summaryrefslogtreecommitdiff
path: root/lib/CodeGen/CGStmtOpenMP.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'lib/CodeGen/CGStmtOpenMP.cpp')
-rw-r--r--lib/CodeGen/CGStmtOpenMP.cpp964
1 files changed, 698 insertions, 266 deletions
diff --git a/lib/CodeGen/CGStmtOpenMP.cpp b/lib/CodeGen/CGStmtOpenMP.cpp
index cf430f860fd8a..f04d28ed0d4a9 100644
--- a/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/lib/CodeGen/CGStmtOpenMP.cpp
@@ -65,6 +65,8 @@ public:
for (auto &C : CS->captures()) {
if (C.capturesVariable() || C.capturesVariableByCopy()) {
auto *VD = C.getCapturedVar();
+ assert(VD == VD->getCanonicalDecl() &&
+ "Canonical decl must be captured.");
DeclRefExpr DRE(const_cast<VarDecl *>(VD),
isCapturedVar(CGF, VD) ||
(CGF.CapturedStmtInfo &&
@@ -119,6 +121,14 @@ public:
/// of used expression from loop statement.
class OMPLoopScope : public CodeGenFunction::RunCleanupsScope {
void emitPreInitStmt(CodeGenFunction &CGF, const OMPLoopDirective &S) {
+ CodeGenFunction::OMPPrivateScope PreCondScope(CGF);
+ for (auto *E : S.counters()) {
+ const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
+ (void)PreCondScope.addPrivate(VD, [&CGF, VD]() {
+ return CGF.CreateMemTemp(VD->getType().getNonReferenceType());
+ });
+ }
+ (void)PreCondScope.Privatize();
if (auto *LD = dyn_cast<OMPLoopDirective>(&S)) {
if (auto *PreInits = cast_or_null<DeclStmt>(LD->getPreInits())) {
for (const auto *I : PreInits->decls())
@@ -136,6 +146,26 @@ public:
} // namespace
+static void emitCommonOMPTargetDirective(CodeGenFunction &CGF,
+ const OMPExecutableDirective &S,
+ const RegionCodeGenTy &CodeGen);
+
+LValue CodeGenFunction::EmitOMPSharedLValue(const Expr *E) {
+ if (auto *OrigDRE = dyn_cast<DeclRefExpr>(E)) {
+ if (auto *OrigVD = dyn_cast<VarDecl>(OrigDRE->getDecl())) {
+ OrigVD = OrigVD->getCanonicalDecl();
+ bool IsCaptured =
+ LambdaCaptureFields.lookup(OrigVD) ||
+ (CapturedStmtInfo && CapturedStmtInfo->lookup(OrigVD)) ||
+ (CurCodeDecl && isa<BlockDecl>(CurCodeDecl));
+ DeclRefExpr DRE(const_cast<VarDecl *>(OrigVD), IsCaptured,
+ OrigDRE->getType(), VK_LValue, OrigDRE->getExprLoc());
+ return EmitLValue(&DRE);
+ }
+ }
+ return EmitLValue(E);
+}
+
llvm::Value *CodeGenFunction::getTypeSize(QualType Ty) {
auto &C = getContext();
llvm::Value *Size = nullptr;
@@ -236,6 +266,12 @@ static QualType getCanonicalParamType(ASTContext &C, QualType T) {
}
if (T->isPointerType())
return C.getPointerType(getCanonicalParamType(C, T->getPointeeType()));
+ if (auto *A = T->getAsArrayTypeUnsafe()) {
+ if (auto *VLA = dyn_cast<VariableArrayType>(A))
+ return getCanonicalParamType(C, VLA->getElementType());
+ else if (!A->isVariablyModifiedType())
+ return C.getCanonicalType(T);
+ }
return C.getCanonicalParamType(T);
}
@@ -246,12 +282,12 @@ namespace {
const CapturedStmt *S = nullptr;
/// true if cast to/from UIntPtr is required for variables captured by
/// value.
- bool UIntPtrCastRequired = true;
- /// true if only casted argumefnts must be registered as local args or VLA
+ const bool UIntPtrCastRequired = true;
+ /// true if only casted arguments must be registered as local args or VLA
/// sizes.
- bool RegisterCastedArgsOnly = false;
+ const bool RegisterCastedArgsOnly = false;
/// Name of the generated function.
