXT:    [[ADD11:%.*]] = add i64 1, [[MUL]], !dbg [[DBG15]]
5059 // CHECK5-NEXT:    store i64 [[ADD11]], i64* [[I4]], align 8, !dbg [[DBG15]]
5060 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]], !dbg [[DBG16:![0-9]+]]
5061 // CHECK5:       omp.body.continue:
5062 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]], !dbg [[DBG14]]
5063 // CHECK5:       omp.inner.for.inc:
5064 // CHECK5-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG15]]
5065 // CHECK5-NEXT:    [[ADD12:%.*]] = add i64 [[TMP20]], 1, !dbg [[DBG15]]
5066 // CHECK5-NEXT:    store i64 [[ADD12]], i64* [[DOTOMP_IV]], align 8, !dbg [[DBG15]]
5067 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !dbg [[DBG14]], !llvm.loop [[LOOP17:![0-9]+]]
5068 // CHECK5:       omp.inner.for.end:
5069 // CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]], !dbg [[DBG14]]
5070 // CHECK5:       omp.dispatch.inc:
5071 // CHECK5-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8, !dbg [[DBG15]]
5072 // CHECK5-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8, !dbg [[DBG15]]
5073 // CHECK5-NEXT:    [[ADD13:%.*]] = add i64 [[TMP21]], [[TMP22]], !dbg [[DBG15]]
5074 // CHECK5-NEXT:    store i64 [[ADD13]], i64* [[DOTOMP_LB]], align 8, !dbg [[DBG15]]
5075 // CHECK5-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !dbg [[DBG15]]
5076 // CHECK5-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8, !dbg [[DBG15]]
5077 // CHECK5-NEXT:    [[ADD14:%.*]] = add i64 [[TMP23]], [[TMP24]], !dbg [[DBG15]]
5078 // CHECK5-NEXT:    store i64 [[ADD14]], i64* [[DOTOMP_UB]], align 8, !dbg [[DBG15]]
5079 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]], !dbg [[DBG14]], !llvm.loop [[LOOP18:![0-9]+]]
5080 // CHECK5:       omp.dispatch.end:
5081 // CHECK5-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !dbg [[DBG14]]
5082 // CHECK5-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4, !dbg [[DBG14]]
5083 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP26]]), !dbg [[DBG14]]
5084 // CHECK5-NEXT:    br label [[OMP_PRECOND_END]], !dbg [[DBG14]]
5085 // CHECK5:       omp.precond.end:
5086 // CHECK5-NEXT:    ret void, !dbg [[DBG16]]
5087 //
5088 //
5089 // CHECK5-LABEL: define {{[^@]+}}@_Z23without_schedule_clausePfS_S_S_
5090 // CHECK5-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] !dbg [[DBG21:![0-9]+]] {
5091 // CHECK5-NEXT:  entry:
5092 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
5093 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
5094 // CHECK5-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
5095 // CHECK5-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
5096 // CHECK5-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
5097 // CHECK5-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
5098 // CHECK5-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
5099 // CHECK5-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
5100 // CHECK5-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
5101 // CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0, !dbg [[DBG22:![0-9]+]]
5102 // CHECK5-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8, !dbg [[DBG22]]
5103 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1, !dbg [[DBG22]]
5104 // CHECK5-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8, !dbg [[DBG22]]
5105 // CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2, !dbg [[DBG22]]
5106 // CHECK5-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8, !dbg [[DBG22]]
5107 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3, !dbg [[DBG22]]
5108 // CHECK5-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8, !dbg [[DBG22]]
5109 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB9:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]]), !dbg [[DBG22]]
5110 // CHECK5-NEXT:    ret void, !dbg [[DBG23:![0-9]+]]
5111 //
5112 //
5113 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..1
5114 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] !dbg [[DBG24:![0-9]+]] {
5115 // CHECK5-NEXT:  entry:
5116 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5117 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5118 // CHECK5-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
5119 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5120 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5121 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5122 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5123 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5124 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5125 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
5126 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5127 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5128 // CHECK5-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
5129 // CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8, !dbg [[DBG25:![0-9]+]]
5130 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0, !dbg [[DBG25]]
5131 // CHECK5-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8, !dbg [[DBG25]]
5132 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1, !dbg [[DBG25]]
5133 // CHECK5-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8, !dbg [[DBG25]]
5134 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 2, !dbg [[DBG25]]
5135 // CHECK5-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8, !dbg [[DBG25]]
5136 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 3, !dbg [[DBG25]]
5137 // CHECK5-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8, !dbg [[DBG25]]
5138 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4, !dbg [[DBG26:![0-9]+]]
5139 // CHECK5-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG26]]
5140 // CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4, !dbg [[DBG26]]
5141 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4, !dbg [[DBG26]]
5142 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !dbg [[DBG25]]
5143 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4, !dbg [[DBG25]]
5144 // CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB6:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1), !dbg [[DBG25]]
5145 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG26]]
5146 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 4571423, !dbg [[DBG26]]
5147 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]], !dbg [[DBG26]]
5148 // CHECK5:       cond.true:
5149 // CHECK5-NEXT:    br label [[COND_END:%.*]], !dbg [[DBG26]]
5150 // CHECK5:       cond.false:
5151 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG26]]
5152 // CHECK5-NEXT:    br label [[COND_END]], !dbg [[DBG26]]
5153 // CHECK5:       cond.end:
5154 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 4571423, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ], !dbg [[DBG26]]
5155 // CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4, !dbg [[DBG26]]
5156 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4, !dbg [[DBG26]]
5157 // CHECK5-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4, !dbg [[DBG26]]
5158 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]], !dbg [[DBG25]]
5159 // CHECK5:       omp.inner.for.cond:
5160 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG26]]
5161 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG26]]
5162 // CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]], !dbg [[DBG26]]
5163 // CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]], !dbg [[DBG25]]
5164 // CHECK5:       omp.inner.for.body:
5165 // CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG26]]
5166 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 7, !dbg [[DBG26]]
5167 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 33, [[MUL]], !dbg [[DBG26]]
5168 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !dbg [[DBG26]]
5169 // CHECK5-NEXT:    [[TMP17:%.*]] = load float*, float** [[TMP4]], align 8, !dbg [[DBG27:![0-9]+]]
5170 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG27]]
5171 // CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64, !dbg [[DBG27]]
5172 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM]], !dbg [[DBG27]]
5173 // CHECK5-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4, !dbg [[DBG27]]
5174 // CHECK5-NEXT:    [[TMP20:%.*]] = load float*, float** [[TMP6]], align 8, !dbg [[DBG27]]
5175 // CHECK5-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG27]]
5176 // CHECK5-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP21]] to i64, !dbg [[DBG27]]
5177 // CHECK5-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM2]], !dbg [[DBG27]]
5178 // CHECK5-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX3]], align 4, !dbg [[DBG27]]
5179 // CHECK5-NEXT:    [[MUL4:%.*]] = fmul float [[TMP19]], [[TMP22]], !dbg [[DBG27]]
5180 // CHECK5-NEXT:    [[TMP23:%.*]] = load float*, float** [[TMP8]], align 8, !dbg [[DBG27]]
5181 // CHECK5-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG27]]
5182 // CHECK5-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP24]] to i64, !dbg [[DBG27]]
5183 // CHECK5-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM5]], !dbg [[DBG27]]
5184 // CHECK5-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX6]], align 4, !dbg [[DBG27]]
5185 // CHECK5-NEXT:    [[MUL7:%.*]] = fmul float [[MUL4]], [[TMP25]], !dbg [[DBG27]]
5186 // CHECK5-NEXT:    [[TMP26:%.*]] = load float*, float** [[TMP2]], align 8, !dbg [[DBG27]]
5187 // CHECK5-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG27]]
5188 // CHECK5-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP27]] to i64, !dbg [[DBG27]]
5189 // CHECK5-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[TMP26]], i64 [[IDXPROM8]], !dbg [[DBG27]]
5190 // CHECK5-NEXT:    store float [[MUL7]], float* [[ARRAYIDX9]], align 4, !dbg [[DBG27]]
5191 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]], !dbg [[DBG28:![0-9]+]]
5192 // CHECK5:       omp.body.continue:
5193 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]], !dbg [[DBG25]]
5194 // CHECK5:       omp.inner.for.inc:
5195 // CHECK5-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG26]]
5196 // CHECK5-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP28]], 1, !dbg [[DBG26]]
5197 // CHECK5-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4, !dbg [[DBG26]]
5198 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !dbg [[DBG25]], !llvm.loop [[LOOP29:![0-9]+]]
5199 // CHECK5:       omp.inner.for.end:
5200 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]], !dbg [[DBG25]]
5201 // CHECK5:       omp.loop.exit:
5202 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB8:[0-9]+]], i32 [[TMP10]]), !dbg [[DBG25]]
5203 // CHECK5-NEXT:    ret void, !dbg [[DBG28]]
5204 //
5205 //
5206 // CHECK5-LABEL: define {{[^@]+}}@_Z18static_not_chunkedPfS_S_S_
5207 // CHECK5-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] !dbg [[DBG30:![0-9]+]] {
5208 // CHECK5-NEXT:  entry:
5209 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
5210 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
5211 // CHECK5-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
5212 // CHECK5-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
5213 // CHECK5-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 8
5214 // CHECK5-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
5215 // CHECK5-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
5216 // CHECK5-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
5217 // CHECK5-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
5218 // CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0, !dbg [[DBG31:![0-9]+]]
5219 // CHECK5-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8, !dbg [[DBG31]]
5220 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1, !dbg [[DBG31]]
5221 // CHECK5-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8, !dbg [[DBG31]]
5222 // CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2, !dbg [[DBG31]]
5223 // CHECK5-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8, !dbg [[DBG31]]
5224 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3, !dbg [[DBG31]]
5225 // CHECK5-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8, !dbg [[DBG31]]
5226 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB14:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]]), !dbg [[DBG31]]
5227 // CHECK5-NEXT:    ret void, !dbg [[DBG32:![0-9]+]]
5228 //
5229 //
5230 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..2
5231 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] !dbg [[DBG33:![0-9]+]] {
5232 // CHECK5-NEXT:  entry:
5233 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5234 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5235 // CHECK5-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
5236 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5237 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5238 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5239 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5240 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5241 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5242 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
5243 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5244 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5245 // CHECK5-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
5246 // CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8, !dbg [[DBG34:![0-9]+]]
5247 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0, !dbg [[DBG34]]
5248 // CHECK5-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8, !dbg [[DBG34]]
5249 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1, !dbg [[DBG34]]
5250 // CHECK5-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8, !dbg [[DBG34]]
5251 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 2, !dbg [[DBG34]]
5252 // CHECK5-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8, !dbg [[DBG34]]
5253 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 3, !dbg [[DBG34]]
5254 // CHECK5-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8, !dbg [[DBG34]]
5255 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4, !dbg [[DBG35:![0-9]+]]
5256 // CHECK5-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG35]]
5257 // CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4, !dbg [[DBG35]]
5258 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4, !dbg [[DBG35]]
5259 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !dbg [[DBG34]]
5260 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4, !dbg [[DBG34]]
5261 // CHECK5-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB11:[0-9]+]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1), !dbg [[DBG34]]
5262 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG35]]
5263 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 4571423, !dbg [[DBG35]]
5264 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]], !dbg [[DBG35]]
5265 // CHECK5:       cond.true:
5266 // CHECK5-NEXT:    br label [[COND_END:%.*]], !dbg [[DBG35]]
5267 // CHECK5:       cond.false:
5268 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG35]]
5269 // CHECK5-NEXT:    br label [[COND_END]], !dbg [[DBG35]]
5270 // CHECK5:       cond.end:
5271 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 4571423, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ], !dbg [[DBG35]]
5272 // CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4, !dbg [[DBG35]]
5273 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4, !dbg [[DBG35]]
5274 // CHECK5-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4, !dbg [[DBG35]]
5275 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]], !dbg [[DBG34]]
5276 // CHECK5:       omp.inner.for.cond:
5277 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG35]]
5278 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG35]]
5279 // CHECK5-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]], !dbg [[DBG35]]
5280 // CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]], !dbg [[DBG34]]
5281 // CHECK5:       omp.inner.for.body:
5282 // CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG35]]
5283 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 7, !dbg [[DBG35]]
5284 // CHECK5-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]], !dbg [[DBG35]]
5285 // CHECK5-NEXT:    store i32 [[SUB]], i32* [[I]], align 4, !dbg [[DBG35]]
5286 // CHECK5-NEXT:    [[TMP17:%.*]] = load float*, float** [[TMP4]], align 8, !dbg [[DBG36:![0-9]+]]
5287 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG36]]
5288 // CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64, !dbg [[DBG36]]
5289 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM]], !dbg [[DBG36]]
5290 // CHECK5-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4, !dbg [[DBG36]]
5291 // CHECK5-NEXT:    [[TMP20:%.*]] = load float*, float** [[TMP6]], align 8, !dbg [[DBG36]]
5292 // CHECK5-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG36]]
5293 // CHECK5-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP21]] to i64, !dbg [[DBG36]]
5294 // CHECK5-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM2]], !dbg [[DBG36]]
5295 // CHECK5-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX3]], align 4, !dbg [[DBG36]]
5296 // CHECK5-NEXT:    [[MUL4:%.*]] = fmul float [[TMP19]], [[TMP22]], !dbg [[DBG36]]
5297 // CHECK5-NEXT:    [[TMP23:%.*]] = load float*, float** [[TMP8]], align 8, !dbg [[DBG36]]
5298 // CHECK5-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG36]]
5299 // CHECK5-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP24]] to i64, !dbg [[DBG36]]
5300 // CHECK5-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM5]], !dbg [[DBG36]]
5301 // CHECK5-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX6]], align 4, !dbg [[DBG36]]
5302 // CHECK5-NEXT:    [[MUL7:%.*]] = fmul float [[MUL4]], [[TMP25]], !dbg [[DBG36]]
5303 // CHECK5-NEXT:    [[TMP26:%.*]] = load float*, float** [[TMP2]], align 8, !dbg [[DBG36]]
5304 // CHECK5-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG36]]
5305 // CHECK5-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP27]] to i64, !dbg [[DBG36]]
5306 // CHECK5-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[TMP26]], i64 [[IDXPROM8]], !dbg [[DBG36]]
5307 // CHECK5-NEXT:    store float [[MUL7]], float* [[ARRAYIDX9]], align 4, !dbg [[DBG36]]
5308 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]], !dbg [[DBG37:![0-9]+]]
5309 // CHECK5:       omp.body.continue:
5310 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]], !dbg [[DBG34]]
5311 // CHECK5:       omp.inner.for.inc:
5312 // CHECK5-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG35]]
5313 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP28]], 1, !dbg [[DBG35]]
5314 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4, !dbg [[DBG35]]
5315 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !dbg [[DBG34]], !llvm.loop [[LOOP38:![0-9]+]]
5316 // CHECK5:       omp.inner.for.end:
5317 // CHECK5-NEXT:    br label [[OMP_LOOP_EXIT:%.*]], !dbg [[DBG34]]
5318 // CHECK5:       omp.loop.exit:
5319 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB13:[0-9]+]], i32 [[TMP10]]), !dbg [[DBG34]]
5320 // CHECK5-NEXT:    ret void, !dbg [[DBG37]]
5321 //
5322 //
5323 // CHECK5-LABEL: define {{[^@]+}}@_Z14static_chunkedPfS_S_S_
5324 // CHECK5-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] !dbg [[DBG39:![0-9]+]] {
5325 // CHECK5-NEXT:  entry:
5326 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
5327 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
5328 // CHECK5-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
5329 // CHECK5-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
5330 // CHECK5-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
5331 // CHECK5-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
5332 // CHECK5-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
5333 // CHECK5-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
5334 // CHECK5-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
5335 // CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0, !dbg [[DBG40:![0-9]+]]
5336 // CHECK5-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8, !dbg [[DBG40]]
5337 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1, !dbg [[DBG40]]
5338 // CHECK5-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8, !dbg [[DBG40]]
5339 // CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2, !dbg [[DBG40]]
5340 // CHECK5-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8, !dbg [[DBG40]]
5341 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3, !dbg [[DBG40]]
5342 // CHECK5-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8, !dbg [[DBG40]]
5343 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB19:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]]), !dbg [[DBG40]]
5344 // CHECK5-NEXT:    ret void, !dbg [[DBG41:![0-9]+]]
5345 //
5346 //
5347 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..3
5348 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] !dbg [[DBG42:![0-9]+]] {
5349 // CHECK5-NEXT:  entry:
5350 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5351 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5352 // CHECK5-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
5353 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5354 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
5355 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5356 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5357 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5358 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5359 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
5360 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5361 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5362 // CHECK5-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
5363 // CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8, !dbg [[DBG43:![0-9]+]]
5364 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0, !dbg [[DBG43]]
5365 // CHECK5-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8, !dbg [[DBG43]]
5366 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1, !dbg [[DBG43]]
5367 // CHECK5-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8, !dbg [[DBG43]]
5368 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2, !dbg [[DBG43]]
5369 // CHECK5-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8, !dbg [[DBG43]]
5370 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3, !dbg [[DBG43]]
5371 // CHECK5-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8, !dbg [[DBG43]]
5372 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4, !dbg [[DBG44:![0-9]+]]
5373 // CHECK5-NEXT:    store i32 16908288, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG44]]
5374 // CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4, !dbg [[DBG44]]
5375 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4, !dbg [[DBG44]]
5376 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !dbg [[DBG43]]
5377 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4, !dbg [[DBG43]]
5378 // CHECK5-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB16:[0-9]+]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 5), !dbg [[DBG43]]
5379 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]], !dbg [[DBG43]]
5380 // CHECK5:       omp.dispatch.cond:
5381 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG44]]
5382 // CHECK5-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP11]], 16908288, !dbg [[DBG44]]
5383 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]], !dbg [[DBG44]]
5384 // CHECK5:       cond.true:
5385 // CHECK5-NEXT:    br label [[COND_END:%.*]], !dbg [[DBG44]]
5386 // CHECK5:       cond.false:
5387 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG44]]
5388 // CHECK5-NEXT:    br label [[COND_END]], !dbg [[DBG44]]
5389 // CHECK5:       cond.end:
5390 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 16908288, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ], !dbg [[DBG44]]
5391 // CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4, !dbg [[DBG44]]
5392 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4, !dbg [[DBG44]]
5393 // CHECK5-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4, !dbg [[DBG44]]
5394 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG44]]
5395 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG44]]
5396 // CHECK5-NEXT:    [[CMP1:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]], !dbg [[DBG44]]
5397 // CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]], !dbg [[DBG43]]
5398 // CHECK5:       omp.dispatch.body:
5399 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]], !dbg [[DBG43]]
5400 // CHECK5:       omp.inner.for.cond:
5401 // CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG44]]
5402 // CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG44]]
5403 // CHECK5-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[TMP16]], [[TMP17]], !dbg [[DBG44]]
5404 // CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]], !dbg [[DBG43]]
5405 // CHECK5:       omp.inner.for.body:
5406 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG44]]
5407 // CHECK5-NEXT:    [[MUL:%.*]] = mul i32 [[TMP18]], 127, !dbg [[DBG44]]
5408 // CHECK5-NEXT:    [[ADD:%.*]] = add i32 131071, [[MUL]], !dbg [[DBG44]]
5409 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !dbg [[DBG44]]
5410 // CHECK5-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP4]], align 8, !dbg [[DBG45:![0-9]+]]
5411 // CHECK5-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG45]]
5412 // CHECK5-NEXT:    [[IDXPROM:%.*]] = zext i32 [[TMP20]] to i64, !dbg [[DBG45]]
5413 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[IDXPROM]], !dbg [[DBG45]]
5414 // CHECK5-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX]], align 4, !dbg [[DBG45]]
5415 // CHECK5-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP6]], align 8, !dbg [[DBG45]]
5416 // CHECK5-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG45]]
5417 // CHECK5-NEXT:    [[IDXPROM3:%.*]] = zext i32 [[TMP23]] to i64, !dbg [[DBG45]]
5418 // CHECK5-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[IDXPROM3]], !dbg [[DBG45]]
5419 // CHECK5-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !dbg [[DBG45]]
5420 // CHECK5-NEXT:    [[MUL5:%.*]] = fmul float [[TMP21]], [[TMP24]], !dbg [[DBG45]]
5421 // CHECK5-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP8]], align 8, !dbg [[DBG45]]
5422 // CHECK5-NEXT:    [[TMP26:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG45]]
5423 // CHECK5-NEXT:    [[IDXPROM6:%.*]] = zext i32 [[TMP26]] to i64, !dbg [[DBG45]]
5424 // CHECK5-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[IDXPROM6]], !dbg [[DBG45]]
5425 // CHECK5-NEXT:    [[TMP27:%.*]] = load float, float* [[ARRAYIDX7]], align 4, !dbg [[DBG45]]
5426 // CHECK5-NEXT:    [[MUL8:%.*]] = fmul float [[MUL5]], [[TMP27]], !dbg [[DBG45]]
5427 // CHECK5-NEXT:    [[TMP28:%.*]] = load float*, float** [[TMP2]], align 8, !dbg [[DBG45]]
5428 // CHECK5-NEXT:    [[TMP29:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG45]]
5429 // CHECK5-NEXT:    [[IDXPROM9:%.*]] = zext i32 [[TMP29]] to i64, !dbg [[DBG45]]
5430 // CHECK5-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP28]], i64 [[IDXPROM9]], !dbg [[DBG45]]
5431 // CHECK5-NEXT:    store float [[MUL8]], float* [[ARRAYIDX10]], align 4, !dbg [[DBG45]]
5432 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]], !dbg [[DBG46:![0-9]+]]
5433 // CHECK5:       omp.body.continue:
5434 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]], !dbg [[DBG43]]
5435 // CHECK5:       omp.inner.for.inc:
5436 // CHECK5-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG44]]
5437 // CHECK5-NEXT:    [[ADD11:%.*]] = add i32 [[TMP30]], 1, !dbg [[DBG44]]
5438 // CHECK5-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4, !dbg [[DBG44]]
