From a3283d64fef21d0ba999679b21ecfaecde3a5376 Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev@hotmail.com>
Date: Tue, 15 Nov 2016 14:26:49 +0000
Subject: [PATCH] Merging r286944:
------------------------------------------------------------------------
r286944 | abataev | 2016-11-15 09:11:50 +0000 (Tue, 15 Nov 2016) | 6 lines
[OPENMP] Fixed codegen for 'omp cancel' construct.
If 'omp cancel' construct is used in a worksharing construct it may cause
hanging of the software in case if reduction clause is used. Patch fixes
this problem by avoiding extra reduction processing for branches that
were canceled.
------------------------------------------------------------------------
git-svn-id: https://llvm.org/svn/llvm-project/cfe/branches/release_39@286968 91177308-0d34-0410-b5e6-96231b3b80d8
---
lib/CodeGen/CGStmtOpenMP.cpp | 41 +++++++++++++++++++++++++++++-----
lib/CodeGen/CodeGenFunction.h | 29 ++++++++++++++++++++++++
test/OpenMP/cancel_codegen.cpp | 7 +++++-
3 files changed, 70 insertions(+), 7 deletions(-)
diff --git a/lib/CodeGen/CGStmtOpenMP.cpp b/lib/CodeGen/CGStmtOpenMP.cpp
index 476b50f8dad..46f00067749 100644
--- a/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/lib/CodeGen/CGStmtOpenMP.cpp
@@ -1767,9 +1767,17 @@ void CodeGenFunction::EmitOMPOuterLoop(bool DynamicOrOrdered, bool IsMonotonic,
EmitBlock(LoopExit.getBlock());
// Tell the runtime we are done.
- if (!DynamicOrOrdered)
- RT.emitForStaticFinish(*this, S.getLocEnd());
+ SourceLocation ELoc = S.getLocEnd();
+ auto &&CodeGen = [DynamicOrOrdered, ELoc](CodeGenFunction &CGF) {
+ if (!DynamicOrOrdered)
+ CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, ELoc);
+ };
+ CodeGen(*this);
+ OpenMPDirectiveKind DKind = S.getDirectiveKind();
+ if (DKind == OMPD_for || DKind == OMPD_parallel_for ||
+ DKind == OMPD_distribute_parallel_for)
+ OMPCancelStack.back().CodeGen = CodeGen;
}
void CodeGenFunction::EmitOMPForOuterLoop(
@@ -1881,6 +1889,7 @@ void CodeGenFunction::EmitOMPDistributeOuterLoop(
void CodeGenFunction::EmitOMPDistributeParallelForDirective(
const OMPDistributeParallelForDirective &S) {
OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
+ OMPCancelStackRAII CancelRegion(*this);
CGM.getOpenMPRuntime().emitInlinedDirective(
*this, OMPD_distribute_parallel_for,
[&S](CodeGenFunction &CGF, PrePostActionTy &) {
@@ -2073,7 +2082,15 @@ bool CodeGenFunction::EmitOMPWorksharingLoop(const OMPLoopDirective &S) {
[](CodeGenFunction &) {});
EmitBlock(LoopExit.getBlock());
// Tell the runtime we are done.
- RT.emitForStaticFinish(*this, S.getLocStart());
+ SourceLocation ELoc = S.getLocEnd();
+ auto &&CodeGen = [ELoc](CodeGenFunction &CGF) {
+ CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, ELoc);
+ };
+ CodeGen(*this);
+ OpenMPDirectiveKind DKind = S.getDirectiveKind();
+ if (DKind == OMPD_for || DKind == OMPD_parallel_for ||
+ DKind == OMPD_distribute_parallel_for)
+ OMPCancelStack.back().CodeGen = CodeGen;
} else {
const bool IsMonotonic =
Ordered || ScheduleKind.Schedule == OMPC_SCHEDULE_static ||
@@ -2127,6 +2144,7 @@ void CodeGenFunction::EmitOMPForDirective(const OMPForDirective &S) {
};
{
OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
+ OMPCancelStackRAII CancelRegion(*this);
CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_for, CodeGen,
S.hasCancel());
}
@@ -2263,7 +2281,12 @@ void CodeGenFunction::EmitSections(const OMPExecutableDirective &S) {
CGF.EmitOMPInnerLoop(S, /*RequiresCleanup=*/false, &Cond, &Inc, BodyGen,
[](CodeGenFunction &) {});
// Tell the runtime we are done.
- CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocStart());
+ SourceLocation ELoc = S.getLocEnd();
+ auto &&CodeGen = [ELoc](CodeGenFunction &CGF) {
+ CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, ELoc);
+ };
+ CodeGen(CGF);
+ CGF.OMPCancelStack.back().CodeGen = CodeGen;
CGF.EmitOMPReductionClauseFinal(S);
// Emit post-update of the reduction variables if IsLastIter != 0.
emitPostUpdateForReductionClause(
@@ -2301,6 +2324,7 @@ void CodeGenFunction::EmitSections(const OMPExecutableDirective &S) {
void CodeGenFunction::EmitOMPSectionsDirective(const OMPSectionsDirective &S) {
{
OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
+ OMPCancelStackRAII CancelRegion(*this);
EmitSections(S);
}
// Emit an implicit barrier at the end.
