This is an archive of the discontinued LLVM Phabricator instance.

Differential D104258

[OPENMP]Fix PR50699: capture locals in combine directrives for aligned clause.
ClosedPublic

Authored by ABataev on Jun 14 2021, 12:35 PM.

Details

Reviewers
lebedev.ri
jdoerfert
Commits
rG45ae766e78e0: [OPENMP]Fix PR50699: capture locals in combine directrives for aligned clause.
Summary

Need to capture locals in aligned clauses for the combined directives to
be fix the crash in the codegen.

Diff Detail

Event Timeline

ABataev created this revision.Jun 14 2021, 12:35 PM
Herald added subscribers: guansong, yaxunl. · View Herald TranscriptJun 14 2021, 12:35 PM
ABataev requested review of this revision.Jun 14 2021, 12:35 PM
Herald added a project: Restricted Project. · View Herald TranscriptJun 14 2021, 12:35 PM
Herald added a subscriber: sstefan1. · View Herald Transcript
lebedev.ri added a comment.Jun 14 2021, 12:38 PM

Thank you!

Harbormaster completed remote builds in B109173: Diff 351966.Jun 14 2021, 1:56 PM
jdoerfert accepted this revision.Jun 14 2021, 2:23 PM

LG

This revision is now accepted and ready to land.Jun 14 2021, 2:23 PM
Closed by commit rG45ae766e78e0: [OPENMP]Fix PR50699: capture locals in combine directrives for aligned clause. (authored by ABataev). · Explain WhyJun 15 2021, 4:58 AM
This revision was automatically updated to reflect the committed changes.
ABataev added a commit: rG45ae766e78e0: [OPENMP]Fix PR50699: capture locals in combine directrives for aligned clause..

Revision Contents

PathSize
clang/
lib/
Sema/
7 lines
test/
OpenMP/
346 lines

Diff 352097

clang/lib/Sema/SemaOpenMP.cpp

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 4,602 Lines • ▼ Show 20 Lines if (ThisCaptureRegion == OMPD_target) {
if (Expr *E = D.AllocatorTraits) if (Expr *E = D.AllocatorTraits)
MarkDeclarationsReferencedInExpr(E); MarkDeclarationsReferencedInExpr(E);
} }
continue; continue;
} }
} }
} }
if (ThisCaptureRegion == OMPD_parallel) { if (ThisCaptureRegion == OMPD_parallel) {
// Capture temp arrays for inscan reductions. // Capture temp arrays for inscan reductions and locals in aligned
// clauses.
for (OMPClause *C : Clauses) { for (OMPClause *C : Clauses) {
if (auto *RC = dyn_cast<OMPReductionClause>(C)) { if (auto *RC = dyn_cast<OMPReductionClause>(C)) {
if (RC->getModifier() != OMPC_REDUCTION_inscan) if (RC->getModifier() != OMPC_REDUCTION_inscan)
continue; continue;
for (Expr *E : RC->copy_array_temps()) for (Expr *E : RC->copy_array_temps())
MarkDeclarationsReferencedInExpr(E); MarkDeclarationsReferencedInExpr(E);
} }
if (auto *AC = dyn_cast<OMPAlignedClause>(C)) {
for (Expr *E : AC->varlists())
MarkDeclarationsReferencedInExpr(E);
}
} }
} }
if (++CompletedRegions == CaptureRegions.size()) if (++CompletedRegions == CaptureRegions.size())
DSAStack->setBodyComplete(); DSAStack->setBodyComplete();
SR = ActOnCapturedRegionEnd(SR.get()); SR = ActOnCapturedRegionEnd(SR.get());
} }
return SR; return SR;
} }
▲ Show 20 Lines • Show All 16,615 Lines • Show Last 20 Lines