5439 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !dbg [[DBG43]], !llvm.loop [[LOOP47:![0-9]+]]
5440 // CHECK5:       omp.inner.for.end:
5441 // CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]], !dbg [[DBG43]]
5442 // CHECK5:       omp.dispatch.inc:
5443 // CHECK5-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4, !dbg [[DBG44]]
5444 // CHECK5-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !dbg [[DBG44]]
5445 // CHECK5-NEXT:    [[ADD12:%.*]] = add i32 [[TMP31]], [[TMP32]], !dbg [[DBG44]]
5446 // CHECK5-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_LB]], align 4, !dbg [[DBG44]]
5447 // CHECK5-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG44]]
5448 // CHECK5-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !dbg [[DBG44]]
5449 // CHECK5-NEXT:    [[ADD13:%.*]] = add i32 [[TMP33]], [[TMP34]], !dbg [[DBG44]]
5450 // CHECK5-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_UB]], align 4, !dbg [[DBG44]]
5451 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]], !dbg [[DBG43]], !llvm.loop [[LOOP48:![0-9]+]]
5452 // CHECK5:       omp.dispatch.end:
5453 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB18:[0-9]+]], i32 [[TMP10]]), !dbg [[DBG43]]
5454 // CHECK5-NEXT:    ret void, !dbg [[DBG46]]
5455 //
5456 //
5457 // CHECK5-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
5458 // CHECK5-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] !dbg [[DBG49:![0-9]+]] {
5459 // CHECK5-NEXT:  entry:
5460 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
5461 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
5462 // CHECK5-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
5463 // CHECK5-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
5464 // CHECK5-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 8
5465 // CHECK5-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
5466 // CHECK5-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
5467 // CHECK5-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
5468 // CHECK5-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
5469 // CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0, !dbg [[DBG50:![0-9]+]]
5470 // CHECK5-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8, !dbg [[DBG50]]
5471 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1, !dbg [[DBG50]]
5472 // CHECK5-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8, !dbg [[DBG50]]
5473 // CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2, !dbg [[DBG50]]
5474 // CHECK5-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8, !dbg [[DBG50]]
5475 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3, !dbg [[DBG50]]
5476 // CHECK5-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8, !dbg [[DBG50]]
5477 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB21:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]]), !dbg [[DBG50]]
5478 // CHECK5-NEXT:    ret void, !dbg [[DBG51:![0-9]+]]
5479 //
5480 //
5481 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..4
5482 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] !dbg [[DBG52:![0-9]+]] {
5483 // CHECK5-NEXT:  entry:
5484 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5485 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5486 // CHECK5-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
5487 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
5488 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i64, align 8
5489 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
5490 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
5491 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
5492 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5493 // CHECK5-NEXT:    [[I:%.*]] = alloca i64, align 8
5494 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5495 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5496 // CHECK5-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
5497 // CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8, !dbg [[DBG53:![0-9]+]]
5498 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0, !dbg [[DBG53]]
5499 // CHECK5-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8, !dbg [[DBG53]]
5500 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1, !dbg [[DBG53]]
5501 // CHECK5-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8, !dbg [[DBG53]]
5502 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2, !dbg [[DBG53]]
5503 // CHECK5-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8, !dbg [[DBG53]]
5504 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3, !dbg [[DBG53]]
5505 // CHECK5-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8, !dbg [[DBG53]]
5506 // CHECK5-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8, !dbg [[DBG54:![0-9]+]]
5507 // CHECK5-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8, !dbg [[DBG54]]
5508 // CHECK5-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8, !dbg [[DBG54]]
5509 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4, !dbg [[DBG54]]
5510 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !dbg [[DBG53]]
5511 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4, !dbg [[DBG53]]
5512 // CHECK5-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB21]], i32 [[TMP10]], i32 1073741859, i64 0, i64 16908287, i64 1, i64 1), !dbg [[DBG53]]
5513 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]], !dbg [[DBG53]]
5514 // CHECK5:       omp.dispatch.cond:
5515 // CHECK5-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB21]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]]), !dbg [[DBG53]]
5516 // CHECK5-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0, !dbg [[DBG53]]
5517 // CHECK5-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]], !dbg [[DBG53]]
5518 // CHECK5:       omp.dispatch.body:
5519 // CHECK5-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8, !dbg [[DBG54]]
5520 // CHECK5-NEXT:    store i64 [[TMP12]], i64* [[DOTOMP_IV]], align 8, !dbg [[DBG54]]
5521 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]], !dbg [[DBG53]]
5522 // CHECK5:       omp.inner.for.cond:
5523 // CHECK5-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG54]], !llvm.access.group !55
5524 // CHECK5-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !dbg [[DBG54]], !llvm.access.group !55
5525 // CHECK5-NEXT:    [[ADD:%.*]] = add i64 [[TMP14]], 1, !dbg [[DBG54]]
5526 // CHECK5-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP13]], [[ADD]], !dbg [[DBG54]]
5527 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]], !dbg [[DBG53]]
5528 // CHECK5:       omp.inner.for.body:
5529 // CHECK5-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG54]], !llvm.access.group !55
5530 // CHECK5-NEXT:    [[MUL:%.*]] = mul i64 [[TMP15]], 127, !dbg [[DBG54]]
5531 // CHECK5-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]], !dbg [[DBG54]]
5532 // CHECK5-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8, !dbg [[DBG54]], !llvm.access.group !55
5533 // CHECK5-NEXT:    [[TMP16:%.*]] = load float*, float** [[TMP4]], align 8, !dbg [[DBG56:![0-9]+]], !llvm.access.group !55
5534 // CHECK5-NEXT:    [[TMP17:%.*]] = load i64, i64* [[I]], align 8, !dbg [[DBG56]], !llvm.access.group !55
5535 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[TMP17]], !dbg [[DBG56]]
5536 // CHECK5-NEXT:    [[TMP18:%.*]] = load float, float* [[ARRAYIDX]], align 4, !dbg [[DBG56]], !llvm.access.group !55
5537 // CHECK5-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP6]], align 8, !dbg [[DBG56]], !llvm.access.group !55
5538 // CHECK5-NEXT:    [[TMP20:%.*]] = load i64, i64* [[I]], align 8, !dbg [[DBG56]], !llvm.access.group !55
5539 // CHECK5-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[TMP20]], !dbg [[DBG56]]
5540 // CHECK5-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX2]], align 4, !dbg [[DBG56]], !llvm.access.group !55
5541 // CHECK5-NEXT:    [[MUL3:%.*]] = fmul float [[TMP18]], [[TMP21]], !dbg [[DBG56]]
5542 // CHECK5-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP8]], align 8, !dbg [[DBG56]], !llvm.access.group !55
5543 // CHECK5-NEXT:    [[TMP23:%.*]] = load i64, i64* [[I]], align 8, !dbg [[DBG56]], !llvm.access.group !55
5544 // CHECK5-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[TMP23]], !dbg [[DBG56]]
5545 // CHECK5-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !dbg [[DBG56]], !llvm.access.group !55
5546 // CHECK5-NEXT:    [[MUL5:%.*]] = fmul float [[MUL3]], [[TMP24]], !dbg [[DBG56]]
5547 // CHECK5-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP2]], align 8, !dbg [[DBG56]], !llvm.access.group !55
5548 // CHECK5-NEXT:    [[TMP26:%.*]] = load i64, i64* [[I]], align 8, !dbg [[DBG56]], !llvm.access.group !55
5549 // CHECK5-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[TMP26]], !dbg [[DBG56]]
5550 // CHECK5-NEXT:    store float [[MUL5]], float* [[ARRAYIDX6]], align 4, !dbg [[DBG56]], !llvm.access.group !55
5551 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]], !dbg [[DBG57:![0-9]+]]
5552 // CHECK5:       omp.body.continue:
5553 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]], !dbg [[DBG53]]
5554 // CHECK5:       omp.inner.for.inc:
5555 // CHECK5-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG54]], !llvm.access.group !55
5556 // CHECK5-NEXT:    [[ADD7:%.*]] = add i64 [[TMP27]], 1, !dbg [[DBG54]]
5557 // CHECK5-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !dbg [[DBG54]], !llvm.access.group !55
5558 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !dbg [[DBG53]], !llvm.loop [[LOOP58:![0-9]+]]
5559 // CHECK5:       omp.inner.for.end:
5560 // CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]], !dbg [[DBG53]]
5561 // CHECK5:       omp.dispatch.inc:
5562 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]], !dbg [[DBG53]], !llvm.loop [[LOOP60:![0-9]+]]
5563 // CHECK5:       omp.dispatch.end:
5564 // CHECK5-NEXT:    ret void, !dbg [[DBG57]]
5565 //
5566 //
5567 // CHECK5-LABEL: define {{[^@]+}}@_Z7guided7PfS_S_S_
5568 // CHECK5-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] !dbg [[DBG61:![0-9]+]] {
5569 // CHECK5-NEXT:  entry:
5570 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
5571 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
5572 // CHECK5-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
5573 // CHECK5-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
5574 // CHECK5-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
5575 // CHECK5-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
5576 // CHECK5-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
5577 // CHECK5-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
5578 // CHECK5-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
5579 // CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0, !dbg [[DBG62:![0-9]+]]
5580 // CHECK5-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8, !dbg [[DBG62]]
5581 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1, !dbg [[DBG62]]
5582 // CHECK5-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8, !dbg [[DBG62]]
5583 // CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2, !dbg [[DBG62]]
5584 // CHECK5-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8, !dbg [[DBG62]]
5585 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3, !dbg [[DBG62]]
5586 // CHECK5-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8, !dbg [[DBG62]]
5587 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB23:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]]), !dbg [[DBG62]]
5588 // CHECK5-NEXT:    ret void, !dbg [[DBG63:![0-9]+]]
5589 //
5590 //
5591 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..5
5592 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] !dbg [[DBG64:![0-9]+]] {
5593 // CHECK5-NEXT:  entry:
5594 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5595 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5596 // CHECK5-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
5597 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
5598 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i64, align 8
5599 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
5600 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
5601 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
5602 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5603 // CHECK5-NEXT:    [[I:%.*]] = alloca i64, align 8
5604 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5605 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5606 // CHECK5-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
5607 // CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8, !dbg [[DBG65:![0-9]+]]
5608 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0, !dbg [[DBG65]]
5609 // CHECK5-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8, !dbg [[DBG65]]
5610 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1, !dbg [[DBG65]]
5611 // CHECK5-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8, !dbg [[DBG65]]
5612 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2, !dbg [[DBG65]]
5613 // CHECK5-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8, !dbg [[DBG65]]
5614 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3, !dbg [[DBG65]]
5615 // CHECK5-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8, !dbg [[DBG65]]
5616 // CHECK5-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8, !dbg [[DBG66:![0-9]+]]
5617 // CHECK5-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8, !dbg [[DBG66]]
5618 // CHECK5-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8, !dbg [[DBG66]]
5619 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4, !dbg [[DBG66]]
5620 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !dbg [[DBG65]]
5621 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4, !dbg [[DBG65]]
5622 // CHECK5-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB23]], i32 [[TMP10]], i32 1073741860, i64 0, i64 16908287, i64 1, i64 7), !dbg [[DBG65]]
5623 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]], !dbg [[DBG65]]
5624 // CHECK5:       omp.dispatch.cond:
5625 // CHECK5-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB23]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]]), !dbg [[DBG65]]
5626 // CHECK5-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0, !dbg [[DBG65]]
5627 // CHECK5-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]], !dbg [[DBG65]]
5628 // CHECK5:       omp.dispatch.body:
5629 // CHECK5-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8, !dbg [[DBG66]]
5630 // CHECK5-NEXT:    store i64 [[TMP12]], i64* [[DOTOMP_IV]], align 8, !dbg [[DBG66]]
5631 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]], !dbg [[DBG65]]
5632 // CHECK5:       omp.inner.for.cond:
5633 // CHECK5-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG66]], !llvm.access.group !67
5634 // CHECK5-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !dbg [[DBG66]], !llvm.access.group !67
5635 // CHECK5-NEXT:    [[ADD:%.*]] = add i64 [[TMP14]], 1, !dbg [[DBG66]]
5636 // CHECK5-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP13]], [[ADD]], !dbg [[DBG66]]
5637 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]], !dbg [[DBG65]]
5638 // CHECK5:       omp.inner.for.body:
5639 // CHECK5-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG66]], !llvm.access.group !67
5640 // CHECK5-NEXT:    [[MUL:%.*]] = mul i64 [[TMP15]], 127, !dbg [[DBG66]]
5641 // CHECK5-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]], !dbg [[DBG66]]
5642 // CHECK5-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8, !dbg [[DBG66]], !llvm.access.group !67
5643 // CHECK5-NEXT:    [[TMP16:%.*]] = load float*, float** [[TMP4]], align 8, !dbg [[DBG68:![0-9]+]], !llvm.access.group !67
5644 // CHECK5-NEXT:    [[TMP17:%.*]] = load i64, i64* [[I]], align 8, !dbg [[DBG68]], !llvm.access.group !67
5645 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[TMP17]], !dbg [[DBG68]]
5646 // CHECK5-NEXT:    [[TMP18:%.*]] = load float, float* [[ARRAYIDX]], align 4, !dbg [[DBG68]], !llvm.access.group !67
5647 // CHECK5-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP6]], align 8, !dbg [[DBG68]], !llvm.access.group !67
5648 // CHECK5-NEXT:    [[TMP20:%.*]] = load i64, i64* [[I]], align 8, !dbg [[DBG68]], !llvm.access.group !67
5649 // CHECK5-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[TMP20]], !dbg [[DBG68]]
5650 // CHECK5-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX2]], align 4, !dbg [[DBG68]], !llvm.access.group !67
5651 // CHECK5-NEXT:    [[MUL3:%.*]] = fmul float [[TMP18]], [[TMP21]], !dbg [[DBG68]]
5652 // CHECK5-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP8]], align 8, !dbg [[DBG68]], !llvm.access.group !67
5653 // CHECK5-NEXT:    [[TMP23:%.*]] = load i64, i64* [[I]], align 8, !dbg [[DBG68]], !llvm.access.group !67
5654 // CHECK5-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[TMP23]], !dbg [[DBG68]]
5655 // CHECK5-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !dbg [[DBG68]], !llvm.access.group !67
5656 // CHECK5-NEXT:    [[MUL5:%.*]] = fmul float [[MUL3]], [[TMP24]], !dbg [[DBG68]]
5657 // CHECK5-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP2]], align 8, !dbg [[DBG68]], !llvm.access.group !67
5658 // CHECK5-NEXT:    [[TMP26:%.*]] = load i64, i64* [[I]], align 8, !dbg [[DBG68]], !llvm.access.group !67
5659 // CHECK5-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[TMP26]], !dbg [[DBG68]]
5660 // CHECK5-NEXT:    store float [[MUL5]], float* [[ARRAYIDX6]], align 4, !dbg [[DBG68]], !llvm.access.group !67
5661 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]], !dbg [[DBG69:![0-9]+]]
5662 // CHECK5:       omp.body.continue:
5663 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]], !dbg [[DBG65]]
5664 // CHECK5:       omp.inner.for.inc:
5665 // CHECK5-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG66]], !llvm.access.group !67
5666 // CHECK5-NEXT:    [[ADD7:%.*]] = add i64 [[TMP27]], 1, !dbg [[DBG66]]
5667 // CHECK5-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !dbg [[DBG66]], !llvm.access.group !67
5668 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !dbg [[DBG65]], !llvm.loop [[LOOP70:![0-9]+]]
5669 // CHECK5:       omp.inner.for.end:
5670 // CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]], !dbg [[DBG65]]
5671 // CHECK5:       omp.dispatch.inc:
5672 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]], !dbg [[DBG65]], !llvm.loop [[LOOP72:![0-9]+]]
5673 // CHECK5:       omp.dispatch.end:
5674 // CHECK5-NEXT:    ret void, !dbg [[DBG69]]
5675 //
5676 //
5677 // CHECK5-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
5678 // CHECK5-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] !dbg [[DBG73:![0-9]+]] {
5679 // CHECK5-NEXT:  entry:
5680 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
5681 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
5682 // CHECK5-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
5683 // CHECK5-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
5684 // CHECK5-NEXT:    [[X:%.*]] = alloca i32, align 4
5685 // CHECK5-NEXT:    [[Y:%.*]] = alloca i32, align 4
5686 // CHECK5-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
5687 // CHECK5-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
5688 // CHECK5-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
5689 // CHECK5-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
5690 // CHECK5-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
5691 // CHECK5-NEXT:    store i32 0, i32* [[X]], align 4, !dbg [[DBG74:![0-9]+]]
5692 // CHECK5-NEXT:    store i32 0, i32* [[Y]], align 4, !dbg [[DBG75:![0-9]+]]
5693 // CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0, !dbg [[DBG76:![0-9]+]]
5694 // CHECK5-NEXT:    store i32* [[Y]], i32** [[TMP0]], align 8, !dbg [[DBG76]]
5695 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1, !dbg [[DBG76]]
5696 // CHECK5-NEXT:    store float** [[A_ADDR]], float*** [[TMP1]], align 8, !dbg [[DBG76]]
5697 // CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2, !dbg [[DBG76]]
5698 // CHECK5-NEXT:    store float** [[B_ADDR]], float*** [[TMP2]], align 8, !dbg [[DBG76]]
5699 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3, !dbg [[DBG76]]
5700 // CHECK5-NEXT:    store float** [[C_ADDR]], float*** [[TMP3]], align 8, !dbg [[DBG76]]
5701 // CHECK5-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4, !dbg [[DBG76]]
5702 // CHECK5-NEXT:    store float** [[D_ADDR]], float*** [[TMP4]], align 8, !dbg [[DBG76]]
5703 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB25:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]]), !dbg [[DBG76]]
5704 // CHECK5-NEXT:    ret void, !dbg [[DBG77:![0-9]+]]
5705 //
5706 //
5707 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..6
5708 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] !dbg [[DBG78:![0-9]+]] {
5709 // CHECK5-NEXT:  entry:
5710 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5711 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5712 // CHECK5-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
5713 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
5714 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i8, align 1
5715 // CHECK5-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
5716 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
5717 // CHECK5-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
5718 // CHECK5-NEXT:    [[I:%.*]] = alloca i8, align 1
5719 // CHECK5-NEXT:    [[X:%.*]] = alloca i32, align 4
5720 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
5721 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
5722 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
5723 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5724 // CHECK5-NEXT:    [[I7:%.*]] = alloca i8, align 1
5725 // CHECK5-NEXT:    [[X8:%.*]] = alloca i32, align 4
5726 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5727 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5728 // CHECK5-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
5729 // CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8, !dbg [[DBG79:![0-9]+]]
5730 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0, !dbg [[DBG79]]
5731 // CHECK5-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8, !dbg [[DBG79]]
5732 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1, !dbg [[DBG79]]
5733 // CHECK5-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8, !dbg [[DBG79]]
5734 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2, !dbg [[DBG79]]
5735 // CHECK5-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8, !dbg [[DBG79]]
5736 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3, !dbg [[DBG79]]
5737 // CHECK5-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8, !dbg [[DBG79]]
5738 // CHECK5-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4, !dbg [[DBG79]]
5739 // CHECK5-NEXT:    [[TMP10:%.*]] = load float**, float*** [[TMP9]], align 8, !dbg [[DBG79]]
5740 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP2]], align 4, !dbg [[DBG80:![0-9]+]]
5741 // CHECK5-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP11]] to i8, !dbg [[DBG80]]
5742 // CHECK5-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1, !dbg [[DBG80]]
5743 // CHECK5-NEXT:    [[TMP12:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !dbg [[DBG80]]
5744 // CHECK5-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP12]] to i32, !dbg [[DBG80]]
5745 // CHECK5-NEXT:    [[SUB:%.*]] = sub i32 57, [[CONV3]], !dbg [[DBG80]]
5746 // CHECK5-NEXT:    [[ADD:%.*]] = add i32 [[SUB]], 1, !dbg [[DBG80]]
5747 // CHECK5-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1, !dbg [[DBG80]]
5748 // CHECK5-NEXT:    [[CONV4:%.*]] = zext i32 [[DIV]] to i64, !dbg [[DBG80]]
5749 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV4]], 11, !dbg [[DBG81:![0-9]+]]
5750 // CHECK5-NEXT:    [[SUB5:%.*]] = sub nsw i64 [[MUL]], 1, !dbg [[DBG81]]
5751 // CHECK5-NEXT:    store i64 [[SUB5]], i64* [[DOTCAPTURE_EXPR_2]], align 8, !dbg [[DBG80]]
5752 // CHECK5-NEXT:    [[TMP13:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !dbg [[DBG80]]
5753 // CHECK5-NEXT:    store i8 [[TMP13]], i8* [[I]], align 1, !dbg [[DBG80]]
5754 // CHECK5-NEXT:    store i32 11, i32* [[X]], align 4, !dbg [[DBG81]]
5755 // CHECK5-NEXT:    [[TMP14:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !dbg [[DBG80]]
5756 // CHECK5-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP14]] to i32, !dbg [[DBG80]]
5757 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[CONV6]], 57, !dbg [[DBG80]]
5758 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]], !dbg [[DBG79]]
5759 // CHECK5:       omp.precond.then:
5760 // CHECK5-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8, !dbg [[DBG80]]
5761 // CHECK5-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8, !dbg [[DBG81]]
5762 // CHECK5-NEXT:    store i64 [[TMP15]], i64* [[DOTOMP_UB]], align 8, !dbg [[DBG80]]
5763 // CHECK5-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8, !dbg [[DBG80]]
5764 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4, !dbg [[DBG80]]
5765 // CHECK5-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8, !dbg [[DBG81]]
5766 // CHECK5-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !dbg [[DBG79]]
5767 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4, !dbg [[DBG79]]
5768 // CHECK5-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB25]], i32 [[TMP18]], i32 1073741862, i64 0, i64 [[TMP16]], i64 1, i64 1), !dbg [[DBG79]]
5769 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]], !dbg [[DBG79]]
5770 // CHECK5:       omp.dispatch.cond:
5771 // CHECK5-NEXT:    [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !dbg [[DBG79]]
5772 // CHECK5-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4, !dbg [[DBG79]]
5773 // CHECK5-NEXT:    [[TMP21:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB25]], i32 [[TMP20]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]]), !dbg [[DBG79]]
5774 // CHECK5-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP21]], 0, !dbg [[DBG79]]
5775 // CHECK5-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]], !dbg [[DBG79]]
5776 // CHECK5:       omp.dispatch.body:
5777 // CHECK5-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8, !dbg [[DBG80]]
5778 // CHECK5-NEXT:    store i64 [[TMP22]], i64* [[DOTOMP_IV]], align 8, !dbg [[DBG80]]
5779 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]], !dbg [[DBG79]]
5780 // CHECK5:       omp.inner.for.cond:
5781 // CHECK5-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG80]], !llvm.access.group !82
5782 // CHECK5-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !dbg [[DBG80]], !llvm.access.group !82
5783 // CHECK5-NEXT:    [[CMP9:%.*]] = icmp sle i64 [[TMP23]], [[TMP24]], !dbg [[DBG80]]
5784 // CHECK5-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]], !dbg [[DBG79]]
5785 // CHECK5:       omp.inner.for.body:
5786 // CHECK5-NEXT:    [[TMP25:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !dbg [[DBG80]], !llvm.access.group !82
5787 // CHECK5-NEXT:    [[CONV10:%.*]] = sext i8 [[TMP25]] to i64, !dbg [[DBG80]]
5788 // CHECK5-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG80]], !llvm.access.group !82
5789 // CHECK5-NEXT:    [[DIV11:%.*]] = sdiv i64 [[TMP26]], 11, !dbg [[DBG80]]
5790 // CHECK5-NEXT:    [[MUL12:%.*]] = mul nsw i64 [[DIV11]], 1, !dbg [[DBG80]]
5791 // CHECK5-NEXT:    [[ADD13:%.*]] = add nsw i64 [[CONV10]], [[MUL12]], !dbg [[DBG80]]
5792 // CHECK5-NEXT:    [[CONV14:%.*]] = trunc i64 [[ADD13]] to i8, !dbg [[DBG80]]
5793 // CHECK5-NEXT:    store i8 [[CONV14]], i8* [[I7]], align 1, !dbg [[DBG80]], !llvm.access.group !82
5794 // CHECK5-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG80]], !llvm.access.group !82
5795 // CHECK5-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG80]], !llvm.access.group !82
5796 // CHECK5-NEXT:    [[DIV15:%.*]] = sdiv i64 [[TMP28]], 11, !dbg [[DBG80]]