@@ -2388,6 +2412,7 @@ void CodeGenFunction::EmitOMPParallelForDirective(
// Emit directive as a combined directive that consists of two implicit
// directives: 'parallel' with 'for' directive.
auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+ OMPCancelStackRAII CancelRegion(CGF);
CGF.EmitOMPWorksharingLoop(S);
};
emitCommonOMPParallelDirective(*this, S, OMPD_for, CodeGen);
@@ -2408,6 +2433,7 @@ void CodeGenFunction::EmitOMPParallelSectionsDirective(
// Emit directive as a combined directive that consists of two implicit
// directives: 'parallel' with 'sections' directive.
auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+ OMPCancelStackRAII CancelRegion(CGF);
CGF.EmitSections(S);
};
emitCommonOMPParallelDirective(*this, S, OMPD_sections, CodeGen);
@@ -3390,8 +3416,11 @@ CodeGenFunction::getOMPCancelDestination(OpenMPDirectiveKind Kind) {
if (Kind == OMPD_parallel || Kind == OMPD_task)
return ReturnBlock;
assert(Kind == OMPD_for || Kind == OMPD_section || Kind == OMPD_sections ||
- Kind == OMPD_parallel_sections || Kind == OMPD_parallel_for);
- return BreakContinueStack.back().BreakBlock;
+ Kind == OMPD_parallel_sections || Kind == OMPD_parallel_for ||
+ Kind == OMPD_distribute_parallel_for);
+ if (!OMPCancelStack.back().ExitBlock.isValid())
+ OMPCancelStack.back().ExitBlock = getJumpDestInCurrentScope("cancel.exit");
+ return OMPCancelStack.back().ExitBlock;
}
// Generate the instructions for '#pragma omp target data' directive.
diff --git a/lib/CodeGen/CodeGenFunction.h b/lib/CodeGen/CodeGenFunction.h
index 45068fa5544..4ad4440bc9c 100644
--- a/lib/CodeGen/CodeGenFunction.h
+++ b/lib/CodeGen/CodeGenFunction.h
@@ -965,6 +965,35 @@ private:
};
SmallVector<BreakContinue, 8> BreakContinueStack;
+ /// Data for exit block for proper support of OpenMP cancellation constructs.
+ struct OMPCancel {
+ JumpDest ExitBlock;
+ llvm::function_ref<void(CodeGenFunction &CGF)> CodeGen;
+ OMPCancel() : CodeGen([](CodeGenFunction &CGF) {}) {}
+ };
+ SmallVector<OMPCancel, 8> OMPCancelStack;
+
+ /// Controls insertion of cancellation exit blocks in worksharing constructs.
+ class OMPCancelStackRAII {
+ CodeGenFunction &CGF;
+
+ public:
+ OMPCancelStackRAII(CodeGenFunction &CGF) : CGF(CGF) {
+ CGF.OMPCancelStack.push_back({});
+ }
+ ~OMPCancelStackRAII() {
+ if (CGF.HaveInsertPoint() &&
+ CGF.OMPCancelStack.back().ExitBlock.isValid()) {
+ auto CJD = CGF.getJumpDestInCurrentScope("cancel.cont");
+ CGF.EmitBranchThroughCleanup(CJD);
+ CGF.EmitBlock(CGF.OMPCancelStack.back().ExitBlock.getBlock());
+ CGF.OMPCancelStack.back().CodeGen(CGF);
+ CGF.EmitBranchThroughCleanup(CJD);
+ CGF.EmitBlock(CJD.getBlock());
+ }
+ }
+ };
+
CodeGenPGO PGO;
/// Calculate branch weights appropriate for PGO data
diff --git a/test/OpenMP/cancel_codegen.cpp b/test/OpenMP/cancel_codegen.cpp
index 8234193e8f9..059a8d3901f 100644
--- a/test/OpenMP/cancel_codegen.cpp
+++ b/test/OpenMP/cancel_codegen.cpp
@@ -91,9 +91,11 @@ for (int i = 0; i < argc; ++i) {
}
}
// CHECK: call void (%ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(
-#pragma omp parallel for
+int r = 0;
+#pragma omp parallel for reduction(+:r)
for (int i = 0; i < argc; ++i) {
#pragma omp cancel for
+ r += i;
}
// CHECK: call void (%ident_t*, i32, void (i32*, i32*, ...)*, ...) @__kmpc_fork_call(
return argc;
@@ -164,6 +166,9 @@ for (int i = 0; i < argc; ++i) {
// CHECK: [[CONTINUE]]
// CHECK: br label
// CHECK: call void @__kmpc_for_static_fini(
+// CHECK: call i32 @__kmpc_reduce_nowait(
+// CHECK: call void @__kmpc_end_reduce_nowait(
+// CHECK: call void @__kmpc_for_static_fini(
// CHECK: ret void
#endif
--
GitLab