- StringRef FunctionName;
+ const StringRef FunctionName;
explicit FunctionOptions(const CapturedStmt *S, bool UIntPtrCastRequired,
bool RegisterCastedArgsOnly,
StringRef FunctionName)
@@ -261,9 +297,9 @@ namespace {
};
}
-static std::pair<llvm::Function *, bool> emitOutlinedFunctionPrologue(
+static llvm::Function *emitOutlinedFunctionPrologue(
CodeGenFunction &CGF, FunctionArgList &Args,
- llvm::DenseMap<const Decl *, std::pair<const VarDecl *, Address>>
+ llvm::MapVector<const Decl *, std::pair<const VarDecl *, Address>>
&LocalAddrs,
llvm::DenseMap<const Decl *, std::pair<const Expr *, llvm::Value *>>
&VLASizes,
@@ -276,10 +312,23 @@ static std::pair<llvm::Function *, bool> emitOutlinedFunctionPrologue(
// Build the argument list.
CodeGenModule &CGM = CGF.CGM;
ASTContext &Ctx = CGM.getContext();
- bool HasUIntPtrArgs = false;
+ FunctionArgList TargetArgs;
Args.append(CD->param_begin(),
std::next(CD->param_begin(), CD->getContextParamPosition()));
+ TargetArgs.append(
+ CD->param_begin(),
+ std::next(CD->param_begin(), CD->getContextParamPosition()));
auto I = FO.S->captures().begin();
+ FunctionDecl *DebugFunctionDecl = nullptr;
+ if (!FO.UIntPtrCastRequired) {
+ FunctionProtoType::ExtProtoInfo EPI;
+ DebugFunctionDecl = FunctionDecl::Create(
+ Ctx, Ctx.getTranslationUnitDecl(), FO.S->getLocStart(),
+ SourceLocation(), DeclarationName(), Ctx.VoidTy,
+ Ctx.getTrivialTypeSourceInfo(
+ Ctx.getFunctionType(Ctx.VoidTy, llvm::None, EPI)),
+ SC_Static, /*isInlineSpecified=*/false, /*hasWrittenPrototype=*/false);
+ }
for (auto *FD : RD->fields()) {
QualType ArgType = FD->getType();
IdentifierInfo *II = nullptr;
@@ -292,7 +341,6 @@ static std::pair<llvm::Function *, bool> emitOutlinedFunctionPrologue(
// outlined function.
if ((I->capturesVariableByCopy() && !ArgType->isAnyPointerType()) ||
I->capturesVariableArrayType()) {
- HasUIntPtrArgs = true;
if (FO.UIntPtrCastRequired)
ArgType = Ctx.getUIntPtrType();
}
@@ -307,20 +355,36 @@ static std::pair<llvm::Function *, bool> emitOutlinedFunctionPrologue(
II = &Ctx.Idents.get("vla");
}
if (ArgType->isVariablyModifiedType())
- ArgType = getCanonicalParamType(Ctx, ArgType.getNonReferenceType());
- Args.push_back(ImplicitParamDecl::Create(Ctx, /*DC=*/nullptr,
- FD->getLocation(), II, ArgType,
- ImplicitParamDecl::Other));
+ ArgType = getCanonicalParamType(Ctx, ArgType);
+ VarDecl *Arg;
+ if (DebugFunctionDecl && (CapVar || I->capturesThis())) {
+ Arg = ParmVarDecl::Create(
+ Ctx, DebugFunctionDecl,
+ CapVar ? CapVar->getLocStart() : FD->getLocStart(),
+ CapVar ? CapVar->getLocation() : FD->getLocation(), II, ArgType,
+ /*TInfo=*/nullptr, SC_None, /*DefArg=*/nullptr);
+ } else {
+ Arg = ImplicitParamDecl::Create(Ctx, /*DC=*/nullptr, FD->getLocation(),
+ II, ArgType, ImplicitParamDecl::Other);
+ }
+ Args.emplace_back(Arg);
+ // Do not cast arguments if we emit function with non-original types.
+ TargetArgs.emplace_back(
+ FO.UIntPtrCastRequired
+ ? Arg
+ : CGM.getOpenMPRuntime().translateParameter(FD, Arg));
++I;
}
Args.append(
std::next(CD->param_begin(), CD->getContextParamPosition() + 1),
CD->param_end());
+ TargetArgs.append(
+ std::next(CD->param_begin(), CD->getContextParamPosition() + 1),
+ CD->param_end());
// Create the function declaration.
- FunctionType::ExtInfo ExtInfo;
const CGFunctionInfo &FuncInfo =
- CGM.getTypes().arrangeBuiltinFunctionDeclaration(Ctx.VoidTy, Args);
+ CGM.getTypes().arrangeBuiltinFunctionDeclaration(Ctx.VoidTy, TargetArgs);
llvm::FunctionType *FuncLLVMTy = CGM.getTypes().GetFunctionType(FuncInfo);
llvm::Function *F =
@@ -328,19 +392,26 @@ static std::pair<llvm::Function *, bool> emitOutlinedFunctionPrologue(
FO.FunctionName, &CGM.getModule());
CGM.SetInternalFunctionAttributes(CD, F, FuncInfo);
if (CD->isNothrow())
- F->addFnAttr(llvm::Attribute::NoUnwind);
+ F->setDoesNotThrow();
// Generate the function.