5797 // CHECK5-NEXT:    [[MUL16:%.*]] = mul nsw i64 [[DIV15]], 11, !dbg [[DBG80]]
5798 // CHECK5-NEXT:    [[SUB17:%.*]] = sub nsw i64 [[TMP27]], [[MUL16]], !dbg [[DBG80]]
5799 // CHECK5-NEXT:    [[MUL18:%.*]] = mul nsw i64 [[SUB17]], 1, !dbg [[DBG81]]
5800 // CHECK5-NEXT:    [[SUB19:%.*]] = sub nsw i64 11, [[MUL18]], !dbg [[DBG81]]
5801 // CHECK5-NEXT:    [[CONV20:%.*]] = trunc i64 [[SUB19]] to i32, !dbg [[DBG81]]
5802 // CHECK5-NEXT:    store i32 [[CONV20]], i32* [[X8]], align 4, !dbg [[DBG81]], !llvm.access.group !82
5803 // CHECK5-NEXT:    [[TMP29:%.*]] = load float*, float** [[TMP6]], align 8, !dbg [[DBG83:![0-9]+]], !llvm.access.group !82
5804 // CHECK5-NEXT:    [[TMP30:%.*]] = load i8, i8* [[I7]], align 1, !dbg [[DBG83]], !llvm.access.group !82
5805 // CHECK5-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP30]] to i64, !dbg [[DBG83]]
5806 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP29]], i64 [[IDXPROM]], !dbg [[DBG83]]
5807 // CHECK5-NEXT:    [[TMP31:%.*]] = load float, float* [[ARRAYIDX]], align 4, !dbg [[DBG83]], !llvm.access.group !82
5808 // CHECK5-NEXT:    [[TMP32:%.*]] = load float*, float** [[TMP8]], align 8, !dbg [[DBG83]], !llvm.access.group !82
5809 // CHECK5-NEXT:    [[TMP33:%.*]] = load i8, i8* [[I7]], align 1, !dbg [[DBG83]], !llvm.access.group !82
5810 // CHECK5-NEXT:    [[IDXPROM21:%.*]] = sext i8 [[TMP33]] to i64, !dbg [[DBG83]]
5811 // CHECK5-NEXT:    [[ARRAYIDX22:%.*]] = getelementptr inbounds float, float* [[TMP32]], i64 [[IDXPROM21]], !dbg [[DBG83]]
5812 // CHECK5-NEXT:    [[TMP34:%.*]] = load float, float* [[ARRAYIDX22]], align 4, !dbg [[DBG83]], !llvm.access.group !82
5813 // CHECK5-NEXT:    [[MUL23:%.*]] = fmul float [[TMP31]], [[TMP34]], !dbg [[DBG83]]
5814 // CHECK5-NEXT:    [[TMP35:%.*]] = load float*, float** [[TMP10]], align 8, !dbg [[DBG83]], !llvm.access.group !82
5815 // CHECK5-NEXT:    [[TMP36:%.*]] = load i8, i8* [[I7]], align 1, !dbg [[DBG83]], !llvm.access.group !82
5816 // CHECK5-NEXT:    [[IDXPROM24:%.*]] = sext i8 [[TMP36]] to i64, !dbg [[DBG83]]
5817 // CHECK5-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds float, float* [[TMP35]], i64 [[IDXPROM24]], !dbg [[DBG83]]
5818 // CHECK5-NEXT:    [[TMP37:%.*]] = load float, float* [[ARRAYIDX25]], align 4, !dbg [[DBG83]], !llvm.access.group !82
5819 // CHECK5-NEXT:    [[MUL26:%.*]] = fmul float [[MUL23]], [[TMP37]], !dbg [[DBG83]]
5820 // CHECK5-NEXT:    [[TMP38:%.*]] = load float*, float** [[TMP4]], align 8, !dbg [[DBG83]], !llvm.access.group !82
5821 // CHECK5-NEXT:    [[TMP39:%.*]] = load i8, i8* [[I7]], align 1, !dbg [[DBG83]], !llvm.access.group !82
5822 // CHECK5-NEXT:    [[IDXPROM27:%.*]] = sext i8 [[TMP39]] to i64, !dbg [[DBG83]]
5823 // CHECK5-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds float, float* [[TMP38]], i64 [[IDXPROM27]], !dbg [[DBG83]]
5824 // CHECK5-NEXT:    store float [[MUL26]], float* [[ARRAYIDX28]], align 4, !dbg [[DBG83]], !llvm.access.group !82
5825 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]], !dbg [[DBG84:![0-9]+]]
5826 // CHECK5:       omp.body.continue:
5827 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]], !dbg [[DBG79]]
5828 // CHECK5:       omp.inner.for.inc:
5829 // CHECK5-NEXT:    [[TMP40:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !dbg [[DBG80]], !llvm.access.group !82
5830 // CHECK5-NEXT:    [[ADD29:%.*]] = add nsw i64 [[TMP40]], 1, !dbg [[DBG80]]
5831 // CHECK5-NEXT:    store i64 [[ADD29]], i64* [[DOTOMP_IV]], align 8, !dbg [[DBG80]], !llvm.access.group !82
5832 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !dbg [[DBG79]], !llvm.loop [[LOOP85:![0-9]+]]
5833 // CHECK5:       omp.inner.for.end:
5834 // CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]], !dbg [[DBG79]]
5835 // CHECK5:       omp.dispatch.inc:
5836 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]], !dbg [[DBG79]], !llvm.loop [[LOOP87:![0-9]+]]
5837 // CHECK5:       omp.dispatch.end:
5838 // CHECK5-NEXT:    br label [[OMP_PRECOND_END]], !dbg [[DBG79]]
5839 // CHECK5:       omp.precond.end:
5840 // CHECK5-NEXT:    ret void, !dbg [[DBG84]]
5841 //
5842 //
5843 // CHECK5-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
5844 // CHECK5-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] !dbg [[DBG88:![0-9]+]] {
5845 // CHECK5-NEXT:  entry:
5846 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
5847 // CHECK5-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
5848 // CHECK5-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
5849 // CHECK5-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
5850 // CHECK5-NEXT:    [[X:%.*]] = alloca i32, align 4
5851 // CHECK5-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 8
5852 // CHECK5-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
5853 // CHECK5-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
5854 // CHECK5-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
5855 // CHECK5-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
5856 // CHECK5-NEXT:    store i32 0, i32* [[X]], align 4, !dbg [[DBG89:![0-9]+]]
5857 // CHECK5-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0, !dbg [[DBG90:![0-9]+]]
5858 // CHECK5-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8, !dbg [[DBG90]]
5859 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1, !dbg [[DBG90]]
5860 // CHECK5-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8, !dbg [[DBG90]]
5861 // CHECK5-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2, !dbg [[DBG90]]
5862 // CHECK5-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8, !dbg [[DBG90]]
5863 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3, !dbg [[DBG90]]
5864 // CHECK5-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8, !dbg [[DBG90]]
5865 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB27:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.6*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]]), !dbg [[DBG90]]
5866 // CHECK5-NEXT:    ret void, !dbg [[DBG91:![0-9]+]]
5867 //
5868 //
5869 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..7
5870 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] !dbg [[DBG92:![0-9]+]] {
5871 // CHECK5-NEXT:  entry:
5872 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
5873 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
5874 // CHECK5-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 8
5875 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
5876 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i8, align 1
5877 // CHECK5-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
5878 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
5879 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
5880 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
5881 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
5882 // CHECK5-NEXT:    [[I:%.*]] = alloca i8, align 1
5883 // CHECK5-NEXT:    [[X:%.*]] = alloca i32, align 4
5884 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
5885 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
5886 // CHECK5-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 8
5887 // CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 8, !dbg [[DBG93:![0-9]+]]
5888 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0, !dbg [[DBG93]]
5889 // CHECK5-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8, !dbg [[DBG93]]
5890 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1, !dbg [[DBG93]]
5891 // CHECK5-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8, !dbg [[DBG93]]
5892 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2, !dbg [[DBG93]]
5893 // CHECK5-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8, !dbg [[DBG93]]
5894 // CHECK5-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3, !dbg [[DBG93]]
5895 // CHECK5-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8, !dbg [[DBG93]]
5896 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4, !dbg [[DBG94:![0-9]+]]
5897 // CHECK5-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG94]]
5898 // CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4, !dbg [[DBG94]]
5899 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4, !dbg [[DBG94]]
5900 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !dbg [[DBG93]]
5901 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4, !dbg [[DBG93]]
5902 // CHECK5-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB27]], i32 [[TMP10]], i32 1073741861, i32 0, i32 199, i32 1, i32 1), !dbg [[DBG93]]
5903 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]], !dbg [[DBG93]]
5904 // CHECK5:       omp.dispatch.cond:
5905 // CHECK5-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB27]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]]), !dbg [[DBG93]]
5906 // CHECK5-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0, !dbg [[DBG93]]
5907 // CHECK5-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]], !dbg [[DBG93]]
5908 // CHECK5:       omp.dispatch.body:
5909 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4, !dbg [[DBG94]]
5910 // CHECK5-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4, !dbg [[DBG94]]
5911 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]], !dbg [[DBG93]]
5912 // CHECK5:       omp.inner.for.cond:
5913 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG94]], !llvm.access.group !95
5914 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG94]], !llvm.access.group !95
5915 // CHECK5-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]], !dbg [[DBG94]]
5916 // CHECK5-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]], !dbg [[DBG93]]
5917 // CHECK5:       omp.inner.for.body:
5918 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG94]], !llvm.access.group !95
5919 // CHECK5-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP15]], 20, !dbg [[DBG94]]
5920 // CHECK5-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1, !dbg [[DBG94]]
5921 // CHECK5-NEXT:    [[ADD:%.*]] = add nsw i32 48, [[MUL]], !dbg [[DBG94]]
5922 // CHECK5-NEXT:    [[CONV:%.*]] = trunc i32 [[ADD]] to i8, !dbg [[DBG94]]
5923 // CHECK5-NEXT:    store i8 [[CONV]], i8* [[I]], align 1, !dbg [[DBG94]], !llvm.access.group !95
5924 // CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG94]], !llvm.access.group !95
5925 // CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG94]], !llvm.access.group !95
5926 // CHECK5-NEXT:    [[DIV2:%.*]] = sdiv i32 [[TMP17]], 20, !dbg [[DBG94]]
5927 // CHECK5-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[DIV2]], 20, !dbg [[DBG94]]
5928 // CHECK5-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP16]], [[MUL3]], !dbg [[DBG94]]
5929 // CHECK5-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[SUB]], 1, !dbg [[DBG96:![0-9]+]]
5930 // CHECK5-NEXT:    [[ADD5:%.*]] = add nsw i32 -10, [[MUL4]], !dbg [[DBG96]]
5931 // CHECK5-NEXT:    store i32 [[ADD5]], i32* [[X]], align 4, !dbg [[DBG96]], !llvm.access.group !95
5932 // CHECK5-NEXT:    [[TMP18:%.*]] = load float*, float** [[TMP4]], align 8, !dbg [[DBG97:![0-9]+]], !llvm.access.group !95
5933 // CHECK5-NEXT:    [[TMP19:%.*]] = load i8, i8* [[I]], align 1, !dbg [[DBG97]], !llvm.access.group !95
5934 // CHECK5-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP19]] to i64, !dbg [[DBG97]]
5935 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP18]], i64 [[IDXPROM]], !dbg [[DBG97]]
5936 // CHECK5-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !dbg [[DBG97]], !llvm.access.group !95
5937 // CHECK5-NEXT:    [[TMP21:%.*]] = load float*, float** [[TMP6]], align 8, !dbg [[DBG97]], !llvm.access.group !95
5938 // CHECK5-NEXT:    [[TMP22:%.*]] = load i8, i8* [[I]], align 1, !dbg [[DBG97]], !llvm.access.group !95
5939 // CHECK5-NEXT:    [[IDXPROM6:%.*]] = zext i8 [[TMP22]] to i64, !dbg [[DBG97]]
5940 // CHECK5-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP21]], i64 [[IDXPROM6]], !dbg [[DBG97]]
5941 // CHECK5-NEXT:    [[TMP23:%.*]] = load float, float* [[ARRAYIDX7]], align 4, !dbg [[DBG97]], !llvm.access.group !95
5942 // CHECK5-NEXT:    [[MUL8:%.*]] = fmul float [[TMP20]], [[TMP23]], !dbg [[DBG97]]
5943 // CHECK5-NEXT:    [[TMP24:%.*]] = load float*, float** [[TMP8]], align 8, !dbg [[DBG97]], !llvm.access.group !95
5944 // CHECK5-NEXT:    [[TMP25:%.*]] = load i8, i8* [[I]], align 1, !dbg [[DBG97]], !llvm.access.group !95
5945 // CHECK5-NEXT:    [[IDXPROM9:%.*]] = zext i8 [[TMP25]] to i64, !dbg [[DBG97]]
5946 // CHECK5-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP24]], i64 [[IDXPROM9]], !dbg [[DBG97]]
5947 // CHECK5-NEXT:    [[TMP26:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !dbg [[DBG97]], !llvm.access.group !95
5948 // CHECK5-NEXT:    [[MUL11:%.*]] = fmul float [[MUL8]], [[TMP26]], !dbg [[DBG97]]
5949 // CHECK5-NEXT:    [[TMP27:%.*]] = load float*, float** [[TMP2]], align 8, !dbg [[DBG97]], !llvm.access.group !95
5950 // CHECK5-NEXT:    [[TMP28:%.*]] = load i8, i8* [[I]], align 1, !dbg [[DBG97]], !llvm.access.group !95
5951 // CHECK5-NEXT:    [[IDXPROM12:%.*]] = zext i8 [[TMP28]] to i64, !dbg [[DBG97]]
5952 // CHECK5-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds float, float* [[TMP27]], i64 [[IDXPROM12]], !dbg [[DBG97]]
5953 // CHECK5-NEXT:    store float [[MUL11]], float* [[ARRAYIDX13]], align 4, !dbg [[DBG97]], !llvm.access.group !95
5954 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]], !dbg [[DBG98:![0-9]+]]
5955 // CHECK5:       omp.body.continue:
5956 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]], !dbg [[DBG93]]
5957 // CHECK5:       omp.inner.for.inc:
5958 // CHECK5-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG94]], !llvm.access.group !95
5959 // CHECK5-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP29]], 1, !dbg [[DBG94]]
5960 // CHECK5-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV]], align 4, !dbg [[DBG94]], !llvm.access.group !95
5961 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !dbg [[DBG93]], !llvm.loop [[LOOP99:![0-9]+]]
5962 // CHECK5:       omp.inner.for.end:
5963 // CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]], !dbg [[DBG93]]
5964 // CHECK5:       omp.dispatch.inc:
5965 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]], !dbg [[DBG93]], !llvm.loop [[LOOP101:![0-9]+]]
5966 // CHECK5:       omp.dispatch.end:
5967 // CHECK5-NEXT:    ret void, !dbg [[DBG98]]
5968 //
5969 //
5970 // CHECK5-LABEL: define {{[^@]+}}@_Z3foov
5971 // CHECK5-SAME: () #[[ATTR3:[0-9]+]] !dbg [[DBG102:![0-9]+]] {
5972 // CHECK5-NEXT:  entry:
5973 // CHECK5-NEXT:    call void @_Z8mayThrowv(), !dbg [[DBG103:![0-9]+]]
5974 // CHECK5-NEXT:    ret i32 0, !dbg [[DBG103]]
5975 //
5976 //
5977 // CHECK5-LABEL: define {{[^@]+}}@_Z12parallel_forPfi
5978 // CHECK5-SAME: (float* [[A:%.*]], i32 [[N:%.*]]) #[[ATTR0]] !dbg [[DBG104:![0-9]+]] {
5979 // CHECK5-NEXT:  entry:
5980 // CHECK5-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
5981 // CHECK5-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
5982 // CHECK5-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
5983 // CHECK5-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
5984 // CHECK5-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 8
5985 // CHECK5-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
5986 // CHECK5-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
5987 // CHECK5-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4, !dbg [[DBG105:![0-9]+]]
5988 // CHECK5-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64, !dbg [[DBG105]]
5989 // CHECK5-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave(), !dbg [[DBG105]]
5990 // CHECK5-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8, !dbg [[DBG105]]
5991 // CHECK5-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP1]], align 16, !dbg [[DBG105]]
5992 // CHECK5-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8, !dbg [[DBG105]]
5993 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0, !dbg [[DBG106:![0-9]+]]
5994 // CHECK5-NEXT:    store float** [[A_ADDR]], float*** [[TMP3]], align 8, !dbg [[DBG106]]
5995 // CHECK5-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1, !dbg [[DBG106]]
5996 // CHECK5-NEXT:    store i64 [[TMP1]], i64* [[TMP4]], align 8, !dbg [[DBG106]]
5997 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2, !dbg [[DBG106]]
5998 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4, !dbg [[DBG107:![0-9]+]]
5999 // CHECK5-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8, !dbg [[DBG106]]
6000 // CHECK5-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB32:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.7*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]]), !dbg [[DBG106]]
6001 // CHECK5-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8, !dbg [[DBG108:![0-9]+]]
6002 // CHECK5-NEXT:    call void @llvm.stackrestore(i8* [[TMP7]]), !dbg [[DBG108]]
6003 // CHECK5-NEXT:    ret void, !dbg [[DBG108]]
6004 //
6005 //
6006 // CHECK5-LABEL: define {{[^@]+}}@.omp_outlined..8
6007 // CHECK5-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) !dbg [[DBG109:![0-9]+]] {
6008 // CHECK5-NEXT:  entry:
6009 // CHECK5-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6010 // CHECK5-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6011 // CHECK5-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 8
6012 // CHECK5-NEXT:    [[N:%.*]] = alloca i32, align 4
6013 // CHECK5-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6014 // CHECK5-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6015 // CHECK5-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6016 // CHECK5-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6017 // CHECK5-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6018 // CHECK5-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6019 // CHECK5-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
6020 // CHECK5-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
6021 // CHECK5-NEXT:    [[I:%.*]] = alloca i32, align 4
6022 // CHECK5-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6023 // CHECK5-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6024 // CHECK5-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 8
6025 // CHECK5-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 8, !dbg [[DBG110:![0-9]+]]
6026 // CHECK5-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0, !dbg [[DBG110]]
6027 // CHECK5-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8, !dbg [[DBG110]]
6028 // CHECK5-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1, !dbg [[DBG110]]
6029 // CHECK5-NEXT:    [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8, !dbg [[DBG110]]
6030 // CHECK5-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2, !dbg [[DBG110]]
6031 // CHECK5-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 8, !dbg [[DBG110]]
6032 // CHECK5-NEXT:    store i32 [[TMP6]], i32* [[N]], align 4, !dbg [[DBG110]]
6033 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4, !dbg [[DBG111:![0-9]+]]
6034 // CHECK5-NEXT:    store i32 16908288, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG111]]
6035 // CHECK5-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4, !dbg [[DBG111]]
6036 // CHECK5-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4, !dbg [[DBG111]]
6037 // CHECK5-NEXT:    [[TMP7:%.*]] = call i8* @llvm.stacksave(), !dbg [[DBG110]]
6038 // CHECK5-NEXT:    store i8* [[TMP7]], i8** [[SAVED_STACK]], align 8, !dbg [[DBG110]]
6039 // CHECK5-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP4]], align 16, !dbg [[DBG110]]
6040 // CHECK5-NEXT:    store i64 [[TMP4]], i64* [[__VLA_EXPR0]], align 8, !dbg [[DBG110]]
6041 // CHECK5-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8, !dbg [[DBG110]]
6042 // CHECK5-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4, !dbg [[DBG110]]
6043 // CHECK5-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB29:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 5), !dbg [[DBG110]]
6044 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND:%.*]], !dbg [[DBG110]]
6045 // CHECK5:       omp.dispatch.cond:
6046 // CHECK5-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG111]]
6047 // CHECK5-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP10]], 16908288, !dbg [[DBG111]]
6048 // CHECK5-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]], !dbg [[DBG111]]
6049 // CHECK5:       cond.true:
6050 // CHECK5-NEXT:    br label [[COND_END:%.*]], !dbg [[DBG111]]
6051 // CHECK5:       cond.false:
6052 // CHECK5-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG111]]
6053 // CHECK5-NEXT:    br label [[COND_END]], !dbg [[DBG111]]
6054 // CHECK5:       cond.end:
6055 // CHECK5-NEXT:    [[COND:%.*]] = phi i32 [ 16908288, [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ], !dbg [[DBG111]]
6056 // CHECK5-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4, !dbg [[DBG111]]
6057 // CHECK5-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4, !dbg [[DBG111]]
6058 // CHECK5-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4, !dbg [[DBG111]]
6059 // CHECK5-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG111]]
6060 // CHECK5-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG111]]
6061 // CHECK5-NEXT:    [[CMP1:%.*]] = icmp ule i32 [[TMP13]], [[TMP14]], !dbg [[DBG111]]
6062 // CHECK5-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_CLEANUP:%.*]], !dbg [[DBG110]]
6063 // CHECK5:       omp.dispatch.cleanup:
6064 // CHECK5-NEXT:    br label [[OMP_DISPATCH_END:%.*]], !dbg [[DBG110]]
6065 // CHECK5:       omp.dispatch.body:
6066 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]], !dbg [[DBG110]]
6067 // CHECK5:       omp.inner.for.cond:
6068 // CHECK5-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG111]]
6069 // CHECK5-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG111]]
6070 // CHECK5-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[TMP15]], [[TMP16]], !dbg [[DBG111]]
6071 // CHECK5-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]], !dbg [[DBG110]]
6072 // CHECK5:       omp.inner.for.cond.cleanup:
6073 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_END:%.*]], !dbg [[DBG110]]
6074 // CHECK5:       omp.inner.for.body:
6075 // CHECK5-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG111]]
6076 // CHECK5-NEXT:    [[MUL:%.*]] = mul i32 [[TMP17]], 127, !dbg [[DBG111]]
6077 // CHECK5-NEXT:    [[ADD:%.*]] = add i32 131071, [[MUL]], !dbg [[DBG111]]
6078 // CHECK5-NEXT:    store i32 [[ADD]], i32* [[I]], align 4, !dbg [[DBG111]]
6079 // CHECK5-NEXT:    [[CALL:%.*]] = invoke i32 @_Z3foov()
6080 // CHECK5-NEXT:    to label [[INVOKE_CONT:%.*]] unwind label [[TERMINATE_LPAD:%.*]], !dbg [[DBG112:![0-9]+]]
6081 // CHECK5:       invoke.cont:
6082 // CHECK5-NEXT:    [[CONV:%.*]] = sitofp i32 [[CALL]] to float, !dbg [[DBG112]]
6083 // CHECK5-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG112]]
6084 // CHECK5-NEXT:    [[IDXPROM:%.*]] = zext i32 [[TMP18]] to i64, !dbg [[DBG112]]
6085 // CHECK5-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[VLA]], i64 [[IDXPROM]], !dbg [[DBG112]]
6086 // CHECK5-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4, !dbg [[DBG112]]
6087 // CHECK5-NEXT:    [[ADD3:%.*]] = fadd float [[CONV]], [[TMP19]], !dbg [[DBG112]]
6088 // CHECK5-NEXT:    [[TMP20:%.*]] = load i32, i32* [[N]], align 4, !dbg [[DBG112]]
6089 // CHECK5-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP20]] to float, !dbg [[DBG112]]
6090 // CHECK5-NEXT:    [[ADD5:%.*]] = fadd float [[ADD3]], [[CONV4]], !dbg [[DBG112]]
6091 // CHECK5-NEXT:    [[TMP21:%.*]] = load float*, float** [[TMP2]], align 8, !dbg [[DBG112]]
6092 // CHECK5-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I]], align 4, !dbg [[DBG112]]
6093 // CHECK5-NEXT:    [[IDXPROM6:%.*]] = zext i32 [[TMP22]] to i64, !dbg [[DBG112]]
6094 // CHECK5-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP21]], i64 [[IDXPROM6]], !dbg [[DBG112]]
6095 // CHECK5-NEXT:    [[TMP23:%.*]] = load float, float* [[ARRAYIDX7]], align 4, !dbg [[DBG112]]
6096 // CHECK5-NEXT:    [[ADD8:%.*]] = fadd float [[TMP23]], [[ADD5]], !dbg [[DBG112]]
6097 // CHECK5-NEXT:    store float [[ADD8]], float* [[ARRAYIDX7]], align 4, !dbg [[DBG112]]
6098 // CHECK5-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]], !dbg [[DBG112]]
6099 // CHECK5:       omp.body.continue:
6100 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]], !dbg [[DBG110]]
6101 // CHECK5:       omp.inner.for.inc:
6102 // CHECK5-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !dbg [[DBG111]]
6103 // CHECK5-NEXT:    [[ADD9:%.*]] = add i32 [[TMP24]], 1, !dbg [[DBG111]]
6104 // CHECK5-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4, !dbg [[DBG111]]
6105 // CHECK5-NEXT:    br label [[OMP_INNER_FOR_COND]], !dbg [[DBG110]], !llvm.loop [[LOOP113:![0-9]+]]
6106 // CHECK5:       omp.inner.for.end:
6107 // CHECK5-NEXT:    br label [[OMP_DISPATCH_INC:%.*]], !dbg [[DBG110]]
6108 // CHECK5:       omp.dispatch.inc:
6109 // CHECK5-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4, !dbg [[DBG111]]
6110 // CHECK5-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !dbg [[DBG111]]
6111 // CHECK5-NEXT:    [[ADD10:%.*]] = add i32 [[TMP25]], [[TMP26]], !dbg [[DBG111]]
6112 // CHECK5-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_LB]], align 4, !dbg [[DBG111]]
6113 // CHECK5-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !dbg [[DBG111]]
6114 // CHECK5-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4, !dbg [[DBG111]]
6115 // CHECK5-NEXT:    [[ADD11:%.*]] = add i32 [[TMP27]], [[TMP28]], !dbg [[DBG111]]
6116 // CHECK5-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_UB]], align 4, !dbg [[DBG111]]
6117 // CHECK5-NEXT:    br label [[OMP_DISPATCH_COND]], !dbg [[DBG110]], !llvm.loop [[LOOP114:![0-9]+]]
6118 // CHECK5:       omp.dispatch.end:
6119 // CHECK5-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB31:[0-9]+]], i32 [[TMP9]]), !dbg [[DBG110]]
6120 // CHECK5-NEXT:    [[TMP29:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8, !dbg [[DBG110]]
6121 // CHECK5-NEXT:    call void @llvm.stackrestore(i8* [[TMP29]]), !dbg [[DBG110]]
6122 // CHECK5-NEXT:    ret void, !dbg [[DBG112]]
6123 // CHECK5:       terminate.lpad:
6124 // CHECK5-NEXT:    [[TMP30:%.*]] = landingpad { i8*, i32 }
6125 // CHECK5-NEXT:    catch i8* null, !dbg [[DBG112]]
6126 // CHECK5-NEXT:    [[TMP31:%.*]] = extractvalue { i8*, i32 } [[TMP30]], 0, !dbg [[DBG112]]
6127 // CHECK5-NEXT:    call void @__clang_call_terminate(i8* [[TMP31]]) #[[ATTR7:[0-9]+]], !dbg [[DBG112]]
6128 // CHECK5-NEXT:    unreachable, !dbg [[DBG112]]
6129 //
6130 //
6131 // CHECK5-LABEL: define {{[^@]+}}@__clang_call_terminate
6132 // CHECK5-SAME: (i8* [[TMP0:%.*]]) #[[ATTR6:[0-9]+]] {
6133 // CHECK5-NEXT:    [[TMP2:%.*]] = call i8* @__cxa_begin_catch(i8* [[TMP0]]) #[[ATTR2:[0-9]+]]
6134 // CHECK5-NEXT:    call void @_ZSt9terminatev() #[[ATTR7]]
6135 // CHECK5-NEXT:    unreachable
6136 //
6137 //
6138 // CHECK6-LABEL: define {{[^@]+}}@_Z17with_var_schedulev
6139 // CHECK6-SAME: () #[[ATTR0:[0-9]+]] {
6140 // CHECK6-NEXT:  entry:
6141 // CHECK6-NEXT:    [[A:%.*]] = alloca double, align 8
6142 // CHECK6-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
6143 // CHECK6-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 1
6144 // CHECK6-NEXT:    store double 5.000000e+00, double* [[A]], align 8
6145 // CHECK6-NEXT:    [[TMP0:%.*]] = load double, double* [[A]], align 8
6146 // CHECK6-NEXT:    [[CONV:%.*]] = fptosi double [[TMP0]] to i8
6147 // CHECK6-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
6148 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6149 // CHECK6-NEXT:    [[TMP2:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
6150 // CHECK6-NEXT:    store i8 [[TMP2]], i8* [[TMP1]], align 1
6151 // CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
6152 // CHECK6-NEXT:    ret void
6153 //
6154 //
6155 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined.