clang/test/OpenMP/parallel_for_simd_aligned_codegen.cpp

  • This file was added.
// 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 _
// RUN: %clang_cc1 -verify -fopenmp -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK1
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --check-prefix=CHECK2
// RUN: %clang_cc1 -verify -fopenmp -x c++ -std=c++11 -DLAMBDA -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK3
// RUN: %clang_cc1 -verify -fopenmp -x c++ -fblocks -DBLOCKS -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --check-prefix=CHECK4
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -fopenmp-simd -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp-simd -x c++ -triple x86_64-unknown-unknown -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -std=c++11 -DLAMBDA -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// RUN: %clang_cc1 -verify -fopenmp-simd -x c++ -fblocks -DBLOCKS -triple x86_64-unknown-linux -emit-llvm %s -o - | FileCheck %s --implicit-check-not="{{__kmpc|__tgt}}"
// expected-no-diagnostics
#ifndef HEADER
#define HEADER
void foo(float *c) {
#pragma omp parallel for simd aligned(c)
for (int i = 0; i < 10; ++i);
}
#endif
// CHECK1-LABEL: define {{[^@]+}}@_Z3fooPf
// CHECK1-SAME: (float* [[C:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK1-NEXT: entry:
// CHECK1-NEXT: [[C_ADDR:%.*]] = alloca float*, align 8
// CHECK1-NEXT: store float* [[C]], float** [[C_ADDR]], align 8
// CHECK1-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**)* @.omp_outlined. to void (i32*, i32*, ...)*), float** [[C_ADDR]])
// CHECK1-NEXT: ret void
//
//
// CHECK1-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK1-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], float** nonnull align 8 dereferenceable(8) [[C:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK1-NEXT: entry:
// CHECK1-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK1-NEXT: [[C_ADDR:%.*]] = alloca float**, align 8
// CHECK1-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK1-NEXT: [[TMP:%.*]] = alloca i32, align 4
// CHECK1-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK1-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK1-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK1-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK1-NEXT: [[I:%.*]] = alloca i32, align 4
// CHECK1-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK1-NEXT: store float** [[C]], float*** [[C_ADDR]], align 8
// CHECK1-NEXT: [[TMP0:%.*]] = load float**, float*** [[C_ADDR]], align 8
// CHECK1-NEXT: [[TMP1:%.*]] = load float*, float** [[TMP0]], align 8
// CHECK1-NEXT: call void @llvm.assume(i1 true) [ "align"(float* [[TMP1]], i64 16) ]
// CHECK1-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK1-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK1-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK1-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK1-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK1-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK1: cond.true:
// CHECK1-NEXT: br label [[COND_END:%.*]]
// CHECK1: cond.false:
// CHECK1-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT: br label [[COND_END]]
// CHECK1: cond.end:
// CHECK1-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK1-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK1-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK1: omp.inner.for.cond:
// CHECK1-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK1-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK1-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK1: omp.inner.for.body:
// CHECK1-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK1-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK1-NEXT: store i32 [[ADD]], i32* [[I]], align 4
// CHECK1-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK1: omp.body.continue:
// CHECK1-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK1: omp.inner.for.inc:
// CHECK1-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK1-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK1-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
// CHECK1: omp.inner.for.end:
// CHECK1-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
// CHECK1: omp.loop.exit:
// CHECK1-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
// CHECK1-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK1-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK1-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK1: .omp.final.then:
// CHECK1-NEXT: store i32 10, i32* [[I]], align 4
// CHECK1-NEXT: br label [[DOTOMP_FINAL_DONE]]
// CHECK1: .omp.final.done:
// CHECK1-NEXT: ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@_Z3fooPf
// CHECK2-SAME: (float* [[C:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK2-NEXT: entry:
// CHECK2-NEXT: [[C_ADDR:%.*]] = alloca float*, align 8
// CHECK2-NEXT: store float* [[C]], float** [[C_ADDR]], align 8
// CHECK2-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**)* @.omp_outlined. to void (i32*, i32*, ...)*), float** [[C_ADDR]])
// CHECK2-NEXT: ret void
//
//
// CHECK2-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK2-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], float** nonnull align 8 dereferenceable(8) [[C:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK2-NEXT: entry:
// CHECK2-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK2-NEXT: [[C_ADDR:%.*]] = alloca float**, align 8