6156 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] {
6157 // CHECK6-NEXT:  entry:
6158 // CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6159 // CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6160 // CHECK6-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
6161 // CHECK6-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
6162 // CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
6163 // CHECK6-NEXT:    [[TMP:%.*]] = alloca i64, align 8
6164 // CHECK6-NEXT:    [[DOTCAPTURE_EXPR_1:%.*]] = alloca double, align 8
6165 // CHECK6-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
6166 // CHECK6-NEXT:    [[I:%.*]] = alloca i64, align 8
6167 // CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
6168 // CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
6169 // CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
6170 // CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6171 // CHECK6-NEXT:    [[A:%.*]] = alloca double, align 8
6172 // CHECK6-NEXT:    [[I4:%.*]] = alloca i64, align 8
6173 // CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6174 // CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6175 // CHECK6-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
6176 // CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
6177 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
6178 // CHECK6-NEXT:    [[TMP2:%.*]] = load i8, i8* [[TMP1]], align 1
6179 // CHECK6-NEXT:    store i8 [[TMP2]], i8* [[DOTCAPTURE_EXPR_]], align 1
6180 // CHECK6-NEXT:    [[TMP3:%.*]] = load double, double* undef, align 8
6181 // CHECK6-NEXT:    [[ADD:%.*]] = fadd double 2.000000e+00, [[TMP3]]
6182 // CHECK6-NEXT:    store double [[ADD]], double* [[DOTCAPTURE_EXPR_1]], align 8
6183 // CHECK6-NEXT:    [[TMP4:%.*]] = load double, double* [[DOTCAPTURE_EXPR_1]], align 8
6184 // CHECK6-NEXT:    [[SUB:%.*]] = fsub double [[TMP4]], 1.000000e+00
6185 // CHECK6-NEXT:    [[DIV:%.*]] = fdiv double [[SUB]], 1.000000e+00
6186 // CHECK6-NEXT:    [[CONV:%.*]] = fptoui double [[DIV]] to i64
6187 // CHECK6-NEXT:    [[SUB3:%.*]] = sub i64 [[CONV]], 1
6188 // CHECK6-NEXT:    store i64 [[SUB3]], i64* [[DOTCAPTURE_EXPR_2]], align 8
6189 // CHECK6-NEXT:    store i64 1, i64* [[I]], align 8
6190 // CHECK6-NEXT:    [[TMP5:%.*]] = load double, double* [[DOTCAPTURE_EXPR_1]], align 8
6191 // CHECK6-NEXT:    [[CMP:%.*]] = fcmp olt double 1.000000e+00, [[TMP5]]
6192 // CHECK6-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
6193 // CHECK6:       omp.precond.then:
6194 // CHECK6-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
6195 // CHECK6-NEXT:    [[TMP6:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
6196 // CHECK6-NEXT:    store i64 [[TMP6]], i64* [[DOTOMP_UB]], align 8
6197 // CHECK6-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
6198 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6199 // CHECK6-NEXT:    [[TMP7:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
6200 // CHECK6-NEXT:    [[CONV5:%.*]] = sext i8 [[TMP7]] to i64
6201 // CHECK6-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6202 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
6203 // CHECK6-NEXT:    call void @__kmpc_for_static_init_8u(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 [[CONV5]])
6204 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
6205 // CHECK6:       omp.dispatch.cond:
6206 // CHECK6-NEXT:    [[TMP10:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
6207 // CHECK6-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
6208 // CHECK6-NEXT:    [[CMP6:%.*]] = icmp ugt i64 [[TMP10]], [[TMP11]]
6209 // CHECK6-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6210 // CHECK6:       cond.true:
6211 // CHECK6-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
6212 // CHECK6-NEXT:    br label [[COND_END:%.*]]
6213 // CHECK6:       cond.false:
6214 // CHECK6-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
6215 // CHECK6-NEXT:    br label [[COND_END]]
6216 // CHECK6:       cond.end:
6217 // CHECK6-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
6218 // CHECK6-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
6219 // CHECK6-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
6220 // CHECK6-NEXT:    store i64 [[TMP14]], i64* [[DOTOMP_IV]], align 8
6221 // CHECK6-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
6222 // CHECK6-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
6223 // CHECK6-NEXT:    [[ADD7:%.*]] = add i64 [[TMP16]], 1
6224 // CHECK6-NEXT:    [[CMP8:%.*]] = icmp ult i64 [[TMP15]], [[ADD7]]
6225 // CHECK6-NEXT:    br i1 [[CMP8]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
6226 // CHECK6:       omp.dispatch.body:
6227 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6228 // CHECK6:       omp.inner.for.cond:
6229 // CHECK6-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
6230 // CHECK6-NEXT:    [[TMP18:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
6231 // CHECK6-NEXT:    [[ADD9:%.*]] = add i64 [[TMP18]], 1
6232 // CHECK6-NEXT:    [[CMP10:%.*]] = icmp ult i64 [[TMP17]], [[ADD9]]
6233 // CHECK6-NEXT:    br i1 [[CMP10]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6234 // CHECK6:       omp.inner.for.body:
6235 // CHECK6-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
6236 // CHECK6-NEXT:    [[MUL:%.*]] = mul i64 [[TMP19]], 1
6237 // CHECK6-NEXT:    [[ADD11:%.*]] = add i64 1, [[MUL]]
6238 // CHECK6-NEXT:    store i64 [[ADD11]], i64* [[I4]], align 8
6239 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6240 // CHECK6:       omp.body.continue:
6241 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6242 // CHECK6:       omp.inner.for.inc:
6243 // CHECK6-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
6244 // CHECK6-NEXT:    [[ADD12:%.*]] = add i64 [[TMP20]], 1
6245 // CHECK6-NEXT:    store i64 [[ADD12]], i64* [[DOTOMP_IV]], align 8
6246 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]]
6247 // CHECK6:       omp.inner.for.end:
6248 // CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
6249 // CHECK6:       omp.dispatch.inc:
6250 // CHECK6-NEXT:    [[TMP21:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
6251 // CHECK6-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
6252 // CHECK6-NEXT:    [[ADD13:%.*]] = add i64 [[TMP21]], [[TMP22]]
6253 // CHECK6-NEXT:    store i64 [[ADD13]], i64* [[DOTOMP_LB]], align 8
6254 // CHECK6-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
6255 // CHECK6-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_STRIDE]], align 8
6256 // CHECK6-NEXT:    [[ADD14:%.*]] = add i64 [[TMP23]], [[TMP24]]
6257 // CHECK6-NEXT:    store i64 [[ADD14]], i64* [[DOTOMP_UB]], align 8
6258 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
6259 // CHECK6:       omp.dispatch.end:
6260 // CHECK6-NEXT:    [[TMP25:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6261 // CHECK6-NEXT:    [[TMP26:%.*]] = load i32, i32* [[TMP25]], align 4
6262 // CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP26]])
6263 // CHECK6-NEXT:    br label [[OMP_PRECOND_END]]
6264 // CHECK6:       omp.precond.end:
6265 // CHECK6-NEXT:    ret void
6266 //
6267 //
6268 // CHECK6-LABEL: define {{[^@]+}}@_Z23without_schedule_clausePfS_S_S_
6269 // CHECK6-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
6270 // CHECK6-NEXT:  entry:
6271 // CHECK6-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
6272 // CHECK6-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
6273 // CHECK6-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
6274 // CHECK6-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
6275 // CHECK6-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
6276 // CHECK6-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
6277 // CHECK6-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
6278 // CHECK6-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
6279 // CHECK6-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
6280 // CHECK6-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6281 // CHECK6-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
6282 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
6283 // CHECK6-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
6284 // CHECK6-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
6285 // CHECK6-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
6286 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
6287 // CHECK6-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
6288 // CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
6289 // CHECK6-NEXT:    ret void
6290 //
6291 //
6292 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..1
6293 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
6294 // CHECK6-NEXT:  entry:
6295 // CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6296 // CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6297 // CHECK6-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
6298 // CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6299 // CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6300 // CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6301 // CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6302 // CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6303 // CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6304 // CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
6305 // CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6306 // CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6307 // CHECK6-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
6308 // CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
6309 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
6310 // CHECK6-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
6311 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 1
6312 // CHECK6-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
6313 // CHECK6-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 2
6314 // CHECK6-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
6315 // CHECK6-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[TMP0]], i32 0, i32 3
6316 // CHECK6-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
6317 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6318 // CHECK6-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
6319 // CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6320 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6321 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6322 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
6323 // CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6324 // CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6325 // CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 4571423
6326 // CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6327 // CHECK6:       cond.true:
6328 // CHECK6-NEXT:    br label [[COND_END:%.*]]
6329 // CHECK6:       cond.false:
6330 // CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6331 // CHECK6-NEXT:    br label [[COND_END]]
6332 // CHECK6:       cond.end:
6333 // CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 4571423, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
6334 // CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6335 // CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6336 // CHECK6-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
6337 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6338 // CHECK6:       omp.inner.for.cond:
6339 // CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6340 // CHECK6-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6341 // CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
6342 // CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6343 // CHECK6:       omp.inner.for.body:
6344 // CHECK6-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6345 // CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 7
6346 // CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 33, [[MUL]]
6347 // CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
6348 // CHECK6-NEXT:    [[TMP17:%.*]] = load float*, float** [[TMP4]], align 8
6349 // CHECK6-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
6350 // CHECK6-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
6351 // CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM]]
6352 // CHECK6-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
6353 // CHECK6-NEXT:    [[TMP20:%.*]] = load float*, float** [[TMP6]], align 8
6354 // CHECK6-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4
6355 // CHECK6-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP21]] to i64
6356 // CHECK6-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM2]]
6357 // CHECK6-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX3]], align 4
6358 // CHECK6-NEXT:    [[MUL4:%.*]] = fmul float [[TMP19]], [[TMP22]]
6359 // CHECK6-NEXT:    [[TMP23:%.*]] = load float*, float** [[TMP8]], align 8
6360 // CHECK6-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4
6361 // CHECK6-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP24]] to i64
6362 // CHECK6-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM5]]
6363 // CHECK6-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX6]], align 4
6364 // CHECK6-NEXT:    [[MUL7:%.*]] = fmul float [[MUL4]], [[TMP25]]
6365 // CHECK6-NEXT:    [[TMP26:%.*]] = load float*, float** [[TMP2]], align 8
6366 // CHECK6-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I]], align 4
6367 // CHECK6-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP27]] to i64
6368 // CHECK6-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[TMP26]], i64 [[IDXPROM8]]
6369 // CHECK6-NEXT:    store float [[MUL7]], float* [[ARRAYIDX9]], align 4
6370 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6371 // CHECK6:       omp.body.continue:
6372 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6373 // CHECK6:       omp.inner.for.inc:
6374 // CHECK6-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6375 // CHECK6-NEXT:    [[ADD10:%.*]] = add nsw i32 [[TMP28]], 1
6376 // CHECK6-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_IV]], align 4
6377 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]]
6378 // CHECK6:       omp.inner.for.end:
6379 // CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6380 // CHECK6:       omp.loop.exit:
6381 // CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
6382 // CHECK6-NEXT:    ret void
6383 //
6384 //
6385 // CHECK6-LABEL: define {{[^@]+}}@_Z18static_not_chunkedPfS_S_S_
6386 // CHECK6-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
6387 // CHECK6-NEXT:  entry:
6388 // CHECK6-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
6389 // CHECK6-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
6390 // CHECK6-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
6391 // CHECK6-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
6392 // CHECK6-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_1:%.*]], align 8
6393 // CHECK6-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
6394 // CHECK6-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
6395 // CHECK6-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
6396 // CHECK6-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
6397 // CHECK6-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6398 // CHECK6-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
6399 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
6400 // CHECK6-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
6401 // CHECK6-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
6402 // CHECK6-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
6403 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
6404 // CHECK6-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
6405 // CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.1*)* @.omp_outlined..2 to void (i32*, i32*, ...)*), %struct.anon.1* [[OMP_OUTLINED_ARG_AGG_]])
6406 // CHECK6-NEXT:    ret void
6407 //
6408 //
6409 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..2
6410 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.1* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
6411 // CHECK6-NEXT:  entry:
6412 // CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6413 // CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6414 // CHECK6-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.1*, align 8
6415 // CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6416 // CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6417 // CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6418 // CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6419 // CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6420 // CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6421 // CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
6422 // CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6423 // CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6424 // CHECK6-NEXT:    store %struct.anon.1* [[__CONTEXT]], %struct.anon.1** [[__CONTEXT_ADDR]], align 8
6425 // CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.anon.1*, %struct.anon.1** [[__CONTEXT_ADDR]], align 8
6426 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_1:%.*]], %struct.anon.1* [[TMP0]], i32 0, i32 0
6427 // CHECK6-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
6428 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 1
6429 // CHECK6-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
6430 // CHECK6-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 2
6431 // CHECK6-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
6432 // CHECK6-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_1]], %struct.anon.1* [[TMP0]], i32 0, i32 3
6433 // CHECK6-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
6434 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6435 // CHECK6-NEXT:    store i32 4571423, i32* [[DOTOMP_UB]], align 4
6436 // CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6437 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6438 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6439 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
6440 // CHECK6-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
6441 // CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6442 // CHECK6-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP11]], 4571423
6443 // CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6444 // CHECK6:       cond.true:
6445 // CHECK6-NEXT:    br label [[COND_END:%.*]]
6446 // CHECK6:       cond.false:
6447 // CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6448 // CHECK6-NEXT:    br label [[COND_END]]
6449 // CHECK6:       cond.end:
6450 // CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 4571423, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
6451 // CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6452 // CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6453 // CHECK6-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
6454 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6455 // CHECK6:       omp.inner.for.cond:
6456 // CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6457 // CHECK6-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6458 // CHECK6-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP14]], [[TMP15]]
6459 // CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6460 // CHECK6:       omp.inner.for.body:
6461 // CHECK6-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6462 // CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP16]], 7
6463 // CHECK6-NEXT:    [[SUB:%.*]] = sub nsw i32 32000000, [[MUL]]
6464 // CHECK6-NEXT:    store i32 [[SUB]], i32* [[I]], align 4
6465 // CHECK6-NEXT:    [[TMP17:%.*]] = load float*, float** [[TMP4]], align 8
6466 // CHECK6-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
6467 // CHECK6-NEXT:    [[IDXPROM:%.*]] = sext i32 [[TMP18]] to i64
6468 // CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP17]], i64 [[IDXPROM]]
6469 // CHECK6-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
6470 // CHECK6-NEXT:    [[TMP20:%.*]] = load float*, float** [[TMP6]], align 8
6471 // CHECK6-NEXT:    [[TMP21:%.*]] = load i32, i32* [[I]], align 4
6472 // CHECK6-NEXT:    [[IDXPROM2:%.*]] = sext i32 [[TMP21]] to i64
6473 // CHECK6-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds float, float* [[TMP20]], i64 [[IDXPROM2]]
6474 // CHECK6-NEXT:    [[TMP22:%.*]] = load float, float* [[ARRAYIDX3]], align 4
6475 // CHECK6-NEXT:    [[MUL4:%.*]] = fmul float [[TMP19]], [[TMP22]]
6476 // CHECK6-NEXT:    [[TMP23:%.*]] = load float*, float** [[TMP8]], align 8
6477 // CHECK6-NEXT:    [[TMP24:%.*]] = load i32, i32* [[I]], align 4
6478 // CHECK6-NEXT:    [[IDXPROM5:%.*]] = sext i32 [[TMP24]] to i64
6479 // CHECK6-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP23]], i64 [[IDXPROM5]]
6480 // CHECK6-NEXT:    [[TMP25:%.*]] = load float, float* [[ARRAYIDX6]], align 4
6481 // CHECK6-NEXT:    [[MUL7:%.*]] = fmul float [[MUL4]], [[TMP25]]
6482 // CHECK6-NEXT:    [[TMP26:%.*]] = load float*, float** [[TMP2]], align 8
6483 // CHECK6-NEXT:    [[TMP27:%.*]] = load i32, i32* [[I]], align 4
6484 // CHECK6-NEXT:    [[IDXPROM8:%.*]] = sext i32 [[TMP27]] to i64
6485 // CHECK6-NEXT:    [[ARRAYIDX9:%.*]] = getelementptr inbounds float, float* [[TMP26]], i64 [[IDXPROM8]]
6486 // CHECK6-NEXT:    store float [[MUL7]], float* [[ARRAYIDX9]], align 4
6487 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6488 // CHECK6:       omp.body.continue:
6489 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6490 // CHECK6:       omp.inner.for.inc:
6491 // CHECK6-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6492 // CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP28]], 1
6493 // CHECK6-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
6494 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]]
6495 // CHECK6:       omp.inner.for.end:
6496 // CHECK6-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
6497 // CHECK6:       omp.loop.exit:
6498 // CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
6499 // CHECK6-NEXT:    ret void
6500 //
6501 //
6502 // CHECK6-LABEL: define {{[^@]+}}@_Z14static_chunkedPfS_S_S_
6503 // CHECK6-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
6504 // CHECK6-NEXT:  entry:
6505 // CHECK6-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
6506 // CHECK6-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
6507 // CHECK6-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
6508 // CHECK6-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
6509 // CHECK6-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_2:%.*]], align 8
6510 // CHECK6-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
6511 // CHECK6-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
6512 // CHECK6-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
6513 // CHECK6-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
6514 // CHECK6-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6515 // CHECK6-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
6516 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
6517 // CHECK6-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
6518 // CHECK6-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
6519 // CHECK6-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
6520 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
6521 // CHECK6-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
6522 // CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.2*)* @.omp_outlined..3 to void (i32*, i32*, ...)*), %struct.anon.2* [[OMP_OUTLINED_ARG_AGG_]])
6523 // CHECK6-NEXT:    ret void
6524 //
6525 //
6526 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..3
6527 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.2* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
6528 // CHECK6-NEXT:  entry:
6529 // CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6530 // CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6531 // CHECK6-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.2*, align 8
6532 // CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
6533 // CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
6534 // CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
6535 // CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
6536 // CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
6537 // CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6538 // CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
6539 // CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6540 // CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6541 // CHECK6-NEXT:    store %struct.anon.2* [[__CONTEXT]], %struct.anon.2** [[__CONTEXT_ADDR]], align 8
6542 // CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.anon.2*, %struct.anon.2** [[__CONTEXT_ADDR]], align 8
6543 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_2:%.*]], %struct.anon.2* [[TMP0]], i32 0, i32 0
6544 // CHECK6-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
6545 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 1
6546 // CHECK6-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
6547 // CHECK6-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 2
6548 // CHECK6-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
6549 // CHECK6-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_2]], %struct.anon.2* [[TMP0]], i32 0, i32 3
6550 // CHECK6-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
6551 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
6552 // CHECK6-NEXT:    store i32 16908288, i32* [[DOTOMP_UB]], align 4
6553 // CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
6554 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6555 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6556 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
6557 // CHECK6-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 5)
6558 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
6559 // CHECK6:       omp.dispatch.cond:
6560 // CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6561 // CHECK6-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP11]], 16908288
6562 // CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
6563 // CHECK6:       cond.true:
6564 // CHECK6-NEXT:    br label [[COND_END:%.*]]
6565 // CHECK6:       cond.false:
6566 // CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6567 // CHECK6-NEXT:    br label [[COND_END]]
6568 // CHECK6:       cond.end:
6569 // CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 16908288, [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
6570 // CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
6571 // CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6572 // CHECK6-NEXT:    store i32 [[TMP13]], i32* [[DOTOMP_IV]], align 4
6573 // CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6574 // CHECK6-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6575 // CHECK6-NEXT:    [[CMP1:%.*]] = icmp ule i32 [[TMP14]], [[TMP15]]
6576 // CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
6577 // CHECK6:       omp.dispatch.body:
6578 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6579 // CHECK6:       omp.inner.for.cond:
6580 // CHECK6-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6581 // CHECK6-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6582 // CHECK6-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[TMP16]], [[TMP17]]
6583 // CHECK6-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6584 // CHECK6:       omp.inner.for.body:
6585 // CHECK6-NEXT:    [[TMP18:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6586 // CHECK6-NEXT:    [[MUL:%.*]] = mul i32 [[TMP18]], 127
6587 // CHECK6-NEXT:    [[ADD:%.*]] = add i32 131071, [[MUL]]
6588 // CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
6589 // CHECK6-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP4]], align 8
6590 // CHECK6-NEXT:    [[TMP20:%.*]] = load i32, i32* [[I]], align 4
6591 // CHECK6-NEXT:    [[IDXPROM:%.*]] = zext i32 [[TMP20]] to i64
6592 // CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[IDXPROM]]
6593 // CHECK6-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX]], align 4
6594 // CHECK6-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP6]], align 8
6595 // CHECK6-NEXT:    [[TMP23:%.*]] = load i32, i32* [[I]], align 4
6596 // CHECK6-NEXT:    [[IDXPROM3:%.*]] = zext i32 [[TMP23]] to i64
6597 // CHECK6-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[IDXPROM3]]
6598 // CHECK6-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4
6599 // CHECK6-NEXT:    [[MUL5:%.*]] = fmul float [[TMP21]], [[TMP24]]
6600 // CHECK6-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP8]], align 8
6601 // CHECK6-NEXT:    [[TMP26:%.*]] = load i32, i32* [[I]], align 4
6602 // CHECK6-NEXT:    [[IDXPROM6:%.*]] = zext i32 [[TMP26]] to i64
6603 // CHECK6-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[IDXPROM6]]
6604 // CHECK6-NEXT:    [[TMP27:%.*]] = load float, float* [[ARRAYIDX7]], align 4
6605 // CHECK6-NEXT:    [[MUL8:%.*]] = fmul float [[MUL5]], [[TMP27]]
6606 // CHECK6-NEXT:    [[TMP28:%.*]] = load float*, float** [[TMP2]], align 8
6607 // CHECK6-NEXT:    [[TMP29:%.*]] = load i32, i32* [[I]], align 4
6608 // CHECK6-NEXT:    [[IDXPROM9:%.*]] = zext i32 [[TMP29]] to i64
6609 // CHECK6-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP28]], i64 [[IDXPROM9]]
6610 // CHECK6-NEXT:    store float [[MUL8]], float* [[ARRAYIDX10]], align 4
6611 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6612 // CHECK6:       omp.body.continue:
6613 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6614 // CHECK6:       omp.inner.for.inc:
6615 // CHECK6-NEXT:    [[TMP30:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
6616 // CHECK6-NEXT:    [[ADD11:%.*]] = add i32 [[TMP30]], 1
6617 // CHECK6-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_IV]], align 4
6618 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]]
6619 // CHECK6:       omp.inner.for.end:
6620 // CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
6621 // CHECK6:       omp.dispatch.inc:
6622 // CHECK6-NEXT:    [[TMP31:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
6623 // CHECK6-NEXT:    [[TMP32:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6624 // CHECK6-NEXT:    [[ADD12:%.*]] = add i32 [[TMP31]], [[TMP32]]
6625 // CHECK6-NEXT:    store i32 [[ADD12]], i32* [[DOTOMP_LB]], align 4
6626 // CHECK6-NEXT:    [[TMP33:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
6627 // CHECK6-NEXT:    [[TMP34:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
6628 // CHECK6-NEXT:    [[ADD13:%.*]] = add i32 [[TMP33]], [[TMP34]]
6629 // CHECK6-NEXT:    store i32 [[ADD13]], i32* [[DOTOMP_UB]], align 4
6630 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
6631 // CHECK6:       omp.dispatch.end:
6632 // CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP10]])
6633 // CHECK6-NEXT:    ret void
6634 //
6635 //
6636 // CHECK6-LABEL: define {{[^@]+}}@_Z8dynamic1PfS_S_S_
6637 // CHECK6-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
6638 // CHECK6-NEXT:  entry:
6639 // CHECK6-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
6640 // CHECK6-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
6641 // CHECK6-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
6642 // CHECK6-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
6643 // CHECK6-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_3:%.*]], align 8
6644 // CHECK6-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
6645 // CHECK6-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
6646 // CHECK6-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
6647 // CHECK6-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
6648 // CHECK6-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6649 // CHECK6-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
6650 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
6651 // CHECK6-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
6652 // CHECK6-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
6653 // CHECK6-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
6654 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
6655 // CHECK6-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
6656 // CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.3*)* @.omp_outlined..4 to void (i32*, i32*, ...)*), %struct.anon.3* [[OMP_OUTLINED_ARG_AGG_]])
6657 // CHECK6-NEXT:    ret void
6658 //
6659 //
6660 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..4
6661 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.3* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
6662 // CHECK6-NEXT:  entry:
6663 // CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6664 // CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6665 // CHECK6-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.3*, align 8
6666 // CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
6667 // CHECK6-NEXT:    [[TMP:%.*]] = alloca i64, align 8
6668 // CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
6669 // CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
6670 // CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
6671 // CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6672 // CHECK6-NEXT:    [[I:%.*]] = alloca i64, align 8
6673 // CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6674 // CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6675 // CHECK6-NEXT:    store %struct.anon.3* [[__CONTEXT]], %struct.anon.3** [[__CONTEXT_ADDR]], align 8
6676 // CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.anon.3*, %struct.anon.3** [[__CONTEXT_ADDR]], align 8
6677 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_3:%.*]], %struct.anon.3* [[TMP0]], i32 0, i32 0
6678 // CHECK6-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
6679 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 1
6680 // CHECK6-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
6681 // CHECK6-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 2
6682 // CHECK6-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
6683 // CHECK6-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_3]], %struct.anon.3* [[TMP0]], i32 0, i32 3
6684 // CHECK6-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
6685 // CHECK6-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
6686 // CHECK6-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
6687 // CHECK6-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
6688 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6689 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6690 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
6691 // CHECK6-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 1073741859, i64 0, i64 16908287, i64 1, i64 1)
6692 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
6693 // CHECK6:       omp.dispatch.cond:
6694 // CHECK6-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
6695 // CHECK6-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
6696 // CHECK6-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
6697 // CHECK6:       omp.dispatch.body:
6698 // CHECK6-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
6699 // CHECK6-NEXT:    store i64 [[TMP12]], i64* [[DOTOMP_IV]], align 8
6700 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6701 // CHECK6:       omp.inner.for.cond:
6702 // CHECK6-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
6703 // CHECK6-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !5
6704 // CHECK6-NEXT:    [[ADD:%.*]] = add i64 [[TMP14]], 1
6705 // CHECK6-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP13]], [[ADD]]
6706 // CHECK6-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6707 // CHECK6:       omp.inner.for.body:
6708 // CHECK6-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
6709 // CHECK6-NEXT:    [[MUL:%.*]] = mul i64 [[TMP15]], 127
6710 // CHECK6-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
6711 // CHECK6-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8, !llvm.access.group !5
6712 // CHECK6-NEXT:    [[TMP16:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !5
6713 // CHECK6-NEXT:    [[TMP17:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
6714 // CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[TMP17]]
6715 // CHECK6-NEXT:    [[TMP18:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !5
6716 // CHECK6-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !5
6717 // CHECK6-NEXT:    [[TMP20:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
6718 // CHECK6-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[TMP20]]
6719 // CHECK6-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX2]], align 4, !llvm.access.group !5
6720 // CHECK6-NEXT:    [[MUL3:%.*]] = fmul float [[TMP18]], [[TMP21]]
6721 // CHECK6-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !5
6722 // CHECK6-NEXT:    [[TMP23:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
6723 // CHECK6-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[TMP23]]
6724 // CHECK6-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !llvm.access.group !5
6725 // CHECK6-NEXT:    [[MUL5:%.*]] = fmul float [[MUL3]], [[TMP24]]
6726 // CHECK6-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !5
6727 // CHECK6-NEXT:    [[TMP26:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !5
6728 // CHECK6-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[TMP26]]
6729 // CHECK6-NEXT:    store float [[MUL5]], float* [[ARRAYIDX6]], align 4, !llvm.access.group !5
6730 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6731 // CHECK6:       omp.body.continue:
6732 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6733 // CHECK6:       omp.inner.for.inc:
6734 // CHECK6-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
6735 // CHECK6-NEXT:    [[ADD7:%.*]] = add i64 [[TMP27]], 1
6736 // CHECK6-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !5
6737 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP6:![0-9]+]]
6738 // CHECK6:       omp.inner.for.end:
6739 // CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
6740 // CHECK6:       omp.dispatch.inc:
6741 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
6742 // CHECK6:       omp.dispatch.end:
6743 // CHECK6-NEXT:    ret void
6744 //
6745 //
6746 // CHECK6-LABEL: define {{[^@]+}}@_Z7guided7PfS_S_S_
6747 // CHECK6-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
6748 // CHECK6-NEXT:  entry:
6749 // CHECK6-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
6750 // CHECK6-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
6751 // CHECK6-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
6752 // CHECK6-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
6753 // CHECK6-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_4:%.*]], align 8
6754 // CHECK6-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
6755 // CHECK6-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
6756 // CHECK6-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
6757 // CHECK6-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
6758 // CHECK6-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6759 // CHECK6-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
6760 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
6761 // CHECK6-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
6762 // CHECK6-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
6763 // CHECK6-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
6764 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
6765 // CHECK6-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
6766 // CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.4*)* @.omp_outlined..5 to void (i32*, i32*, ...)*), %struct.anon.4* [[OMP_OUTLINED_ARG_AGG_]])
6767 // CHECK6-NEXT:    ret void
6768 //
6769 //
6770 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..5
6771 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.4* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
6772 // CHECK6-NEXT:  entry:
6773 // CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6774 // CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6775 // CHECK6-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.4*, align 8
6776 // CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
6777 // CHECK6-NEXT:    [[TMP:%.*]] = alloca i64, align 8
6778 // CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
6779 // CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
6780 // CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
6781 // CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6782 // CHECK6-NEXT:    [[I:%.*]] = alloca i64, align 8
6783 // CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6784 // CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6785 // CHECK6-NEXT:    store %struct.anon.4* [[__CONTEXT]], %struct.anon.4** [[__CONTEXT_ADDR]], align 8
6786 // CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.anon.4*, %struct.anon.4** [[__CONTEXT_ADDR]], align 8
6787 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_4:%.*]], %struct.anon.4* [[TMP0]], i32 0, i32 0
6788 // CHECK6-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
6789 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 1
6790 // CHECK6-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
6791 // CHECK6-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 2
6792 // CHECK6-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
6793 // CHECK6-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_4]], %struct.anon.4* [[TMP0]], i32 0, i32 3
6794 // CHECK6-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
6795 // CHECK6-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
6796 // CHECK6-NEXT:    store i64 16908287, i64* [[DOTOMP_UB]], align 8
6797 // CHECK6-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
6798 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6799 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6800 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
6801 // CHECK6-NEXT:    call void @__kmpc_dispatch_init_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 1073741860, i64 0, i64 16908287, i64 1, i64 7)
6802 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
6803 // CHECK6:       omp.dispatch.cond:
6804 // CHECK6-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_8u(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
6805 // CHECK6-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
6806 // CHECK6-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
6807 // CHECK6:       omp.dispatch.body:
6808 // CHECK6-NEXT:    [[TMP12:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
6809 // CHECK6-NEXT:    store i64 [[TMP12]], i64* [[DOTOMP_IV]], align 8
6810 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6811 // CHECK6:       omp.inner.for.cond:
6812 // CHECK6-NEXT:    [[TMP13:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
6813 // CHECK6-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !8
6814 // CHECK6-NEXT:    [[ADD:%.*]] = add i64 [[TMP14]], 1
6815 // CHECK6-NEXT:    [[CMP:%.*]] = icmp ult i64 [[TMP13]], [[ADD]]
6816 // CHECK6-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6817 // CHECK6:       omp.inner.for.body:
6818 // CHECK6-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
6819 // CHECK6-NEXT:    [[MUL:%.*]] = mul i64 [[TMP15]], 127
6820 // CHECK6-NEXT:    [[ADD1:%.*]] = add i64 131071, [[MUL]]
6821 // CHECK6-NEXT:    store i64 [[ADD1]], i64* [[I]], align 8, !llvm.access.group !8
6822 // CHECK6-NEXT:    [[TMP16:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !8
6823 // CHECK6-NEXT:    [[TMP17:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
6824 // CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP16]], i64 [[TMP17]]
6825 // CHECK6-NEXT:    [[TMP18:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !8
6826 // CHECK6-NEXT:    [[TMP19:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !8
6827 // CHECK6-NEXT:    [[TMP20:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
6828 // CHECK6-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, float* [[TMP19]], i64 [[TMP20]]
6829 // CHECK6-NEXT:    [[TMP21:%.*]] = load float, float* [[ARRAYIDX2]], align 4, !llvm.access.group !8
6830 // CHECK6-NEXT:    [[MUL3:%.*]] = fmul float [[TMP18]], [[TMP21]]
6831 // CHECK6-NEXT:    [[TMP22:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !8
6832 // CHECK6-NEXT:    [[TMP23:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
6833 // CHECK6-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, float* [[TMP22]], i64 [[TMP23]]
6834 // CHECK6-NEXT:    [[TMP24:%.*]] = load float, float* [[ARRAYIDX4]], align 4, !llvm.access.group !8
6835 // CHECK6-NEXT:    [[MUL5:%.*]] = fmul float [[MUL3]], [[TMP24]]
6836 // CHECK6-NEXT:    [[TMP25:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !8
6837 // CHECK6-NEXT:    [[TMP26:%.*]] = load i64, i64* [[I]], align 8, !llvm.access.group !8
6838 // CHECK6-NEXT:    [[ARRAYIDX6:%.*]] = getelementptr inbounds float, float* [[TMP25]], i64 [[TMP26]]
6839 // CHECK6-NEXT:    store float [[MUL5]], float* [[ARRAYIDX6]], align 4, !llvm.access.group !8
6840 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
6841 // CHECK6:       omp.body.continue:
6842 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
6843 // CHECK6:       omp.inner.for.inc:
6844 // CHECK6-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
6845 // CHECK6-NEXT:    [[ADD7:%.*]] = add i64 [[TMP27]], 1
6846 // CHECK6-NEXT:    store i64 [[ADD7]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !8
6847 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
6848 // CHECK6:       omp.inner.for.end:
6849 // CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
6850 // CHECK6:       omp.dispatch.inc:
6851 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
6852 // CHECK6:       omp.dispatch.end:
6853 // CHECK6-NEXT:    ret void
6854 //
6855 //
6856 // CHECK6-LABEL: define {{[^@]+}}@_Z9test_autoPfS_S_S_
6857 // CHECK6-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
6858 // CHECK6-NEXT:  entry:
6859 // CHECK6-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
6860 // CHECK6-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
6861 // CHECK6-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
6862 // CHECK6-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
6863 // CHECK6-NEXT:    [[X:%.*]] = alloca i32, align 4
6864 // CHECK6-NEXT:    [[Y:%.*]] = alloca i32, align 4
6865 // CHECK6-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_5:%.*]], align 8
6866 // CHECK6-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
6867 // CHECK6-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
6868 // CHECK6-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
6869 // CHECK6-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
6870 // CHECK6-NEXT:    store i32 0, i32* [[X]], align 4
6871 // CHECK6-NEXT:    store i32 0, i32* [[Y]], align 4
6872 // CHECK6-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
6873 // CHECK6-NEXT:    store i32* [[Y]], i32** [[TMP0]], align 8
6874 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
6875 // CHECK6-NEXT:    store float** [[A_ADDR]], float*** [[TMP1]], align 8
6876 // CHECK6-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
6877 // CHECK6-NEXT:    store float** [[B_ADDR]], float*** [[TMP2]], align 8
6878 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
6879 // CHECK6-NEXT:    store float** [[C_ADDR]], float*** [[TMP3]], align 8
6880 // CHECK6-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 4
6881 // CHECK6-NEXT:    store float** [[D_ADDR]], float*** [[TMP4]], align 8
6882 // CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.5*)* @.omp_outlined..6 to void (i32*, i32*, ...)*), %struct.anon.5* [[OMP_OUTLINED_ARG_AGG_]])
6883 // CHECK6-NEXT:    ret void
6884 //
6885 //
6886 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..6
6887 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.5* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
6888 // CHECK6-NEXT:  entry:
6889 // CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
6890 // CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
6891 // CHECK6-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.5*, align 8
6892 // CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
6893 // CHECK6-NEXT:    [[TMP:%.*]] = alloca i8, align 1
6894 // CHECK6-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
6895 // CHECK6-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i8, align 1
6896 // CHECK6-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i64, align 8
6897 // CHECK6-NEXT:    [[I:%.*]] = alloca i8, align 1
6898 // CHECK6-NEXT:    [[X:%.*]] = alloca i32, align 4
6899 // CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
6900 // CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
6901 // CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
6902 // CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
6903 // CHECK6-NEXT:    [[I7:%.*]] = alloca i8, align 1
6904 // CHECK6-NEXT:    [[X8:%.*]] = alloca i32, align 4
6905 // CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
6906 // CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
6907 // CHECK6-NEXT:    store %struct.anon.5* [[__CONTEXT]], %struct.anon.5** [[__CONTEXT_ADDR]], align 8
6908 // CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.anon.5*, %struct.anon.5** [[__CONTEXT_ADDR]], align 8
6909 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_5:%.*]], %struct.anon.5* [[TMP0]], i32 0, i32 0
6910 // CHECK6-NEXT:    [[TMP2:%.*]] = load i32*, i32** [[TMP1]], align 8
6911 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 1
6912 // CHECK6-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
6913 // CHECK6-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 2
6914 // CHECK6-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
6915 // CHECK6-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 3
6916 // CHECK6-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
6917 // CHECK6-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_ANON_5]], %struct.anon.5* [[TMP0]], i32 0, i32 4
6918 // CHECK6-NEXT:    [[TMP10:%.*]] = load float**, float*** [[TMP9]], align 8
6919 // CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[TMP2]], align 4
6920 // CHECK6-NEXT:    [[CONV:%.*]] = trunc i32 [[TMP11]] to i8
6921 // CHECK6-NEXT:    store i8 [[CONV]], i8* [[DOTCAPTURE_EXPR_]], align 1
6922 // CHECK6-NEXT:    [[TMP12:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
6923 // CHECK6-NEXT:    [[CONV3:%.*]] = sext i8 [[TMP12]] to i32
6924 // CHECK6-NEXT:    [[SUB:%.*]] = sub i32 57, [[CONV3]]
6925 // CHECK6-NEXT:    [[ADD:%.*]] = add i32 [[SUB]], 1
6926 // CHECK6-NEXT:    [[DIV:%.*]] = udiv i32 [[ADD]], 1
6927 // CHECK6-NEXT:    [[CONV4:%.*]] = zext i32 [[DIV]] to i64
6928 // CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i64 [[CONV4]], 11
6929 // CHECK6-NEXT:    [[SUB5:%.*]] = sub nsw i64 [[MUL]], 1
6930 // CHECK6-NEXT:    store i64 [[SUB5]], i64* [[DOTCAPTURE_EXPR_2]], align 8
6931 // CHECK6-NEXT:    [[TMP13:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
6932 // CHECK6-NEXT:    store i8 [[TMP13]], i8* [[I]], align 1
6933 // CHECK6-NEXT:    store i32 11, i32* [[X]], align 4
6934 // CHECK6-NEXT:    [[TMP14:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1
6935 // CHECK6-NEXT:    [[CONV6:%.*]] = sext i8 [[TMP14]] to i32
6936 // CHECK6-NEXT:    [[CMP:%.*]] = icmp sle i32 [[CONV6]], 57
6937 // CHECK6-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
6938 // CHECK6:       omp.precond.then:
6939 // CHECK6-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
6940 // CHECK6-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
6941 // CHECK6-NEXT:    store i64 [[TMP15]], i64* [[DOTOMP_UB]], align 8
6942 // CHECK6-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
6943 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
6944 // CHECK6-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_2]], align 8
6945 // CHECK6-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6946 // CHECK6-NEXT:    [[TMP18:%.*]] = load i32, i32* [[TMP17]], align 4
6947 // CHECK6-NEXT:    call void @__kmpc_dispatch_init_8(%struct.ident_t* @[[GLOB2]], i32 [[TMP18]], i32 1073741862, i64 0, i64 [[TMP16]], i64 1, i64 1)
6948 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
6949 // CHECK6:       omp.dispatch.cond:
6950 // CHECK6-NEXT:    [[TMP19:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
6951 // CHECK6-NEXT:    [[TMP20:%.*]] = load i32, i32* [[TMP19]], align 4
6952 // CHECK6-NEXT:    [[TMP21:%.*]] = call i32 @__kmpc_dispatch_next_8(%struct.ident_t* @[[GLOB2]], i32 [[TMP20]], i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]])
6953 // CHECK6-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP21]], 0
6954 // CHECK6-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
6955 // CHECK6:       omp.dispatch.body:
6956 // CHECK6-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
6957 // CHECK6-NEXT:    store i64 [[TMP22]], i64* [[DOTOMP_IV]], align 8
6958 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
6959 // CHECK6:       omp.inner.for.cond:
6960 // CHECK6-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
6961 // CHECK6-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8, !llvm.access.group !11
6962 // CHECK6-NEXT:    [[CMP9:%.*]] = icmp sle i64 [[TMP23]], [[TMP24]]
6963 // CHECK6-NEXT:    br i1 [[CMP9]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
6964 // CHECK6:       omp.inner.for.body:
6965 // CHECK6-NEXT:    [[TMP25:%.*]] = load i8, i8* [[DOTCAPTURE_EXPR_]], align 1, !llvm.access.group !11
6966 // CHECK6-NEXT:    [[CONV10:%.*]] = sext i8 [[TMP25]] to i64
6967 // CHECK6-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
6968 // CHECK6-NEXT:    [[DIV11:%.*]] = sdiv i64 [[TMP26]], 11
6969 // CHECK6-NEXT:    [[MUL12:%.*]] = mul nsw i64 [[DIV11]], 1
6970 // CHECK6-NEXT:    [[ADD13:%.*]] = add nsw i64 [[CONV10]], [[MUL12]]
6971 // CHECK6-NEXT:    [[CONV14:%.*]] = trunc i64 [[ADD13]] to i8
6972 // CHECK6-NEXT:    store i8 [[CONV14]], i8* [[I7]], align 1, !llvm.access.group !11
6973 // CHECK6-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
6974 // CHECK6-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
6975 // CHECK6-NEXT:    [[DIV15:%.*]] = sdiv i64 [[TMP28]], 11
6976 // CHECK6-NEXT:    [[MUL16:%.*]] = mul nsw i64 [[DIV15]], 11
6977 // CHECK6-NEXT:    [[SUB17:%.*]] = sub nsw i64 [[TMP27]], [[MUL16]]
6978 // CHECK6-NEXT:    [[MUL18:%.*]] = mul nsw i64 [[SUB17]], 1
6979 // CHECK6-NEXT:    [[SUB19:%.*]] = sub nsw i64 11, [[MUL18]]
6980 // CHECK6-NEXT:    [[CONV20:%.*]] = trunc i64 [[SUB19]] to i32
6981 // CHECK6-NEXT:    store i32 [[CONV20]], i32* [[X8]], align 4, !llvm.access.group !11
6982 // CHECK6-NEXT:    [[TMP29:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !11
6983 // CHECK6-NEXT:    [[TMP30:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
6984 // CHECK6-NEXT:    [[IDXPROM:%.*]] = sext i8 [[TMP30]] to i64
6985 // CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP29]], i64 [[IDXPROM]]
6986 // CHECK6-NEXT:    [[TMP31:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !11
6987 // CHECK6-NEXT:    [[TMP32:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !11
6988 // CHECK6-NEXT:    [[TMP33:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
6989 // CHECK6-NEXT:    [[IDXPROM21:%.*]] = sext i8 [[TMP33]] to i64
6990 // CHECK6-NEXT:    [[ARRAYIDX22:%.*]] = getelementptr inbounds float, float* [[TMP32]], i64 [[IDXPROM21]]
6991 // CHECK6-NEXT:    [[TMP34:%.*]] = load float, float* [[ARRAYIDX22]], align 4, !llvm.access.group !11
6992 // CHECK6-NEXT:    [[MUL23:%.*]] = fmul float [[TMP31]], [[TMP34]]
6993 // CHECK6-NEXT:    [[TMP35:%.*]] = load float*, float** [[TMP10]], align 8, !llvm.access.group !11
6994 // CHECK6-NEXT:    [[TMP36:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