// CHECK2-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK2-NEXT: [[TMP:%.*]] = alloca i32, align 4
// CHECK2-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK2-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK2-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK2-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK2-NEXT: [[I:%.*]] = alloca i32, align 4
// CHECK2-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK2-NEXT: store float** [[C]], float*** [[C_ADDR]], align 8
// CHECK2-NEXT: [[TMP0:%.*]] = load float**, float*** [[C_ADDR]], align 8
// CHECK2-NEXT: [[TMP1:%.*]] = load float*, float** [[TMP0]], align 8
// CHECK2-NEXT: call void @llvm.assume(i1 true) [ "align"(float* [[TMP1]], i64 16) ]
// CHECK2-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK2-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK2-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK2-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK2-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK2-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK2: cond.true:
// CHECK2-NEXT: br label [[COND_END:%.*]]
// CHECK2: cond.false:
// CHECK2-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT: br label [[COND_END]]
// CHECK2: cond.end:
// CHECK2-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK2-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK2-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK2: omp.inner.for.cond:
// CHECK2-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK2-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK2-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK2: omp.inner.for.body:
// CHECK2-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK2-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK2-NEXT: store i32 [[ADD]], i32* [[I]], align 4
// CHECK2-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK2: omp.body.continue:
// CHECK2-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK2: omp.inner.for.inc:
// CHECK2-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK2-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK2-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
// CHECK2: omp.inner.for.end:
// CHECK2-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
// CHECK2: omp.loop.exit:
// CHECK2-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
// CHECK2-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK2-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK2-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK2: .omp.final.then:
// CHECK2-NEXT: store i32 10, i32* [[I]], align 4
// CHECK2-NEXT: br label [[DOTOMP_FINAL_DONE]]
// CHECK2: .omp.final.done:
// CHECK2-NEXT: ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@_Z3fooPf
// CHECK3-SAME: (float* [[C:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK3-NEXT: entry:
// CHECK3-NEXT: [[C_ADDR:%.*]] = alloca float*, align 8
// CHECK3-NEXT: store float* [[C]], float** [[C_ADDR]], align 8
// CHECK3-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**)* @.omp_outlined. to void (i32*, i32*, ...)*), float** [[C_ADDR]])
// CHECK3-NEXT: ret void
//
//
// CHECK3-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK3-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], float** nonnull align 8 dereferenceable(8) [[C:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK3-NEXT: entry:
// CHECK3-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK3-NEXT: [[C_ADDR:%.*]] = alloca float**, align 8
// CHECK3-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK3-NEXT: [[TMP:%.*]] = alloca i32, align 4
// CHECK3-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK3-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK3-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK3-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK3-NEXT: [[I:%.*]] = alloca i32, align 4
// CHECK3-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK3-NEXT: store float** [[C]], float*** [[C_ADDR]], align 8
// CHECK3-NEXT: [[TMP0:%.*]] = load float**, float*** [[C_ADDR]], align 8
// CHECK3-NEXT: [[TMP1:%.*]] = load float*, float** [[TMP0]], align 8
// CHECK3-NEXT: call void @llvm.assume(i1 true) [ "align"(float* [[TMP1]], i64 16) ]
// CHECK3-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK3-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK3-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK3-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK3-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK3-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK3: cond.true:
// CHECK3-NEXT: br label [[COND_END:%.*]]
// CHECK3: cond.false:
// CHECK3-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT: br label [[COND_END]]
// CHECK3: cond.end:
// CHECK3-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK3-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK3-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK3: omp.inner.for.cond:
// CHECK3-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK3-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK3-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK3: omp.inner.for.body:
// CHECK3-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK3-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK3-NEXT: store i32 [[ADD]], i32* [[I]], align 4
// CHECK3-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK3: omp.body.continue:
// CHECK3-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK3: omp.inner.for.inc:
// CHECK3-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK3-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK3-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
// CHECK3: omp.inner.for.end:
// CHECK3-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
// CHECK3: omp.loop.exit:
// CHECK3-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
// CHECK3-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK3-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK3-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK3: .omp.final.then:
// CHECK3-NEXT: store i32 10, i32* [[I]], align 4
// CHECK3-NEXT: br label [[DOTOMP_FINAL_DONE]]
// CHECK3: .omp.final.done:
// CHECK3-NEXT: ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@_Z3fooPf
// CHECK4-SAME: (float* [[C:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK4-NEXT: entry:
// CHECK4-NEXT: [[C_ADDR:%.*]] = alloca float*, align 8
// CHECK4-NEXT: store float* [[C]], float** [[C_ADDR]], align 8
// CHECK4-NEXT: call void (%struct.ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(%struct.ident_t* @[[GLOB2:[0-9]+]], i32 1, void (i32*, i32*, ...)* bitcast (void (i32*, i32*, float**)* @.omp_outlined. to void (i32*, i32*, ...)*), float** [[C_ADDR]])
// CHECK4-NEXT: ret void
//
//
// CHECK4-LABEL: define {{[^@]+}}@.omp_outlined.
// CHECK4-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]], float** nonnull align 8 dereferenceable(8) [[C:%.*]]) #[[ATTR1:[0-9]+]] {
// CHECK4-NEXT: entry:
// CHECK4-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca i32*, align 8
// CHECK4-NEXT: [[C_ADDR:%.*]] = alloca float**, align 8
// CHECK4-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4
// CHECK4-NEXT: [[TMP:%.*]] = alloca i32, align 4
// CHECK4-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4
// CHECK4-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4
// CHECK4-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4
// CHECK4-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4
// CHECK4-NEXT: [[I:%.*]] = alloca i32, align 4
// CHECK4-NEXT: store i32* [[DOTGLOBAL_TID_]], i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT: store i32* [[DOTBOUND_TID_]], i32** [[DOTBOUND_TID__ADDR]], align 8
// CHECK4-NEXT: store float** [[C]], float*** [[C_ADDR]], align 8
// CHECK4-NEXT: [[TMP0:%.*]] = load float**, float*** [[C_ADDR]], align 8
// CHECK4-NEXT: [[TMP1:%.*]] = load float*, float** [[TMP0]], align 8
// CHECK4-NEXT: call void @llvm.assume(i1 true) [ "align"(float* [[TMP1]], i64 16) ]
// CHECK4-NEXT: store i32 0, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT: store i32 9, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT: store i32 1, i32* [[DOTOMP_STRIDE]], align 4
// CHECK4-NEXT: store i32 0, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT: [[TMP2:%.*]] = load i32*, i32** [[DOTGLOBAL_TID__ADDR]], align 8
// CHECK4-NEXT: [[TMP3:%.*]] = load i32, i32* [[TMP2]], align 4
// CHECK4-NEXT: call void @__kmpc_for_static_init_4(%struct.ident_t* @[[GLOB1:[0-9]+]], i32 [[TMP3]], i32 34, i32* [[DOTOMP_IS_LAST]], i32* [[DOTOMP_LB]], i32* [[DOTOMP_UB]], i32* [[DOTOMP_STRIDE]], i32 1, i32 1)
// CHECK4-NEXT: [[TMP4:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT: [[CMP:%.*]] = icmp sgt i32 [[TMP4]], 9
// CHECK4-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
// CHECK4: cond.true:
// CHECK4-NEXT: br label [[COND_END:%.*]]
// CHECK4: cond.false:
// CHECK4-NEXT: [[TMP5:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT: br label [[COND_END]]
// CHECK4: cond.end:
// CHECK4-NEXT: [[COND:%.*]] = phi i32 [ 9, [[COND_TRUE]] ], [ [[TMP5]], [[COND_FALSE]] ]
// CHECK4-NEXT: store i32 [[COND]], i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT: [[TMP6:%.*]] = load i32, i32* [[DOTOMP_LB]], align 4
// CHECK4-NEXT: store i32 [[TMP6]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
// CHECK4: omp.inner.for.cond:
// CHECK4-NEXT: [[TMP7:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT: [[TMP8:%.*]] = load i32, i32* [[DOTOMP_UB]], align 4
// CHECK4-NEXT: [[CMP1:%.*]] = icmp sle i32 [[TMP7]], [[TMP8]]
// CHECK4-NEXT: br i1 [[CMP1]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
// CHECK4: omp.inner.for.body:
// CHECK4-NEXT: [[TMP9:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP9]], 1
// CHECK4-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
// CHECK4-NEXT: store i32 [[ADD]], i32* [[I]], align 4
// CHECK4-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
// CHECK4: omp.body.continue:
// CHECK4-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
// CHECK4: omp.inner.for.inc:
// CHECK4-NEXT: [[TMP10:%.*]] = load i32, i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT: [[ADD2:%.*]] = add nsw i32 [[TMP10]], 1
// CHECK4-NEXT: store i32 [[ADD2]], i32* [[DOTOMP_IV]], align 4
// CHECK4-NEXT: br label [[OMP_INNER_FOR_COND]], !llvm.loop [[LOOP2:![0-9]+]]
// CHECK4: omp.inner.for.end:
// CHECK4-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
// CHECK4: omp.loop.exit:
// CHECK4-NEXT: call void @__kmpc_for_static_fini(%struct.ident_t* @[[GLOB1]], i32 [[TMP3]])
// CHECK4-NEXT: [[TMP11:%.*]] = load i32, i32* [[DOTOMP_IS_LAST]], align 4
// CHECK4-NEXT: [[TMP12:%.*]] = icmp ne i32 [[TMP11]], 0
// CHECK4-NEXT: br i1 [[TMP12]], label [[DOTOMP_FINAL_THEN:%.*]], label [[DOTOMP_FINAL_DONE:%.*]]
// CHECK4: .omp.final.then:
// CHECK4-NEXT: store i32 10, i32* [[I]], align 4
// CHECK4-NEXT: br label [[DOTOMP_FINAL_DONE]]
// CHECK4: .omp.final.done:
// CHECK4-NEXT: ret void
//