6995 // CHECK6-NEXT:    [[IDXPROM24:%.*]] = sext i8 [[TMP36]] to i64
6996 // CHECK6-NEXT:    [[ARRAYIDX25:%.*]] = getelementptr inbounds float, float* [[TMP35]], i64 [[IDXPROM24]]
6997 // CHECK6-NEXT:    [[TMP37:%.*]] = load float, float* [[ARRAYIDX25]], align 4, !llvm.access.group !11
6998 // CHECK6-NEXT:    [[MUL26:%.*]] = fmul float [[MUL23]], [[TMP37]]
6999 // CHECK6-NEXT:    [[TMP38:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !11
7000 // CHECK6-NEXT:    [[TMP39:%.*]] = load i8, i8* [[I7]], align 1, !llvm.access.group !11
7001 // CHECK6-NEXT:    [[IDXPROM27:%.*]] = sext i8 [[TMP39]] to i64
7002 // CHECK6-NEXT:    [[ARRAYIDX28:%.*]] = getelementptr inbounds float, float* [[TMP38]], i64 [[IDXPROM27]]
7003 // CHECK6-NEXT:    store float [[MUL26]], float* [[ARRAYIDX28]], align 4, !llvm.access.group !11
7004 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7005 // CHECK6:       omp.body.continue:
7006 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7007 // CHECK6:       omp.inner.for.inc:
7008 // CHECK6-NEXT:    [[TMP40:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
7009 // CHECK6-NEXT:    [[ADD29:%.*]] = add nsw i64 [[TMP40]], 1
7010 // CHECK6-NEXT:    store i64 [[ADD29]], i64* [[DOTOMP_IV]], align 8, !llvm.access.group !11
7011 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP12:![0-9]+]]
7012 // CHECK6:       omp.inner.for.end:
7013 // CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
7014 // CHECK6:       omp.dispatch.inc:
7015 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
7016 // CHECK6:       omp.dispatch.end:
7017 // CHECK6-NEXT:    br label [[OMP_PRECOND_END]]
7018 // CHECK6:       omp.precond.end:
7019 // CHECK6-NEXT:    ret void
7020 //
7021 //
7022 // CHECK6-LABEL: define {{[^@]+}}@_Z7runtimePfS_S_S_
7023 // CHECK6-SAME: (float* [[A:%.*]], float* [[B:%.*]], float* [[C:%.*]], float* [[D:%.*]]) #[[ATTR0]] {
7024 // CHECK6-NEXT:  entry:
7025 // CHECK6-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
7026 // CHECK6-NEXT:    [[B_ADDR:%.*]] = alloca float*, align 8
7027 // CHECK6-NEXT:    [[C_ADDR:%.*]] = alloca float*, align 8
7028 // CHECK6-NEXT:    [[D_ADDR:%.*]] = alloca float*, align 8
7029 // CHECK6-NEXT:    [[X:%.*]] = alloca i32, align 4
7030 // CHECK6-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_6:%.*]], align 8
7031 // CHECK6-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
7032 // CHECK6-NEXT:    store float* [[B]], float** [[B_ADDR]], align 8
7033 // CHECK6-NEXT:    store float* [[C]], float** [[C_ADDR]], align 8
7034 // CHECK6-NEXT:    store float* [[D]], float** [[D_ADDR]], align 8
7035 // CHECK6-NEXT:    store i32 0, i32* [[X]], align 4
7036 // CHECK6-NEXT:    [[TMP0:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7037 // CHECK6-NEXT:    store float** [[A_ADDR]], float*** [[TMP0]], align 8
7038 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
7039 // CHECK6-NEXT:    store float** [[B_ADDR]], float*** [[TMP1]], align 8
7040 // CHECK6-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
7041 // CHECK6-NEXT:    store float** [[C_ADDR]], float*** [[TMP2]], align 8
7042 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 3
7043 // CHECK6-NEXT:    store float** [[D_ADDR]], float*** [[TMP3]], align 8
7044 // CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.6*)* @.omp_outlined..7 to void (i32*, i32*, ...)*), %struct.anon.6* [[OMP_OUTLINED_ARG_AGG_]])
7045 // CHECK6-NEXT:    ret void
7046 //
7047 //
7048 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..7
7049 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.6* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
7050 // CHECK6-NEXT:  entry:
7051 // CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7052 // CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7053 // CHECK6-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.6*, align 8
7054 // CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7055 // CHECK6-NEXT:    [[TMP:%.*]] = alloca i8, align 1
7056 // CHECK6-NEXT:    [[_TMP1:%.*]] = alloca i32, align 4
7057 // CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7058 // CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7059 // CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7060 // CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7061 // CHECK6-NEXT:    [[I:%.*]] = alloca i8, align 1
7062 // CHECK6-NEXT:    [[X:%.*]] = alloca i32, align 4
7063 // CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7064 // CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7065 // CHECK6-NEXT:    store %struct.anon.6* [[__CONTEXT]], %struct.anon.6** [[__CONTEXT_ADDR]], align 8
7066 // CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.anon.6*, %struct.anon.6** [[__CONTEXT_ADDR]], align 8
7067 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_6:%.*]], %struct.anon.6* [[TMP0]], i32 0, i32 0
7068 // CHECK6-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
7069 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 1
7070 // CHECK6-NEXT:    [[TMP4:%.*]] = load float**, float*** [[TMP3]], align 8
7071 // CHECK6-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 2
7072 // CHECK6-NEXT:    [[TMP6:%.*]] = load float**, float*** [[TMP5]], align 8
7073 // CHECK6-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [[STRUCT_ANON_6]], %struct.anon.6* [[TMP0]], i32 0, i32 3
7074 // CHECK6-NEXT:    [[TMP8:%.*]] = load float**, float*** [[TMP7]], align 8
7075 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7076 // CHECK6-NEXT:    store i32 199, i32* [[DOTOMP_UB]], align 4
7077 // CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7078 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7079 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7080 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[TMP9]], align 4
7081 // CHECK6-NEXT:    call void @__kmpc_dispatch_init_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32 1073741861, i32 0, i32 199, i32 1, i32 1)
7082 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
7083 // CHECK6:       omp.dispatch.cond:
7084 // CHECK6-NEXT:    [[TMP11:%.*]] = call i32 @__kmpc_dispatch_next_4(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]], i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]])
7085 // CHECK6-NEXT:    [[TOBOOL:%.*]] = icmp ne i32 [[TMP11]], 0
7086 // CHECK6-NEXT:    br i1 [[TOBOOL]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_END:%.*]]
7087 // CHECK6:       omp.dispatch.body:
7088 // CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7089 // CHECK6-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
7090 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7091 // CHECK6:       omp.inner.for.cond:
7092 // CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
7093 // CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4, !llvm.access.group !14
7094 // CHECK6-NEXT:    [[CMP:%.*]] = icmp sle i32 [[TMP13]], [[TMP14]]
7095 // CHECK6-NEXT:    br i1 [[CMP]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7096 // CHECK6:       omp.inner.for.body:
7097 // CHECK6-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
7098 // CHECK6-NEXT:    [[DIV:%.*]] = sdiv i32 [[TMP15]], 20
7099 // CHECK6-NEXT:    [[MUL:%.*]] = mul nsw i32 [[DIV]], 1
7100 // CHECK6-NEXT:    [[ADD:%.*]] = add nsw i32 48, [[MUL]]
7101 // CHECK6-NEXT:    [[CONV:%.*]] = trunc i32 [[ADD]] to i8
7102 // CHECK6-NEXT:    store i8 [[CONV]], i8* [[I]], align 1, !llvm.access.group !14
7103 // CHECK6-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
7104 // CHECK6-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
7105 // CHECK6-NEXT:    [[DIV2:%.*]] = sdiv i32 [[TMP17]], 20
7106 // CHECK6-NEXT:    [[MUL3:%.*]] = mul nsw i32 [[DIV2]], 20
7107 // CHECK6-NEXT:    [[SUB:%.*]] = sub nsw i32 [[TMP16]], [[MUL3]]
7108 // CHECK6-NEXT:    [[MUL4:%.*]] = mul nsw i32 [[SUB]], 1
7109 // CHECK6-NEXT:    [[ADD5:%.*]] = add nsw i32 -10, [[MUL4]]
7110 // CHECK6-NEXT:    store i32 [[ADD5]], i32* [[X]], align 4, !llvm.access.group !14
7111 // CHECK6-NEXT:    [[TMP18:%.*]] = load float*, float** [[TMP4]], align 8, !llvm.access.group !14
7112 // CHECK6-NEXT:    [[TMP19:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
7113 // CHECK6-NEXT:    [[IDXPROM:%.*]] = zext i8 [[TMP19]] to i64
7114 // CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[TMP18]], i64 [[IDXPROM]]
7115 // CHECK6-NEXT:    [[TMP20:%.*]] = load float, float* [[ARRAYIDX]], align 4, !llvm.access.group !14
7116 // CHECK6-NEXT:    [[TMP21:%.*]] = load float*, float** [[TMP6]], align 8, !llvm.access.group !14
7117 // CHECK6-NEXT:    [[TMP22:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
7118 // CHECK6-NEXT:    [[IDXPROM6:%.*]] = zext i8 [[TMP22]] to i64
7119 // CHECK6-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP21]], i64 [[IDXPROM6]]
7120 // CHECK6-NEXT:    [[TMP23:%.*]] = load float, float* [[ARRAYIDX7]], align 4, !llvm.access.group !14
7121 // CHECK6-NEXT:    [[MUL8:%.*]] = fmul float [[TMP20]], [[TMP23]]
7122 // CHECK6-NEXT:    [[TMP24:%.*]] = load float*, float** [[TMP8]], align 8, !llvm.access.group !14
7123 // CHECK6-NEXT:    [[TMP25:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
7124 // CHECK6-NEXT:    [[IDXPROM9:%.*]] = zext i8 [[TMP25]] to i64
7125 // CHECK6-NEXT:    [[ARRAYIDX10:%.*]] = getelementptr inbounds float, float* [[TMP24]], i64 [[IDXPROM9]]
7126 // CHECK6-NEXT:    [[TMP26:%.*]] = load float, float* [[ARRAYIDX10]], align 4, !llvm.access.group !14
7127 // CHECK6-NEXT:    [[MUL11:%.*]] = fmul float [[MUL8]], [[TMP26]]
7128 // CHECK6-NEXT:    [[TMP27:%.*]] = load float*, float** [[TMP2]], align 8, !llvm.access.group !14
7129 // CHECK6-NEXT:    [[TMP28:%.*]] = load i8, i8* [[I]], align 1, !llvm.access.group !14
7130 // CHECK6-NEXT:    [[IDXPROM12:%.*]] = zext i8 [[TMP28]] to i64
7131 // CHECK6-NEXT:    [[ARRAYIDX13:%.*]] = getelementptr inbounds float, float* [[TMP27]], i64 [[IDXPROM12]]
7132 // CHECK6-NEXT:    store float [[MUL11]], float* [[ARRAYIDX13]], align 4, !llvm.access.group !14
7133 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7134 // CHECK6:       omp.body.continue:
7135 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7136 // CHECK6:       omp.inner.for.inc:
7137 // CHECK6-NEXT:    [[TMP29:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
7138 // CHECK6-NEXT:    [[ADD14:%.*]] = add nsw i32 [[TMP29]], 1
7139 // CHECK6-NEXT:    store i32 [[ADD14]], i32* [[DOTOMP_IV]], align 4, !llvm.access.group !14
7140 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP15:![0-9]+]]
7141 // CHECK6:       omp.inner.for.end:
7142 // CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
7143 // CHECK6:       omp.dispatch.inc:
7144 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
7145 // CHECK6:       omp.dispatch.end:
7146 // CHECK6-NEXT:    ret void
7147 //
7148 //
7149 // CHECK6-LABEL: define {{[^@]+}}@_Z3foov
7150 // CHECK6-SAME: () #[[ATTR0]] {
7151 // CHECK6-NEXT:  entry:
7152 // CHECK6-NEXT:    call void @_Z8mayThrowv()
7153 // CHECK6-NEXT:    ret i32 0
7154 //
7155 //
7156 // CHECK6-LABEL: define {{[^@]+}}@_Z12parallel_forPfi
7157 // CHECK6-SAME: (float* [[A:%.*]], i32 [[N:%.*]]) #[[ATTR0]] {
7158 // CHECK6-NEXT:  entry:
7159 // CHECK6-NEXT:    [[A_ADDR:%.*]] = alloca float*, align 8
7160 // CHECK6-NEXT:    [[N_ADDR:%.*]] = alloca i32, align 4
7161 // CHECK6-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
7162 // CHECK6-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
7163 // CHECK6-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_7:%.*]], align 8
7164 // CHECK6-NEXT:    store float* [[A]], float** [[A_ADDR]], align 8
7165 // CHECK6-NEXT:    store i32 [[N]], i32* [[N_ADDR]], align 4
7166 // CHECK6-NEXT:    [[TMP0:%.*]] = load i32, i32* [[N_ADDR]], align 4
7167 // CHECK6-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
7168 // CHECK6-NEXT:    [[TMP2:%.*]] = call i8* @llvm.stacksave()
7169 // CHECK6-NEXT:    store i8* [[TMP2]], i8** [[SAVED_STACK]], align 8
7170 // CHECK6-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP1]], align 16
7171 // CHECK6-NEXT:    store i64 [[TMP1]], i64* [[__VLA_EXPR0]], align 8
7172 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7173 // CHECK6-NEXT:    store float** [[A_ADDR]], float*** [[TMP3]], align 8
7174 // CHECK6-NEXT:    [[TMP4:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 1
7175 // CHECK6-NEXT:    store i64 [[TMP1]], i64* [[TMP4]], align 8
7176 // CHECK6-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 2
7177 // CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[N_ADDR]], align 4
7178 // CHECK6-NEXT:    store i32 [[TMP6]], i32* [[TMP5]], align 8
7179 // CHECK6-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.7*)* @.omp_outlined..8 to void (i32*, i32*, ...)*), %struct.anon.7* [[OMP_OUTLINED_ARG_AGG_]])
7180 // CHECK6-NEXT:    [[TMP7:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
7181 // CHECK6-NEXT:    call void @llvm.stackrestore(i8* [[TMP7]])
7182 // CHECK6-NEXT:    ret void
7183 //
7184 //
7185 // CHECK6-LABEL: define {{[^@]+}}@.omp_outlined..8
7186 // CHECK6-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.7* noalias [[__CONTEXT:%.*]]) #[[ATTR1]] {
7187 // CHECK6-NEXT:  entry:
7188 // CHECK6-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7189 // CHECK6-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7190 // CHECK6-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.7*, align 8
7191 // CHECK6-NEXT:    [[N:%.*]] = alloca i32, align 4
7192 // CHECK6-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7193 // CHECK6-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7194 // CHECK6-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7195 // CHECK6-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7196 // CHECK6-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7197 // CHECK6-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7198 // CHECK6-NEXT:    [[SAVED_STACK:%.*]] = alloca i8*, align 8
7199 // CHECK6-NEXT:    [[__VLA_EXPR0:%.*]] = alloca i64, align 8
7200 // CHECK6-NEXT:    [[I:%.*]] = alloca i32, align 4
7201 // CHECK6-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7202 // CHECK6-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7203 // CHECK6-NEXT:    store %struct.anon.7* [[__CONTEXT]], %struct.anon.7** [[__CONTEXT_ADDR]], align 8
7204 // CHECK6-NEXT:    [[TMP0:%.*]] = load %struct.anon.7*, %struct.anon.7** [[__CONTEXT_ADDR]], align 8
7205 // CHECK6-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_7:%.*]], %struct.anon.7* [[TMP0]], i32 0, i32 0
7206 // CHECK6-NEXT:    [[TMP2:%.*]] = load float**, float*** [[TMP1]], align 8
7207 // CHECK6-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 1
7208 // CHECK6-NEXT:    [[TMP4:%.*]] = load i64, i64* [[TMP3]], align 8
7209 // CHECK6-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_ANON_7]], %struct.anon.7* [[TMP0]], i32 0, i32 2
7210 // CHECK6-NEXT:    [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 8
7211 // CHECK6-NEXT:    store i32 [[TMP6]], i32* [[N]], align 4
7212 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7213 // CHECK6-NEXT:    store i32 16908288, i32* [[DOTOMP_UB]], align 4
7214 // CHECK6-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7215 // CHECK6-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7216 // CHECK6-NEXT:    [[TMP7:%.*]] = call i8* @llvm.stacksave()
7217 // CHECK6-NEXT:    store i8* [[TMP7]], i8** [[SAVED_STACK]], align 8
7218 // CHECK6-NEXT:    [[VLA:%.*]] = alloca float, i64 [[TMP4]], align 16
7219 // CHECK6-NEXT:    store i64 [[TMP4]], i64* [[__VLA_EXPR0]], align 8
7220 // CHECK6-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7221 // CHECK6-NEXT:    [[TMP9:%.*]] = load i32, i32* [[TMP8]], align 4
7222 // CHECK6-NEXT:    call void @__kmpc_for_static_init_4u(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]], i32 33, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 5)
7223 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND:%.*]]
7224 // CHECK6:       omp.dispatch.cond:
7225 // CHECK6-NEXT:    [[TMP10:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7226 // CHECK6-NEXT:    [[CMP:%.*]] = icmp ugt i32 [[TMP10]], 16908288
7227 // CHECK6-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7228 // CHECK6:       cond.true:
7229 // CHECK6-NEXT:    br label [[COND_END:%.*]]
7230 // CHECK6:       cond.false:
7231 // CHECK6-NEXT:    [[TMP11:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7232 // CHECK6-NEXT:    br label [[COND_END]]
7233 // CHECK6:       cond.end:
7234 // CHECK6-NEXT:    [[COND:%.*]] = phi i32 [ 16908288, [[COND_TRUE]] ], [ [[TMP11]], [[COND_FALSE]] ]
7235 // CHECK6-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7236 // CHECK6-NEXT:    [[TMP12:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7237 // CHECK6-NEXT:    store i32 [[TMP12]], i32* [[DOTOMP_IV]], align 4
7238 // CHECK6-NEXT:    [[TMP13:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7239 // CHECK6-NEXT:    [[TMP14:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7240 // CHECK6-NEXT:    [[CMP1:%.*]] = icmp ule i32 [[TMP13]], [[TMP14]]
7241 // CHECK6-NEXT:    br i1 [[CMP1]], label [[OMP_DISPATCH_BODY:%.*]], label [[OMP_DISPATCH_CLEANUP:%.*]]
7242 // CHECK6:       omp.dispatch.cleanup:
7243 // CHECK6-NEXT:    br label [[OMP_DISPATCH_END:%.*]]
7244 // CHECK6:       omp.dispatch.body:
7245 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7246 // CHECK6:       omp.inner.for.cond:
7247 // CHECK6-NEXT:    [[TMP15:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7248 // CHECK6-NEXT:    [[TMP16:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7249 // CHECK6-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[TMP15]], [[TMP16]]
7250 // CHECK6-NEXT:    br i1 [[CMP2]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_COND_CLEANUP:%.*]]
7251 // CHECK6:       omp.inner.for.cond.cleanup:
7252 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_END:%.*]]
7253 // CHECK6:       omp.inner.for.body:
7254 // CHECK6-NEXT:    [[TMP17:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7255 // CHECK6-NEXT:    [[MUL:%.*]] = mul i32 [[TMP17]], 127
7256 // CHECK6-NEXT:    [[ADD:%.*]] = add i32 131071, [[MUL]]
7257 // CHECK6-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
7258 // CHECK6-NEXT:    [[CALL:%.*]] = call i32 @_Z3foov()
7259 // CHECK6-NEXT:    [[CONV:%.*]] = sitofp i32 [[CALL]] to float
7260 // CHECK6-NEXT:    [[TMP18:%.*]] = load i32, i32* [[I]], align 4
7261 // CHECK6-NEXT:    [[IDXPROM:%.*]] = zext i32 [[TMP18]] to i64
7262 // CHECK6-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, float* [[VLA]], i64 [[IDXPROM]]
7263 // CHECK6-NEXT:    [[TMP19:%.*]] = load float, float* [[ARRAYIDX]], align 4
7264 // CHECK6-NEXT:    [[ADD3:%.*]] = fadd float [[CONV]], [[TMP19]]
7265 // CHECK6-NEXT:    [[TMP20:%.*]] = load i32, i32* [[N]], align 4
7266 // CHECK6-NEXT:    [[CONV4:%.*]] = sitofp i32 [[TMP20]] to float
7267 // CHECK6-NEXT:    [[ADD5:%.*]] = fadd float [[ADD3]], [[CONV4]]
7268 // CHECK6-NEXT:    [[TMP21:%.*]] = load float*, float** [[TMP2]], align 8
7269 // CHECK6-NEXT:    [[TMP22:%.*]] = load i32, i32* [[I]], align 4
7270 // CHECK6-NEXT:    [[IDXPROM6:%.*]] = zext i32 [[TMP22]] to i64
7271 // CHECK6-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, float* [[TMP21]], i64 [[IDXPROM6]]
7272 // CHECK6-NEXT:    [[TMP23:%.*]] = load float, float* [[ARRAYIDX7]], align 4
7273 // CHECK6-NEXT:    [[ADD8:%.*]] = fadd float [[TMP23]], [[ADD5]]
7274 // CHECK6-NEXT:    store float [[ADD8]], float* [[ARRAYIDX7]], align 4
7275 // CHECK6-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7276 // CHECK6:       omp.body.continue:
7277 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7278 // CHECK6:       omp.inner.for.inc:
7279 // CHECK6-NEXT:    [[TMP24:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7280 // CHECK6-NEXT:    [[ADD9:%.*]] = add i32 [[TMP24]], 1
7281 // CHECK6-NEXT:    store i32 [[ADD9]], i32* [[DOTOMP_IV]], align 4
7282 // CHECK6-NEXT:    br label [[OMP_INNER_FOR_COND]]
7283 // CHECK6:       omp.inner.for.end:
7284 // CHECK6-NEXT:    br label [[OMP_DISPATCH_INC:%.*]]
7285 // CHECK6:       omp.dispatch.inc:
7286 // CHECK6-NEXT:    [[TMP25:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7287 // CHECK6-NEXT:    [[TMP26:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7288 // CHECK6-NEXT:    [[ADD10:%.*]] = add i32 [[TMP25]], [[TMP26]]
7289 // CHECK6-NEXT:    store i32 [[ADD10]], i32* [[DOTOMP_LB]], align 4
7290 // CHECK6-NEXT:    [[TMP27:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7291 // CHECK6-NEXT:    [[TMP28:%.*]] = load i32, i32* [[DOTOMP_STRIDE]], align 4
7292 // CHECK6-NEXT:    [[ADD11:%.*]] = add i32 [[TMP27]], [[TMP28]]
7293 // CHECK6-NEXT:    store i32 [[ADD11]], i32* [[DOTOMP_UB]], align 4
7294 // CHECK6-NEXT:    br label [[OMP_DISPATCH_COND]]
7295 // CHECK6:       omp.dispatch.end:
7296 // CHECK6-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP9]])
7297 // CHECK6-NEXT:    [[TMP29:%.*]] = load i8*, i8** [[SAVED_STACK]], align 8
7298 // CHECK6-NEXT:    call void @llvm.stackrestore(i8* [[TMP29]])
7299 // CHECK6-NEXT:    ret void
7300 //
7301 //
7302 // CHECK11-LABEL: define {{[^@]+}}@_Z9incrementv
7303 // CHECK11-SAME: () #[[ATTR0:[0-9]+]] {
7304 // CHECK11-NEXT:  entry:
7305 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7306 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7307 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7308 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7309 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7310 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7311 // CHECK11-NEXT:    [[I:%.*]] = alloca i32, align 4
7312 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
7313 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7314 // CHECK11-NEXT:    store i32 4, i32* [[DOTOMP_UB]], align 4
7315 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7316 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7317 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7318 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7319 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 4
7320 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7321 // CHECK11:       cond.true:
7322 // CHECK11-NEXT:    br label [[COND_END:%.*]]
7323 // CHECK11:       cond.false:
7324 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7325 // CHECK11-NEXT:    br label [[COND_END]]
7326 // CHECK11:       cond.end:
7327 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 4, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
7328 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7329 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7330 // CHECK11-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_IV]], align 4
7331 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7332 // CHECK11:       omp.inner.for.cond:
7333 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7334 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7335 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
7336 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7337 // CHECK11:       omp.inner.for.body:
7338 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7339 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
7340 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7341 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
7342 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7343 // CHECK11:       omp.body.continue:
7344 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7345 // CHECK11:       omp.inner.for.inc:
7346 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7347 // CHECK11-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP7]], 1
7348 // CHECK11-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
7349 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
7350 // CHECK11:       omp.inner.for.end:
7351 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7352 // CHECK11:       omp.loop.exit:
7353 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
7354 // CHECK11-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]])
7355 // CHECK11-NEXT:    ret i32 0
7356 //
7357 //
7358 // CHECK11-LABEL: define {{[^@]+}}@_Z16decrement_nowaitv
7359 // CHECK11-SAME: () #[[ATTR0]] {
7360 // CHECK11-NEXT:  entry:
7361 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7362 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7363 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7364 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7365 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7366 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7367 // CHECK11-NEXT:    [[J:%.*]] = alloca i32, align 4
7368 // CHECK11-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
7369 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7370 // CHECK11-NEXT:    store i32 4, i32* [[DOTOMP_UB]], align 4
7371 // CHECK11-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7372 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7373 // CHECK11-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7374 // CHECK11-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7375 // CHECK11-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 4
7376 // CHECK11-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7377 // CHECK11:       cond.true:
7378 // CHECK11-NEXT:    br label [[COND_END:%.*]]
7379 // CHECK11:       cond.false:
7380 // CHECK11-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7381 // CHECK11-NEXT:    br label [[COND_END]]
7382 // CHECK11:       cond.end:
7383 // CHECK11-NEXT:    [[COND:%.*]] = phi i32 [ 4, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
7384 // CHECK11-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7385 // CHECK11-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7386 // CHECK11-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_IV]], align 4
7387 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7388 // CHECK11:       omp.inner.for.cond:
7389 // CHECK11-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7390 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7391 // CHECK11-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
7392 // CHECK11-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7393 // CHECK11:       omp.inner.for.body:
7394 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7395 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
7396 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i32 5, [[MUL]]
7397 // CHECK11-NEXT:    store i32 [[SUB]], i32* [[J]], align 4
7398 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7399 // CHECK11:       omp.body.continue:
7400 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7401 // CHECK11:       omp.inner.for.inc:
7402 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7403 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
7404 // CHECK11-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7405 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
7406 // CHECK11:       omp.inner.for.end:
7407 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7408 // CHECK11:       omp.loop.exit:
7409 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
7410 // CHECK11-NEXT:    ret i32 0
7411 //
7412 //
7413 // CHECK11-LABEL: define {{[^@]+}}@_Z16range_for_singlev
7414 // CHECK11-SAME: () #[[ATTR3:[0-9]+]] {
7415 // CHECK11-NEXT:  entry:
7416 // CHECK11-NEXT:    [[ARR:%.*]] = alloca [10 x i32], align 16
7417 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
7418 // CHECK11-NEXT:    [[TMP0:%.*]] = bitcast [10 x i32]* [[ARR]] to i8*
7419 // CHECK11-NEXT:    call void @llvm.memset.p0i8.i64(i8* align 16 [[TMP0]], i8 0, i64 40, i1 false)
7420 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7421 // CHECK11-NEXT:    store [10 x i32]* [[ARR]], [10 x i32]** [[TMP1]], align 8
7422 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
7423 // CHECK11-NEXT:    ret void
7424 //
7425 //
7426 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined.
7427 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR5:[0-9]+]] {
7428 // CHECK11-NEXT:  entry:
7429 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7430 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7431 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
7432 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
7433 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32*, align 8
7434 // CHECK11-NEXT:    [[__RANGE1:%.*]] = alloca [10 x i32]*, align 8
7435 // CHECK11-NEXT:    [[__END1:%.*]] = alloca i32*, align 8
7436 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32*, align 8
7437 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32*, align 8
7438 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i64, align 8
7439 // CHECK11-NEXT:    [[__BEGIN1:%.*]] = alloca i32*, align 8
7440 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
7441 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
7442 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
7443 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7444 // CHECK11-NEXT:    [[__BEGIN15:%.*]] = alloca i32*, align 8
7445 // CHECK11-NEXT:    [[A:%.*]] = alloca i32*, align 8
7446 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7447 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7448 // CHECK11-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
7449 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
7450 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
7451 // CHECK11-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP1]], align 8
7452 // CHECK11-NEXT:    store [10 x i32]* [[TMP2]], [10 x i32]** [[__RANGE1]], align 8
7453 // CHECK11-NEXT:    [[TMP3:%.*]] = load [10 x i32]*, [10 x i32]** [[__RANGE1]], align 8
7454 // CHECK11-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP3]], i64 0, i64 0
7455 // CHECK11-NEXT:    [[ADD_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAYDECAY]], i64 10
7456 // CHECK11-NEXT:    store i32* [[ADD_PTR]], i32** [[__END1]], align 8
7457 // CHECK11-NEXT:    [[TMP4:%.*]] = load [10 x i32]*, [10 x i32]** [[__RANGE1]], align 8
7458 // CHECK11-NEXT:    [[ARRAYDECAY1:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP4]], i64 0, i64 0
7459 // CHECK11-NEXT:    store i32* [[ARRAYDECAY1]], i32** [[DOTCAPTURE_EXPR_]], align 8
7460 // CHECK11-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[__END1]], align 8
7461 // CHECK11-NEXT:    store i32* [[TMP5]], i32** [[DOTCAPTURE_EXPR_2]], align 8
7462 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_2]], align 8
7463 // CHECK11-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
7464 // CHECK11-NEXT:    [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint i32* [[TMP6]] to i64
7465 // CHECK11-NEXT:    [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint i32* [[TMP7]] to i64
7466 // CHECK11-NEXT:    [[SUB_PTR_SUB:%.*]] = sub i64 [[SUB_PTR_LHS_CAST]], [[SUB_PTR_RHS_CAST]]
7467 // CHECK11-NEXT:    [[SUB_PTR_DIV:%.*]] = sdiv exact i64 [[SUB_PTR_SUB]], 4
7468 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i64 [[SUB_PTR_DIV]], 1
7469 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 [[SUB]], 1
7470 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i64 [[ADD]], 1
7471 // CHECK11-NEXT:    [[SUB4:%.*]] = sub nsw i64 [[DIV]], 1
7472 // CHECK11-NEXT:    store i64 [[SUB4]], i64* [[DOTCAPTURE_EXPR_3]], align 8
7473 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
7474 // CHECK11-NEXT:    store i32* [[TMP8]], i32** [[__BEGIN1]], align 8
7475 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
7476 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_2]], align 8
7477 // CHECK11-NEXT:    [[CMP:%.*]] = icmp ult i32* [[TMP9]], [[TMP10]]
7478 // CHECK11-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
7479 // CHECK11:       omp.precond.then:
7480 // CHECK11-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
7481 // CHECK11-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_3]], align 8
7482 // CHECK11-NEXT:    store i64 [[TMP11]], i64* [[DOTOMP_UB]], align 8
7483 // CHECK11-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
7484 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7485 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7486 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
7487 // CHECK11-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
7488 // CHECK11-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
7489 // CHECK11-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_3]], align 8
7490 // CHECK11-NEXT:    [[CMP6:%.*]] = icmp sgt i64 [[TMP14]], [[TMP15]]
7491 // CHECK11-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7492 // CHECK11:       cond.true:
7493 // CHECK11-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_3]], align 8
7494 // CHECK11-NEXT:    br label [[COND_END:%.*]]
7495 // CHECK11:       cond.false:
7496 // CHECK11-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
7497 // CHECK11-NEXT:    br label [[COND_END]]
7498 // CHECK11:       cond.end:
7499 // CHECK11-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP16]], [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
7500 // CHECK11-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
7501 // CHECK11-NEXT:    [[TMP18:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
7502 // CHECK11-NEXT:    store i64 [[TMP18]], i64* [[DOTOMP_IV]], align 8
7503 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7504 // CHECK11:       omp.inner.for.cond:
7505 // CHECK11-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
7506 // CHECK11-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
7507 // CHECK11-NEXT:    [[CMP7:%.*]] = icmp sle i64 [[TMP19]], [[TMP20]]
7508 // CHECK11-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7509 // CHECK11:       omp.inner.for.body:
7510 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
7511 // CHECK11-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
7512 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP22]], 1
7513 // CHECK11-NEXT:    [[ADD_PTR8:%.*]] = getelementptr inbounds i32, i32* [[TMP21]], i64 [[MUL]]
7514 // CHECK11-NEXT:    store i32* [[ADD_PTR8]], i32** [[__BEGIN15]], align 8
7515 // CHECK11-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[__BEGIN15]], align 8
7516 // CHECK11-NEXT:    store i32* [[TMP23]], i32** [[A]], align 8
7517 // CHECK11-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[A]], align 8
7518 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7519 // CHECK11:       omp.body.continue:
7520 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7521 // CHECK11:       omp.inner.for.inc:
7522 // CHECK11-NEXT:    [[TMP25:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
7523 // CHECK11-NEXT:    [[ADD9:%.*]] = add nsw i64 [[TMP25]], 1
7524 // CHECK11-NEXT:    store i64 [[ADD9]], i64* [[DOTOMP_IV]], align 8
7525 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
7526 // CHECK11:       omp.inner.for.end:
7527 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7528 // CHECK11:       omp.loop.exit:
7529 // CHECK11-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7530 // CHECK11-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
7531 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP27]])
7532 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
7533 // CHECK11:       omp.precond.end:
7534 // CHECK11-NEXT:    ret void
7535 //
7536 //
7537 // CHECK11-LABEL: define {{[^@]+}}@_Z19range_for_collapsedv
7538 // CHECK11-SAME: () #[[ATTR3]] {
7539 // CHECK11-NEXT:  entry:
7540 // CHECK11-NEXT:    [[ARR:%.*]] = alloca [10 x i32], align 16
7541 // CHECK11-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
7542 // CHECK11-NEXT:    [[TMP0:%.*]] = bitcast [10 x i32]* [[ARR]] to i8*
7543 // CHECK11-NEXT:    call void @llvm.memset.p0i8.i64(i8* align 16 [[TMP0]], i8 0, i64 40, i1 false)
7544 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7545 // CHECK11-NEXT:    store [10 x i32]* [[ARR]], [10 x i32]** [[TMP1]], align 8
7546 // CHECK11-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
7547 // CHECK11-NEXT:    ret void
7548 //
7549 //
7550 // CHECK11-LABEL: define {{[^@]+}}@.omp_outlined..1
7551 // CHECK11-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR5]] {
7552 // CHECK11-NEXT:  entry:
7553 // CHECK11-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7554 // CHECK11-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7555 // CHECK11-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
7556 // CHECK11-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
7557 // CHECK11-NEXT:    [[TMP:%.*]] = alloca i32*, align 8
7558 // CHECK11-NEXT:    [[_TMP1:%.*]] = alloca i32*, align 8
7559 // CHECK11-NEXT:    [[__RANGE1:%.*]] = alloca [10 x i32]*, align 8
7560 // CHECK11-NEXT:    [[__END1:%.*]] = alloca i32*, align 8
7561 // CHECK11-NEXT:    [[__RANGE2:%.*]] = alloca [10 x i32]*, align 8
7562 // CHECK11-NEXT:    [[__END2:%.*]] = alloca i32*, align 8
7563 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32*, align 8
7564 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32*, align 8
7565 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_6:%.*]] = alloca i32*, align 8
7566 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_8:%.*]] = alloca i32*, align 8
7567 // CHECK11-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i64, align 8
7568 // CHECK11-NEXT:    [[__BEGIN1:%.*]] = alloca i32*, align 8
7569 // CHECK11-NEXT:    [[__BEGIN2:%.*]] = alloca i32*, align 8
7570 // CHECK11-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
7571 // CHECK11-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
7572 // CHECK11-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
7573 // CHECK11-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7574 // CHECK11-NEXT:    [[__BEGIN119:%.*]] = alloca i32*, align 8
7575 // CHECK11-NEXT:    [[__BEGIN220:%.*]] = alloca i32*, align 8
7576 // CHECK11-NEXT:    [[A:%.*]] = alloca i32*, align 8
7577 // CHECK11-NEXT:    [[B:%.*]] = alloca i32, align 4
7578 // CHECK11-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7579 // CHECK11-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7580 // CHECK11-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
7581 // CHECK11-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
7582 // CHECK11-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
7583 // CHECK11-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP1]], align 8
7584 // CHECK11-NEXT:    store [10 x i32]* [[TMP2]], [10 x i32]** [[__RANGE1]], align 8
7585 // CHECK11-NEXT:    [[TMP3:%.*]] = load [10 x i32]*, [10 x i32]** [[__RANGE1]], align 8
7586 // CHECK11-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP3]], i64 0, i64 0
7587 // CHECK11-NEXT:    [[ADD_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAYDECAY]], i64 10
7588 // CHECK11-NEXT:    store i32* [[ADD_PTR]], i32** [[__END1]], align 8
7589 // CHECK11-NEXT:    store [10 x i32]* [[TMP2]], [10 x i32]** [[__RANGE2]], align 8
7590 // CHECK11-NEXT:    [[TMP4:%.*]] = load [10 x i32]*, [10 x i32]** [[__RANGE2]], align 8
7591 // CHECK11-NEXT:    [[ARRAYDECAY2:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP4]], i64 0, i64 0
7592 // CHECK11-NEXT:    [[ADD_PTR3:%.*]] = getelementptr inbounds i32, i32* [[ARRAYDECAY2]], i64 10
7593 // CHECK11-NEXT:    store i32* [[ADD_PTR3]], i32** [[__END2]], align 8
7594 // CHECK11-NEXT:    [[TMP5:%.*]] = load [10 x i32]*, [10 x i32]** [[__RANGE1]], align 8
7595 // CHECK11-NEXT:    [[ARRAYDECAY4:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP5]], i64 0, i64 0
7596 // CHECK11-NEXT:    store i32* [[ARRAYDECAY4]], i32** [[DOTCAPTURE_EXPR_]], align 8
7597 // CHECK11-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[__END1]], align 8
7598 // CHECK11-NEXT:    store i32* [[TMP6]], i32** [[DOTCAPTURE_EXPR_5]], align 8
7599 // CHECK11-NEXT:    [[TMP7:%.*]] = load [10 x i32]*, [10 x i32]** [[__RANGE2]], align 8
7600 // CHECK11-NEXT:    [[ARRAYDECAY7:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP7]], i64 0, i64 0
7601 // CHECK11-NEXT:    store i32* [[ARRAYDECAY7]], i32** [[DOTCAPTURE_EXPR_6]], align 8
7602 // CHECK11-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[__END2]], align 8
7603 // CHECK11-NEXT:    store i32* [[TMP8]], i32** [[DOTCAPTURE_EXPR_8]], align 8
7604 // CHECK11-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_5]], align 8
7605 // CHECK11-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
7606 // CHECK11-NEXT:    [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint i32* [[TMP9]] to i64
7607 // CHECK11-NEXT:    [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint i32* [[TMP10]] to i64
7608 // CHECK11-NEXT:    [[SUB_PTR_SUB:%.*]] = sub i64 [[SUB_PTR_LHS_CAST]], [[SUB_PTR_RHS_CAST]]
7609 // CHECK11-NEXT:    [[SUB_PTR_DIV:%.*]] = sdiv exact i64 [[SUB_PTR_SUB]], 4
7610 // CHECK11-NEXT:    [[SUB:%.*]] = sub nsw i64 [[SUB_PTR_DIV]], 1
7611 // CHECK11-NEXT:    [[ADD:%.*]] = add nsw i64 [[SUB]], 1
7612 // CHECK11-NEXT:    [[DIV:%.*]] = sdiv i64 [[ADD]], 1
7613 // CHECK11-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_8]], align 8
7614 // CHECK11-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
7615 // CHECK11-NEXT:    [[SUB_PTR_LHS_CAST10:%.*]] = ptrtoint i32* [[TMP11]] to i64
7616 // CHECK11-NEXT:    [[SUB_PTR_RHS_CAST11:%.*]] = ptrtoint i32* [[TMP12]] to i64
7617 // CHECK11-NEXT:    [[SUB_PTR_SUB12:%.*]] = sub i64 [[SUB_PTR_LHS_CAST10]], [[SUB_PTR_RHS_CAST11]]
7618 // CHECK11-NEXT:    [[SUB_PTR_DIV13:%.*]] = sdiv exact i64 [[SUB_PTR_SUB12]], 4
7619 // CHECK11-NEXT:    [[SUB14:%.*]] = sub nsw i64 [[SUB_PTR_DIV13]], 1
7620 // CHECK11-NEXT:    [[ADD15:%.*]] = add nsw i64 [[SUB14]], 1
7621 // CHECK11-NEXT:    [[DIV16:%.*]] = sdiv i64 [[ADD15]], 1
7622 // CHECK11-NEXT:    [[MUL:%.*]] = mul nsw i64 [[DIV]], [[DIV16]]
7623 // CHECK11-NEXT:    [[SUB17:%.*]] = sub nsw i64 [[MUL]], 1
7624 // CHECK11-NEXT:    store i64 [[SUB17]], i64* [[DOTCAPTURE_EXPR_9]], align 8
7625 // CHECK11-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
7626 // CHECK11-NEXT:    store i32* [[TMP13]], i32** [[__BEGIN1]], align 8
7627 // CHECK11-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
7628 // CHECK11-NEXT:    store i32* [[TMP14]], i32** [[__BEGIN2]], align 8
7629 // CHECK11-NEXT:    [[TMP15:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
7630 // CHECK11-NEXT:    [[TMP16:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_5]], align 8
7631 // CHECK11-NEXT:    [[CMP:%.*]] = icmp ult i32* [[TMP15]], [[TMP16]]
7632 // CHECK11-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
7633 // CHECK11:       land.lhs.true:
7634 // CHECK11-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
7635 // CHECK11-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_8]], align 8
7636 // CHECK11-NEXT:    [[CMP18:%.*]] = icmp ult i32* [[TMP17]], [[TMP18]]
7637 // CHECK11-NEXT:    br i1 [[CMP18]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
7638 // CHECK11:       omp.precond.then:
7639 // CHECK11-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
7640 // CHECK11-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_9]], align 8
7641 // CHECK11-NEXT:    store i64 [[TMP19]], i64* [[DOTOMP_UB]], align 8
7642 // CHECK11-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
7643 // CHECK11-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7644 // CHECK11-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7645 // CHECK11-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
7646 // CHECK11-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
7647 // CHECK11-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
7648 // CHECK11-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_9]], align 8
7649 // CHECK11-NEXT:    [[CMP21:%.*]] = icmp sgt i64 [[TMP22]], [[TMP23]]
7650 // CHECK11-NEXT:    br i1 [[CMP21]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7651 // CHECK11:       cond.true:
7652 // CHECK11-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_9]], align 8
7653 // CHECK11-NEXT:    br label [[COND_END:%.*]]
7654 // CHECK11:       cond.false:
7655 // CHECK11-NEXT:    [[TMP25:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
7656 // CHECK11-NEXT:    br label [[COND_END]]
7657 // CHECK11:       cond.end:
7658 // CHECK11-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP24]], [[COND_TRUE]] ], [ [[TMP25]], [[COND_FALSE]] ]
7659 // CHECK11-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
7660 // CHECK11-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
7661 // CHECK11-NEXT:    store i64 [[TMP26]], i64* [[DOTOMP_IV]], align 8
7662 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7663 // CHECK11:       omp.inner.for.cond:
7664 // CHECK11-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
7665 // CHECK11-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
7666 // CHECK11-NEXT:    [[CMP22:%.*]] = icmp sle i64 [[TMP27]], [[TMP28]]
7667 // CHECK11-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7668 // CHECK11:       omp.inner.for.body:
7669 // CHECK11-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
7670 // CHECK11-NEXT:    [[TMP30:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
7671 // CHECK11-NEXT:    [[TMP31:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_8]], align 8
7672 // CHECK11-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
7673 // CHECK11-NEXT:    [[SUB_PTR_LHS_CAST23:%.*]] = ptrtoint i32* [[TMP31]] to i64
7674 // CHECK11-NEXT:    [[SUB_PTR_RHS_CAST24:%.*]] = ptrtoint i32* [[TMP32]] to i64
7675 // CHECK11-NEXT:    [[SUB_PTR_SUB25:%.*]] = sub i64 [[SUB_PTR_LHS_CAST23]], [[SUB_PTR_RHS_CAST24]]
7676 // CHECK11-NEXT:    [[SUB_PTR_DIV26:%.*]] = sdiv exact i64 [[SUB_PTR_SUB25]], 4
7677 // CHECK11-NEXT:    [[SUB27:%.*]] = sub nsw i64 [[SUB_PTR_DIV26]], 1
7678 // CHECK11-NEXT:    [[ADD28:%.*]] = add nsw i64 [[SUB27]], 1
7679 // CHECK11-NEXT:    [[DIV29:%.*]] = sdiv i64 [[ADD28]], 1
7680 // CHECK11-NEXT:    [[MUL30:%.*]] = mul nsw i64 1, [[DIV29]]
7681 // CHECK11-NEXT:    [[DIV31:%.*]] = sdiv i64 [[TMP30]], [[MUL30]]
7682 // CHECK11-NEXT:    [[MUL32:%.*]] = mul nsw i64 [[DIV31]], 1
7683 // CHECK11-NEXT:    [[ADD_PTR33:%.*]] = getelementptr inbounds i32, i32* [[TMP29]], i64 [[MUL32]]
7684 // CHECK11-NEXT:    store i32* [[ADD_PTR33]], i32** [[__BEGIN119]], align 8
7685 // CHECK11-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
7686 // CHECK11-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
7687 // CHECK11-NEXT:    [[TMP35:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
7688 // CHECK11-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_8]], align 8
7689 // CHECK11-NEXT:    [[TMP37:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
7690 // CHECK11-NEXT:    [[SUB_PTR_LHS_CAST34:%.*]] = ptrtoint i32* [[TMP36]] to i64
7691 // CHECK11-NEXT:    [[SUB_PTR_RHS_CAST35:%.*]] = ptrtoint i32* [[TMP37]] to i64
7692 // CHECK11-NEXT:    [[SUB_PTR_SUB36:%.*]] = sub i64 [[SUB_PTR_LHS_CAST34]], [[SUB_PTR_RHS_CAST35]]
7693 // CHECK11-NEXT:    [[SUB_PTR_DIV37:%.*]] = sdiv exact i64 [[SUB_PTR_SUB36]], 4
7694 // CHECK11-NEXT:    [[SUB38:%.*]] = sub nsw i64 [[SUB_PTR_DIV37]], 1
7695 // CHECK11-NEXT:    [[ADD39:%.*]] = add nsw i64 [[SUB38]], 1
7696 // CHECK11-NEXT:    [[DIV40:%.*]] = sdiv i64 [[ADD39]], 1
7697 // CHECK11-NEXT:    [[MUL41:%.*]] = mul nsw i64 1, [[DIV40]]
7698 // CHECK11-NEXT:    [[DIV42:%.*]] = sdiv i64 [[TMP35]], [[MUL41]]
7699 // CHECK11-NEXT:    [[TMP38:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_8]], align 8
7700 // CHECK11-NEXT:    [[TMP39:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
7701 // CHECK11-NEXT:    [[SUB_PTR_LHS_CAST43:%.*]] = ptrtoint i32* [[TMP38]] to i64
7702 // CHECK11-NEXT:    [[SUB_PTR_RHS_CAST44:%.*]] = ptrtoint i32* [[TMP39]] to i64
7703 // CHECK11-NEXT:    [[SUB_PTR_SUB45:%.*]] = sub i64 [[SUB_PTR_LHS_CAST43]], [[SUB_PTR_RHS_CAST44]]
7704 // CHECK11-NEXT:    [[SUB_PTR_DIV46:%.*]] = sdiv exact i64 [[SUB_PTR_SUB45]], 4
7705 // CHECK11-NEXT:    [[SUB47:%.*]] = sub nsw i64 [[SUB_PTR_DIV46]], 1
7706 // CHECK11-NEXT:    [[ADD48:%.*]] = add nsw i64 [[SUB47]], 1
7707 // CHECK11-NEXT:    [[DIV49:%.*]] = sdiv i64 [[ADD48]], 1
7708 // CHECK11-NEXT:    [[MUL50:%.*]] = mul nsw i64 1, [[DIV49]]
7709 // CHECK11-NEXT:    [[MUL51:%.*]] = mul nsw i64 [[DIV42]], [[MUL50]]
7710 // CHECK11-NEXT:    [[SUB52:%.*]] = sub nsw i64 [[TMP34]], [[MUL51]]
7711 // CHECK11-NEXT:    [[MUL53:%.*]] = mul nsw i64 [[SUB52]], 1
7712 // CHECK11-NEXT:    [[ADD_PTR54:%.*]] = getelementptr inbounds i32, i32* [[TMP33]], i64 [[MUL53]]
7713 // CHECK11-NEXT:    store i32* [[ADD_PTR54]], i32** [[__BEGIN220]], align 8
7714 // CHECK11-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[__BEGIN119]], align 8
7715 // CHECK11-NEXT:    store i32* [[TMP40]], i32** [[A]], align 8
7716 // CHECK11-NEXT:    [[TMP41:%.*]] = load i32*, i32** [[__BEGIN220]], align 8
7717 // CHECK11-NEXT:    [[TMP42:%.*]] = load i32, i32* [[TMP41]], align 4
7718 // CHECK11-NEXT:    store i32 [[TMP42]], i32* [[B]], align 4
7719 // CHECK11-NEXT:    [[TMP43:%.*]] = load i32, i32* [[B]], align 4
7720 // CHECK11-NEXT:    [[TMP44:%.*]] = load i32*, i32** [[A]], align 8
7721 // CHECK11-NEXT:    store i32 [[TMP43]], i32* [[TMP44]], align 4
7722 // CHECK11-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7723 // CHECK11:       omp.body.continue:
7724 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7725 // CHECK11:       omp.inner.for.inc:
7726 // CHECK11-NEXT:    [[TMP45:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
7727 // CHECK11-NEXT:    [[ADD55:%.*]] = add nsw i64 [[TMP45]], 1
7728 // CHECK11-NEXT:    store i64 [[ADD55]], i64* [[DOTOMP_IV]], align 8
7729 // CHECK11-NEXT:    br label [[OMP_INNER_FOR_COND]]
7730 // CHECK11:       omp.inner.for.end:
7731 // CHECK11-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7732 // CHECK11:       omp.loop.exit:
7733 // CHECK11-NEXT:    [[TMP46:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7734 // CHECK11-NEXT:    [[TMP47:%.*]] = load i32, i32* [[TMP46]], align 4
7735 // CHECK11-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP47]])
7736 // CHECK11-NEXT:    br label [[OMP_PRECOND_END]]
7737 // CHECK11:       omp.precond.end:
7738 // CHECK11-NEXT:    ret void
7739 //
7740 //
7741 // CHECK12-LABEL: define {{[^@]+}}@_Z9incrementv
7742 // CHECK12-SAME: () #[[ATTR0:[0-9]+]] {
7743 // CHECK12-NEXT:  entry:
7744 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7745 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7746 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7747 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7748 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7749 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7750 // CHECK12-NEXT:    [[I:%.*]] = alloca i32, align 4
7751 // CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2:[0-9]+]])
7752 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7753 // CHECK12-NEXT:    store i32 4, i32* [[DOTOMP_UB]], align 4
7754 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7755 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7756 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7757 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7758 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 4
7759 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7760 // CHECK12:       cond.true:
7761 // CHECK12-NEXT:    br label [[COND_END:%.*]]
7762 // CHECK12:       cond.false:
7763 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7764 // CHECK12-NEXT:    br label [[COND_END]]
7765 // CHECK12:       cond.end:
7766 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 4, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
7767 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7768 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7769 // CHECK12-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_IV]], align 4
7770 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7771 // CHECK12:       omp.inner.for.cond:
7772 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7773 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7774 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
7775 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7776 // CHECK12:       omp.inner.for.body:
7777 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7778 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
7779 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 0, [[MUL]]
7780 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[I]], align 4
7781 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7782 // CHECK12:       omp.body.continue:
7783 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7784 // CHECK12:       omp.inner.for.inc:
7785 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7786 // CHECK12-NEXT:    [[ADD2:%.*]] = add nsw i32 [[TMP7]], 1
7787 // CHECK12-NEXT:    store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
7788 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
7789 // CHECK12:       omp.inner.for.end:
7790 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7791 // CHECK12:       omp.loop.exit:
7792 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
7793 // CHECK12-NEXT:    call void @__kmpc_barrier(%struct.ident_t* @[[GLOB3:[0-9]+]], i32 [[TMP0]])
7794 // CHECK12-NEXT:    ret i32 0
7795 //
7796 //
7797 // CHECK12-LABEL: define {{[^@]+}}@_Z16decrement_nowaitv
7798 // CHECK12-SAME: () #[[ATTR0]] {
7799 // CHECK12-NEXT:  entry:
7800 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i32, align 4
7801 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32, align 4
7802 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i32, align 4
7803 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i32, align 4
7804 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
7805 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7806 // CHECK12-NEXT:    [[J:%.*]] = alloca i32, align 4
7807 // CHECK12-NEXT:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @[[GLOB2]])
7808 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_LB]], align 4
7809 // CHECK12-NEXT:    store i32 4, i32* [[DOTOMP_UB]], align 4
7810 // CHECK12-NEXT:    store i32 1, i32* [[DOTOMP_STRIDE]], align 4
7811 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7812 // CHECK12-NEXT:    call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
7813 // CHECK12-NEXT:    [[TMP1:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7814 // CHECK12-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[TMP1]], 4
7815 // CHECK12-NEXT:    br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7816 // CHECK12:       cond.true:
7817 // CHECK12-NEXT:    br label [[COND_END:%.*]]
7818 // CHECK12:       cond.false:
7819 // CHECK12-NEXT:    [[TMP2:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7820 // CHECK12-NEXT:    br label [[COND_END]]
7821 // CHECK12:       cond.end:
7822 // CHECK12-NEXT:    [[COND:%.*]] = phi i32 [ 4, [[COND_TRUE]] ], [ [[TMP2]], [[COND_FALSE]] ]
7823 // CHECK12-NEXT:    store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
7824 // CHECK12-NEXT:    [[TMP3:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
7825 // CHECK12-NEXT:    store i32 [[TMP3]], i32* [[DOTOMP_IV]], align 4
7826 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7827 // CHECK12:       omp.inner.for.cond:
7828 // CHECK12-NEXT:    [[TMP4:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7829 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
7830 // CHECK12-NEXT:    [[CMP1:%.*]] = icmp sle i32 [[TMP4]], [[TMP5]]
7831 // CHECK12-NEXT:    br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7832 // CHECK12:       omp.inner.for.body:
7833 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7834 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i32 [[TMP6]], 1
7835 // CHECK12-NEXT:    [[SUB:%.*]] = sub nsw i32 5, [[MUL]]
7836 // CHECK12-NEXT:    store i32 [[SUB]], i32* [[J]], align 4
7837 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7838 // CHECK12:       omp.body.continue:
7839 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7840 // CHECK12:       omp.inner.for.inc:
7841 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
7842 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP7]], 1
7843 // CHECK12-NEXT:    store i32 [[ADD]], i32* [[DOTOMP_IV]], align 4
7844 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
7845 // CHECK12:       omp.inner.for.end:
7846 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7847 // CHECK12:       omp.loop.exit:
7848 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]])
7849 // CHECK12-NEXT:    ret i32 0
7850 //
7851 //
7852 // CHECK12-LABEL: define {{[^@]+}}@_Z16range_for_singlev
7853 // CHECK12-SAME: () #[[ATTR3:[0-9]+]] {
7854 // CHECK12-NEXT:  entry:
7855 // CHECK12-NEXT:    [[ARR:%.*]] = alloca [10 x i32], align 16
7856 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8
7857 // CHECK12-NEXT:    [[TMP0:%.*]] = bitcast [10 x i32]* [[ARR]] to i8*
7858 // CHECK12-NEXT:    call void @llvm.memset.p0i8.i64(i8* align 16 [[TMP0]], i8 0, i64 40, i1 false)
7859 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON]], %struct.anon* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7860 // CHECK12-NEXT:    store [10 x i32]* [[ARR]], [10 x i32]** [[TMP1]], align 8
7861 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon*)* @.omp_outlined. to void (i32*, i32*, ...)*), %struct.anon* [[OMP_OUTLINED_ARG_AGG_]])
7862 // CHECK12-NEXT:    ret void
7863 //
7864 //
7865 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined.
7866 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon* noalias [[__CONTEXT:%.*]]) #[[ATTR5:[0-9]+]] {
7867 // CHECK12-NEXT:  entry:
7868 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7869 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7870 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon*, align 8
7871 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
7872 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32*, align 8
7873 // CHECK12-NEXT:    [[__RANGE1:%.*]] = alloca [10 x i32]*, align 8
7874 // CHECK12-NEXT:    [[__END1:%.*]] = alloca i32*, align 8
7875 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32*, align 8
7876 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32*, align 8
7877 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR_3:%.*]] = alloca i64, align 8
7878 // CHECK12-NEXT:    [[__BEGIN1:%.*]] = alloca i32*, align 8
7879 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
7880 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
7881 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
7882 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
7883 // CHECK12-NEXT:    [[__BEGIN15:%.*]] = alloca i32*, align 8
7884 // CHECK12-NEXT:    [[A:%.*]] = alloca i32*, align 8
7885 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
7886 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
7887 // CHECK12-NEXT:    store %struct.anon* [[__CONTEXT]], %struct.anon** [[__CONTEXT_ADDR]], align 8
7888 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon*, %struct.anon** [[__CONTEXT_ADDR]], align 8
7889 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON:%.*]], %struct.anon* [[TMP0]], i32 0, i32 0
7890 // CHECK12-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP1]], align 8
7891 // CHECK12-NEXT:    store [10 x i32]* [[TMP2]], [10 x i32]** [[__RANGE1]], align 8
7892 // CHECK12-NEXT:    [[TMP3:%.*]] = load [10 x i32]*, [10 x i32]** [[__RANGE1]], align 8
7893 // CHECK12-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP3]], i64 0, i64 0
7894 // CHECK12-NEXT:    [[ADD_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAYDECAY]], i64 10
7895 // CHECK12-NEXT:    store i32* [[ADD_PTR]], i32** [[__END1]], align 8
7896 // CHECK12-NEXT:    [[TMP4:%.*]] = load [10 x i32]*, [10 x i32]** [[__RANGE1]], align 8
7897 // CHECK12-NEXT:    [[ARRAYDECAY1:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP4]], i64 0, i64 0
7898 // CHECK12-NEXT:    store i32* [[ARRAYDECAY1]], i32** [[DOTCAPTURE_EXPR_]], align 8
7899 // CHECK12-NEXT:    [[TMP5:%.*]] = load i32*, i32** [[__END1]], align 8
7900 // CHECK12-NEXT:    store i32* [[TMP5]], i32** [[DOTCAPTURE_EXPR_2]], align 8
7901 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_2]], align 8
7902 // CHECK12-NEXT:    [[TMP7:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
7903 // CHECK12-NEXT:    [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint i32* [[TMP6]] to i64
7904 // CHECK12-NEXT:    [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint i32* [[TMP7]] to i64
7905 // CHECK12-NEXT:    [[SUB_PTR_SUB:%.*]] = sub i64 [[SUB_PTR_LHS_CAST]], [[SUB_PTR_RHS_CAST]]
7906 // CHECK12-NEXT:    [[SUB_PTR_DIV:%.*]] = sdiv exact i64 [[SUB_PTR_SUB]], 4
7907 // CHECK12-NEXT:    [[SUB:%.*]] = sub nsw i64 [[SUB_PTR_DIV]], 1
7908 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i64 [[SUB]], 1
7909 // CHECK12-NEXT:    [[DIV:%.*]] = sdiv i64 [[ADD]], 1
7910 // CHECK12-NEXT:    [[SUB4:%.*]] = sub nsw i64 [[DIV]], 1
7911 // CHECK12-NEXT:    store i64 [[SUB4]], i64* [[DOTCAPTURE_EXPR_3]], align 8
7912 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
7913 // CHECK12-NEXT:    store i32* [[TMP8]], i32** [[__BEGIN1]], align 8
7914 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
7915 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_2]], align 8
7916 // CHECK12-NEXT:    [[CMP:%.*]] = icmp ult i32* [[TMP9]], [[TMP10]]
7917 // CHECK12-NEXT:    br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
7918 // CHECK12:       omp.precond.then:
7919 // CHECK12-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
7920 // CHECK12-NEXT:    [[TMP11:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_3]], align 8
7921 // CHECK12-NEXT:    store i64 [[TMP11]], i64* [[DOTOMP_UB]], align 8
7922 // CHECK12-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
7923 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
7924 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7925 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32, i32* [[TMP12]], align 4
7926 // CHECK12-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP13]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
7927 // CHECK12-NEXT:    [[TMP14:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
7928 // CHECK12-NEXT:    [[TMP15:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_3]], align 8
7929 // CHECK12-NEXT:    [[CMP6:%.*]] = icmp sgt i64 [[TMP14]], [[TMP15]]
7930 // CHECK12-NEXT:    br i1 [[CMP6]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
7931 // CHECK12:       cond.true:
7932 // CHECK12-NEXT:    [[TMP16:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_3]], align 8
7933 // CHECK12-NEXT:    br label [[COND_END:%.*]]
7934 // CHECK12:       cond.false:
7935 // CHECK12-NEXT:    [[TMP17:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
7936 // CHECK12-NEXT:    br label [[COND_END]]
7937 // CHECK12:       cond.end:
7938 // CHECK12-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP16]], [[COND_TRUE]] ], [ [[TMP17]], [[COND_FALSE]] ]
7939 // CHECK12-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
7940 // CHECK12-NEXT:    [[TMP18:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
7941 // CHECK12-NEXT:    store i64 [[TMP18]], i64* [[DOTOMP_IV]], align 8
7942 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
7943 // CHECK12:       omp.inner.for.cond:
7944 // CHECK12-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
7945 // CHECK12-NEXT:    [[TMP20:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
7946 // CHECK12-NEXT:    [[CMP7:%.*]] = icmp sle i64 [[TMP19]], [[TMP20]]
7947 // CHECK12-NEXT:    br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
7948 // CHECK12:       omp.inner.for.body:
7949 // CHECK12-NEXT:    [[TMP21:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
7950 // CHECK12-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
7951 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i64 [[TMP22]], 1
7952 // CHECK12-NEXT:    [[ADD_PTR8:%.*]] = getelementptr inbounds i32, i32* [[TMP21]], i64 [[MUL]]
7953 // CHECK12-NEXT:    store i32* [[ADD_PTR8]], i32** [[__BEGIN15]], align 8
7954 // CHECK12-NEXT:    [[TMP23:%.*]] = load i32*, i32** [[__BEGIN15]], align 8
7955 // CHECK12-NEXT:    store i32* [[TMP23]], i32** [[A]], align 8
7956 // CHECK12-NEXT:    [[TMP24:%.*]] = load i32*, i32** [[A]], align 8
7957 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
7958 // CHECK12:       omp.body.continue:
7959 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
7960 // CHECK12:       omp.inner.for.inc:
7961 // CHECK12-NEXT:    [[TMP25:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
7962 // CHECK12-NEXT:    [[ADD9:%.*]] = add nsw i64 [[TMP25]], 1
7963 // CHECK12-NEXT:    store i64 [[ADD9]], i64* [[DOTOMP_IV]], align 8
7964 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
7965 // CHECK12:       omp.inner.for.end:
7966 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
7967 // CHECK12:       omp.loop.exit:
7968 // CHECK12-NEXT:    [[TMP26:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
7969 // CHECK12-NEXT:    [[TMP27:%.*]] = load i32, i32* [[TMP26]], align 4
7970 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP27]])
7971 // CHECK12-NEXT:    br label [[OMP_PRECOND_END]]
7972 // CHECK12:       omp.precond.end:
7973 // CHECK12-NEXT:    ret void
7974 //
7975 //
7976 // CHECK12-LABEL: define {{[^@]+}}@_Z19range_for_collapsedv
7977 // CHECK12-SAME: () #[[ATTR3]] {
7978 // CHECK12-NEXT:  entry:
7979 // CHECK12-NEXT:    [[ARR:%.*]] = alloca [10 x i32], align 16
7980 // CHECK12-NEXT:    [[OMP_OUTLINED_ARG_AGG_:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 8
7981 // CHECK12-NEXT:    [[TMP0:%.*]] = bitcast [10 x i32]* [[ARR]] to i8*
7982 // CHECK12-NEXT:    call void @llvm.memset.p0i8.i64(i8* align 16 [[TMP0]], i8 0, i64 40, i1 false)
7983 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0]], %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]], i32 0, i32 0
7984 // CHECK12-NEXT:    store [10 x i32]* [[ARR]], [10 x i32]** [[TMP1]], align 8
7985 // CHECK12-NEXT:    call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, %struct.anon.0*)* @.omp_outlined..1 to void (i32*, i32*, ...)*), %struct.anon.0* [[OMP_OUTLINED_ARG_AGG_]])
7986 // CHECK12-NEXT:    ret void
7987 //
7988 //
7989 // CHECK12-LABEL: define {{[^@]+}}@.omp_outlined..1
7990 // CHECK12-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], %struct.anon.0* noalias [[__CONTEXT:%.*]]) #[[ATTR5]] {
7991 // CHECK12-NEXT:  entry:
7992 // CHECK12-NEXT:    [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
7993 // CHECK12-NEXT:    [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
7994 // CHECK12-NEXT:    [[__CONTEXT_ADDR:%.*]] = alloca %struct.anon.0*, align 8
7995 // CHECK12-NEXT:    [[DOTOMP_IV:%.*]] = alloca i64, align 8
7996 // CHECK12-NEXT:    [[TMP:%.*]] = alloca i32*, align 8
7997 // CHECK12-NEXT:    [[_TMP1:%.*]] = alloca i32*, align 8
7998 // CHECK12-NEXT:    [[__RANGE1:%.*]] = alloca [10 x i32]*, align 8
7999 // CHECK12-NEXT:    [[__END1:%.*]] = alloca i32*, align 8
8000 // CHECK12-NEXT:    [[__RANGE2:%.*]] = alloca [10 x i32]*, align 8
8001 // CHECK12-NEXT:    [[__END2:%.*]] = alloca i32*, align 8
8002 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32*, align 8
8003 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR_5:%.*]] = alloca i32*, align 8
8004 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR_6:%.*]] = alloca i32*, align 8
8005 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR_8:%.*]] = alloca i32*, align 8
8006 // CHECK12-NEXT:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i64, align 8
8007 // CHECK12-NEXT:    [[__BEGIN1:%.*]] = alloca i32*, align 8
8008 // CHECK12-NEXT:    [[__BEGIN2:%.*]] = alloca i32*, align 8
8009 // CHECK12-NEXT:    [[DOTOMP_LB:%.*]] = alloca i64, align 8
8010 // CHECK12-NEXT:    [[DOTOMP_UB:%.*]] = alloca i64, align 8
8011 // CHECK12-NEXT:    [[DOTOMP_STRIDE:%.*]] = alloca i64, align 8
8012 // CHECK12-NEXT:    [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
8013 // CHECK12-NEXT:    [[__BEGIN119:%.*]] = alloca i32*, align 8
8014 // CHECK12-NEXT:    [[__BEGIN220:%.*]] = alloca i32*, align 8
8015 // CHECK12-NEXT:    [[A:%.*]] = alloca i32*, align 8
8016 // CHECK12-NEXT:    [[B:%.*]] = alloca i32, align 4
8017 // CHECK12-NEXT:    store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
8018 // CHECK12-NEXT:    store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
8019 // CHECK12-NEXT:    store %struct.anon.0* [[__CONTEXT]], %struct.anon.0** [[__CONTEXT_ADDR]], align 8
8020 // CHECK12-NEXT:    [[TMP0:%.*]] = load %struct.anon.0*, %struct.anon.0** [[__CONTEXT_ADDR]], align 8
8021 // CHECK12-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [[STRUCT_ANON_0:%.*]], %struct.anon.0* [[TMP0]], i32 0, i32 0
8022 // CHECK12-NEXT:    [[TMP2:%.*]] = load [10 x i32]*, [10 x i32]** [[TMP1]], align 8
8023 // CHECK12-NEXT:    store [10 x i32]* [[TMP2]], [10 x i32]** [[__RANGE1]], align 8
8024 // CHECK12-NEXT:    [[TMP3:%.*]] = load [10 x i32]*, [10 x i32]** [[__RANGE1]], align 8
8025 // CHECK12-NEXT:    [[ARRAYDECAY:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP3]], i64 0, i64 0
8026 // CHECK12-NEXT:    [[ADD_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAYDECAY]], i64 10
8027 // CHECK12-NEXT:    store i32* [[ADD_PTR]], i32** [[__END1]], align 8
8028 // CHECK12-NEXT:    store [10 x i32]* [[TMP2]], [10 x i32]** [[__RANGE2]], align 8
8029 // CHECK12-NEXT:    [[TMP4:%.*]] = load [10 x i32]*, [10 x i32]** [[__RANGE2]], align 8
8030 // CHECK12-NEXT:    [[ARRAYDECAY2:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP4]], i64 0, i64 0
8031 // CHECK12-NEXT:    [[ADD_PTR3:%.*]] = getelementptr inbounds i32, i32* [[ARRAYDECAY2]], i64 10
8032 // CHECK12-NEXT:    store i32* [[ADD_PTR3]], i32** [[__END2]], align 8
8033 // CHECK12-NEXT:    [[TMP5:%.*]] = load [10 x i32]*, [10 x i32]** [[__RANGE1]], align 8
8034 // CHECK12-NEXT:    [[ARRAYDECAY4:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP5]], i64 0, i64 0
8035 // CHECK12-NEXT:    store i32* [[ARRAYDECAY4]], i32** [[DOTCAPTURE_EXPR_]], align 8
8036 // CHECK12-NEXT:    [[TMP6:%.*]] = load i32*, i32** [[__END1]], align 8
8037 // CHECK12-NEXT:    store i32* [[TMP6]], i32** [[DOTCAPTURE_EXPR_5]], align 8
8038 // CHECK12-NEXT:    [[TMP7:%.*]] = load [10 x i32]*, [10 x i32]** [[__RANGE2]], align 8
8039 // CHECK12-NEXT:    [[ARRAYDECAY7:%.*]] = getelementptr inbounds [10 x i32], [10 x i32]* [[TMP7]], i64 0, i64 0
8040 // CHECK12-NEXT:    store i32* [[ARRAYDECAY7]], i32** [[DOTCAPTURE_EXPR_6]], align 8
8041 // CHECK12-NEXT:    [[TMP8:%.*]] = load i32*, i32** [[__END2]], align 8
8042 // CHECK12-NEXT:    store i32* [[TMP8]], i32** [[DOTCAPTURE_EXPR_8]], align 8
8043 // CHECK12-NEXT:    [[TMP9:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_5]], align 8
8044 // CHECK12-NEXT:    [[TMP10:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
8045 // CHECK12-NEXT:    [[SUB_PTR_LHS_CAST:%.*]] = ptrtoint i32* [[TMP9]] to i64
8046 // CHECK12-NEXT:    [[SUB_PTR_RHS_CAST:%.*]] = ptrtoint i32* [[TMP10]] to i64
8047 // CHECK12-NEXT:    [[SUB_PTR_SUB:%.*]] = sub i64 [[SUB_PTR_LHS_CAST]], [[SUB_PTR_RHS_CAST]]
8048 // CHECK12-NEXT:    [[SUB_PTR_DIV:%.*]] = sdiv exact i64 [[SUB_PTR_SUB]], 4
8049 // CHECK12-NEXT:    [[SUB:%.*]] = sub nsw i64 [[SUB_PTR_DIV]], 1
8050 // CHECK12-NEXT:    [[ADD:%.*]] = add nsw i64 [[SUB]], 1
8051 // CHECK12-NEXT:    [[DIV:%.*]] = sdiv i64 [[ADD]], 1
8052 // CHECK12-NEXT:    [[TMP11:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_8]], align 8
8053 // CHECK12-NEXT:    [[TMP12:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
8054 // CHECK12-NEXT:    [[SUB_PTR_LHS_CAST10:%.*]] = ptrtoint i32* [[TMP11]] to i64
8055 // CHECK12-NEXT:    [[SUB_PTR_RHS_CAST11:%.*]] = ptrtoint i32* [[TMP12]] to i64
8056 // CHECK12-NEXT:    [[SUB_PTR_SUB12:%.*]] = sub i64 [[SUB_PTR_LHS_CAST10]], [[SUB_PTR_RHS_CAST11]]
8057 // CHECK12-NEXT:    [[SUB_PTR_DIV13:%.*]] = sdiv exact i64 [[SUB_PTR_SUB12]], 4
8058 // CHECK12-NEXT:    [[SUB14:%.*]] = sub nsw i64 [[SUB_PTR_DIV13]], 1
8059 // CHECK12-NEXT:    [[ADD15:%.*]] = add nsw i64 [[SUB14]], 1
8060 // CHECK12-NEXT:    [[DIV16:%.*]] = sdiv i64 [[ADD15]], 1
8061 // CHECK12-NEXT:    [[MUL:%.*]] = mul nsw i64 [[DIV]], [[DIV16]]
8062 // CHECK12-NEXT:    [[SUB17:%.*]] = sub nsw i64 [[MUL]], 1
8063 // CHECK12-NEXT:    store i64 [[SUB17]], i64* [[DOTCAPTURE_EXPR_9]], align 8
8064 // CHECK12-NEXT:    [[TMP13:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
8065 // CHECK12-NEXT:    store i32* [[TMP13]], i32** [[__BEGIN1]], align 8
8066 // CHECK12-NEXT:    [[TMP14:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
8067 // CHECK12-NEXT:    store i32* [[TMP14]], i32** [[__BEGIN2]], align 8
8068 // CHECK12-NEXT:    [[TMP15:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
8069 // CHECK12-NEXT:    [[TMP16:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_5]], align 8
8070 // CHECK12-NEXT:    [[CMP:%.*]] = icmp ult i32* [[TMP15]], [[TMP16]]
8071 // CHECK12-NEXT:    br i1 [[CMP]], label [[LAND_LHS_TRUE:%.*]], label [[OMP_PRECOND_END:%.*]]
8072 // CHECK12:       land.lhs.true:
8073 // CHECK12-NEXT:    [[TMP17:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
8074 // CHECK12-NEXT:    [[TMP18:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_8]], align 8
8075 // CHECK12-NEXT:    [[CMP18:%.*]] = icmp ult i32* [[TMP17]], [[TMP18]]
8076 // CHECK12-NEXT:    br i1 [[CMP18]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END]]
8077 // CHECK12:       omp.precond.then:
8078 // CHECK12-NEXT:    store i64 0, i64* [[DOTOMP_LB]], align 8
8079 // CHECK12-NEXT:    [[TMP19:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_9]], align 8
8080 // CHECK12-NEXT:    store i64 [[TMP19]], i64* [[DOTOMP_UB]], align 8
8081 // CHECK12-NEXT:    store i64 1, i64* [[DOTOMP_STRIDE]], align 8
8082 // CHECK12-NEXT:    store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
8083 // CHECK12-NEXT:    [[TMP20:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8084 // CHECK12-NEXT:    [[TMP21:%.*]] = load i32, i32* [[TMP20]], align 4
8085 // CHECK12-NEXT:    call void @__kmpc_for_static_init_8(%struct.ident_t* @[[GLOB1]], i32 [[TMP21]], i32 34, i32* [[DOTOMP_IS_LAST]], i64* [[DOTOMP_LB]], i64* [[DOTOMP_UB]], i64* [[DOTOMP_STRIDE]], i64 1, i64 1)
8086 // CHECK12-NEXT:    [[TMP22:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
8087 // CHECK12-NEXT:    [[TMP23:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_9]], align 8
8088 // CHECK12-NEXT:    [[CMP21:%.*]] = icmp sgt i64 [[TMP22]], [[TMP23]]
8089 // CHECK12-NEXT:    br i1 [[CMP21]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
8090 // CHECK12:       cond.true:
8091 // CHECK12-NEXT:    [[TMP24:%.*]] = load i64, i64* [[DOTCAPTURE_EXPR_9]], align 8
8092 // CHECK12-NEXT:    br label [[COND_END:%.*]]
8093 // CHECK12:       cond.false:
8094 // CHECK12-NEXT:    [[TMP25:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
8095 // CHECK12-NEXT:    br label [[COND_END]]
8096 // CHECK12:       cond.end:
8097 // CHECK12-NEXT:    [[COND:%.*]] = phi i64 [ [[TMP24]], [[COND_TRUE]] ], [ [[TMP25]], [[COND_FALSE]] ]
8098 // CHECK12-NEXT:    store i64 [[COND]], i64* [[DOTOMP_UB]], align 8
8099 // CHECK12-NEXT:    [[TMP26:%.*]] = load i64, i64* [[DOTOMP_LB]], align 8
8100 // CHECK12-NEXT:    store i64 [[TMP26]], i64* [[DOTOMP_IV]], align 8
8101 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND:%.*]]
8102 // CHECK12:       omp.inner.for.cond:
8103 // CHECK12-NEXT:    [[TMP27:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
8104 // CHECK12-NEXT:    [[TMP28:%.*]] = load i64, i64* [[DOTOMP_UB]], align 8
8105 // CHECK12-NEXT:    [[CMP22:%.*]] = icmp sle i64 [[TMP27]], [[TMP28]]
8106 // CHECK12-NEXT:    br i1 [[CMP22]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
8107 // CHECK12:       omp.inner.for.body:
8108 // CHECK12-NEXT:    [[TMP29:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_]], align 8
8109 // CHECK12-NEXT:    [[TMP30:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
8110 // CHECK12-NEXT:    [[TMP31:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_8]], align 8
8111 // CHECK12-NEXT:    [[TMP32:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
8112 // CHECK12-NEXT:    [[SUB_PTR_LHS_CAST23:%.*]] = ptrtoint i32* [[TMP31]] to i64
8113 // CHECK12-NEXT:    [[SUB_PTR_RHS_CAST24:%.*]] = ptrtoint i32* [[TMP32]] to i64
8114 // CHECK12-NEXT:    [[SUB_PTR_SUB25:%.*]] = sub i64 [[SUB_PTR_LHS_CAST23]], [[SUB_PTR_RHS_CAST24]]
8115 // CHECK12-NEXT:    [[SUB_PTR_DIV26:%.*]] = sdiv exact i64 [[SUB_PTR_SUB25]], 4
8116 // CHECK12-NEXT:    [[SUB27:%.*]] = sub nsw i64 [[SUB_PTR_DIV26]], 1
8117 // CHECK12-NEXT:    [[ADD28:%.*]] = add nsw i64 [[SUB27]], 1
8118 // CHECK12-NEXT:    [[DIV29:%.*]] = sdiv i64 [[ADD28]], 1
8119 // CHECK12-NEXT:    [[MUL30:%.*]] = mul nsw i64 1, [[DIV29]]
8120 // CHECK12-NEXT:    [[DIV31:%.*]] = sdiv i64 [[TMP30]], [[MUL30]]
8121 // CHECK12-NEXT:    [[MUL32:%.*]] = mul nsw i64 [[DIV31]], 1
8122 // CHECK12-NEXT:    [[ADD_PTR33:%.*]] = getelementptr inbounds i32, i32* [[TMP29]], i64 [[MUL32]]
8123 // CHECK12-NEXT:    store i32* [[ADD_PTR33]], i32** [[__BEGIN119]], align 8
8124 // CHECK12-NEXT:    [[TMP33:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
8125 // CHECK12-NEXT:    [[TMP34:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
8126 // CHECK12-NEXT:    [[TMP35:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
8127 // CHECK12-NEXT:    [[TMP36:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_8]], align 8
8128 // CHECK12-NEXT:    [[TMP37:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
8129 // CHECK12-NEXT:    [[SUB_PTR_LHS_CAST34:%.*]] = ptrtoint i32* [[TMP36]] to i64
8130 // CHECK12-NEXT:    [[SUB_PTR_RHS_CAST35:%.*]] = ptrtoint i32* [[TMP37]] to i64
8131 // CHECK12-NEXT:    [[SUB_PTR_SUB36:%.*]] = sub i64 [[SUB_PTR_LHS_CAST34]], [[SUB_PTR_RHS_CAST35]]
8132 // CHECK12-NEXT:    [[SUB_PTR_DIV37:%.*]] = sdiv exact i64 [[SUB_PTR_SUB36]], 4
8133 // CHECK12-NEXT:    [[SUB38:%.*]] = sub nsw i64 [[SUB_PTR_DIV37]], 1
8134 // CHECK12-NEXT:    [[ADD39:%.*]] = add nsw i64 [[SUB38]], 1
8135 // CHECK12-NEXT:    [[DIV40:%.*]] = sdiv i64 [[ADD39]], 1
8136 // CHECK12-NEXT:    [[MUL41:%.*]] = mul nsw i64 1, [[DIV40]]
8137 // CHECK12-NEXT:    [[DIV42:%.*]] = sdiv i64 [[TMP35]], [[MUL41]]
8138 // CHECK12-NEXT:    [[TMP38:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_8]], align 8
8139 // CHECK12-NEXT:    [[TMP39:%.*]] = load i32*, i32** [[DOTCAPTURE_EXPR_6]], align 8
8140 // CHECK12-NEXT:    [[SUB_PTR_LHS_CAST43:%.*]] = ptrtoint i32* [[TMP38]] to i64
8141 // CHECK12-NEXT:    [[SUB_PTR_RHS_CAST44:%.*]] = ptrtoint i32* [[TMP39]] to i64
8142 // CHECK12-NEXT:    [[SUB_PTR_SUB45:%.*]] = sub i64 [[SUB_PTR_LHS_CAST43]], [[SUB_PTR_RHS_CAST44]]
8143 // CHECK12-NEXT:    [[SUB_PTR_DIV46:%.*]] = sdiv exact i64 [[SUB_PTR_SUB45]], 4
8144 // CHECK12-NEXT:    [[SUB47:%.*]] = sub nsw i64 [[SUB_PTR_DIV46]], 1
8145 // CHECK12-NEXT:    [[ADD48:%.*]] = add nsw i64 [[SUB47]], 1
8146 // CHECK12-NEXT:    [[DIV49:%.*]] = sdiv i64 [[ADD48]], 1
8147 // CHECK12-NEXT:    [[MUL50:%.*]] = mul nsw i64 1, [[DIV49]]
8148 // CHECK12-NEXT:    [[MUL51:%.*]] = mul nsw i64 [[DIV42]], [[MUL50]]
8149 // CHECK12-NEXT:    [[SUB52:%.*]] = sub nsw i64 [[TMP34]], [[MUL51]]
8150 // CHECK12-NEXT:    [[MUL53:%.*]] = mul nsw i64 [[SUB52]], 1
8151 // CHECK12-NEXT:    [[ADD_PTR54:%.*]] = getelementptr inbounds i32, i32* [[TMP33]], i64 [[MUL53]]
8152 // CHECK12-NEXT:    store i32* [[ADD_PTR54]], i32** [[__BEGIN220]], align 8
8153 // CHECK12-NEXT:    [[TMP40:%.*]] = load i32*, i32** [[__BEGIN119]], align 8
8154 // CHECK12-NEXT:    store i32* [[TMP40]], i32** [[A]], align 8
8155 // CHECK12-NEXT:    [[TMP41:%.*]] = load i32*, i32** [[__BEGIN220]], align 8
8156 // CHECK12-NEXT:    [[TMP42:%.*]] = load i32, i32* [[TMP41]], align 4
8157 // CHECK12-NEXT:    store i32 [[TMP42]], i32* [[B]], align 4
8158 // CHECK12-NEXT:    [[TMP43:%.*]] = load i32, i32* [[B]], align 4
8159 // CHECK12-NEXT:    [[TMP44:%.*]] = load i32*, i32** [[A]], align 8
8160 // CHECK12-NEXT:    store i32 [[TMP43]], i32* [[TMP44]], align 4
8161 // CHECK12-NEXT:    br label [[OMP_BODY_CONTINUE:%.*]]
8162 // CHECK12:       omp.body.continue:
8163 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_INC:%.*]]
8164 // CHECK12:       omp.inner.for.inc:
8165 // CHECK12-NEXT:    [[TMP45:%.*]] = load i64, i64* [[DOTOMP_IV]], align 8
8166 // CHECK12-NEXT:    [[ADD55:%.*]] = add nsw i64 [[TMP45]], 1
8167 // CHECK12-NEXT:    store i64 [[ADD55]], i64* [[DOTOMP_IV]], align 8
8168 // CHECK12-NEXT:    br label [[OMP_INNER_FOR_COND]]
8169 // CHECK12:       omp.inner.for.end:
8170 // CHECK12-NEXT:    br label [[OMP_LOOP_EXIT:%.*]]
8171 // CHECK12:       omp.loop.exit:
8172 // CHECK12-NEXT:    [[TMP46:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
8173 // CHECK12-NEXT:    [[TMP47:%.*]] = load i32, i32* [[TMP46]], align 4
8174 // CHECK12-NEXT:    call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP47]])
8175 // CHECK12-NEXT:    br label [[OMP_PRECOND_END]]
8176 // CHECK12:       omp.precond.end:
8177 // CHECK12-NEXT:    ret void
8178 //
8179