- CGF.StartFunction(CD, Ctx.VoidTy, F, FuncInfo, Args, CD->getLocation(),
- CD->getBody()->getLocStart());
+ CGF.StartFunction(CD, Ctx.VoidTy, F, FuncInfo, TargetArgs,
+ FO.S->getLocStart(), CD->getBody()->getLocStart());
unsigned Cnt = CD->getContextParamPosition();
I = FO.S->captures().begin();
for (auto *FD : RD->fields()) {
+ // Do not map arguments if we emit function with non-original types.
+ Address LocalAddr(Address::invalid());
+ if (!FO.UIntPtrCastRequired && Args[Cnt] != TargetArgs[Cnt]) {
+ LocalAddr = CGM.getOpenMPRuntime().getParameterAddress(CGF, Args[Cnt],
+ TargetArgs[Cnt]);
+ } else {
+ LocalAddr = CGF.GetAddrOfLocalVar(Args[Cnt]);
+ }
// If we are capturing a pointer by copy we don't need to do anything, just
// use the value that we get from the arguments.
if (I->capturesVariableByCopy() && FD->getType()->isAnyPointerType()) {
const VarDecl *CurVD = I->getCapturedVar();
- Address LocalAddr = CGF.GetAddrOfLocalVar(Args[Cnt]);
// If the variable is a reference we need to materialize it here.
if (CurVD->getType()->isReferenceType()) {
Address RefAddr = CGF.CreateMemTemp(
@@ -356,15 +427,14 @@ static std::pair<llvm::Function *, bool> emitOutlinedFunctionPrologue(
continue;
}
- LValueBaseInfo BaseInfo(AlignmentSource::Decl, false);
- LValue ArgLVal = CGF.MakeAddrLValue(CGF.GetAddrOfLocalVar(Args[Cnt]),
- Args[Cnt]->getType(), BaseInfo);
+ LValue ArgLVal = CGF.MakeAddrLValue(LocalAddr, Args[Cnt]->getType(),
+ AlignmentSource::Decl);
if (FD->hasCapturedVLAType()) {
if (FO.UIntPtrCastRequired) {
ArgLVal = CGF.MakeAddrLValue(castValueFromUintptr(CGF, FD->getType(),
Args[Cnt]->getName(),
ArgLVal),
- FD->getType(), BaseInfo);
+ FD->getType(), AlignmentSource::Decl);
}
auto *ExprArg =
CGF.EmitLoadOfLValue(ArgLVal, SourceLocation()).getScalarVal();
@@ -376,8 +446,7 @@ static std::pair<llvm::Function *, bool> emitOutlinedFunctionPrologue(
Address ArgAddr = ArgLVal.getAddress();
if (!VarTy->isReferenceType()) {
if (ArgLVal.getType()->isLValueReferenceType()) {
- ArgAddr = CGF.EmitLoadOfReference(
- ArgAddr, ArgLVal.getType()->castAs<ReferenceType>());
+ ArgAddr = CGF.EmitLoadOfReference(ArgLVal);
} else if (!VarTy->isVariablyModifiedType() || !VarTy->isPointerType()) {
assert(ArgLVal.getType()->isPointerType());
ArgAddr = CGF.EmitLoadOfPointer(
@@ -412,7 +481,7 @@ static std::pair<llvm::Function *, bool> emitOutlinedFunctionPrologue(
++I;
}
- return {F, HasUIntPtrArgs};
+ return F;
}
llvm::Function *
@@ -426,14 +495,17 @@ CodeGenFunction::GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S) {
getDebugInfo() &&
CGM.getCodeGenOpts().getDebugInfo() >= codegenoptions::LimitedDebugInfo;
FunctionArgList Args;
- llvm::DenseMap<const Decl *, std::pair<const VarDecl *, Address>> LocalAddrs;
+ llvm::MapVector<const Decl *, std::pair<const VarDecl *, Address>> LocalAddrs;
llvm::DenseMap<const Decl *, std::pair<const Expr *, llvm::Value *>> VLASizes;
+ SmallString<256> Buffer;
+ llvm::raw_svector_ostream Out(Buffer);
+ Out << CapturedStmtInfo->getHelperName();
+ if (NeedWrapperFunction)
+ Out << "_debug__";
FunctionOptions FO(&S, !NeedWrapperFunction, /*RegisterCastedArgsOnly=*/false,
- CapturedStmtInfo->getHelperName());
- llvm::Function *F;
- bool HasUIntPtrArgs;
- std::tie(F, HasUIntPtrArgs) = emitOutlinedFunctionPrologue(
- *this, Args, LocalAddrs, VLASizes, CXXThisValue, FO);
+ Out.str());
+ llvm::Function *F = emitOutlinedFunctionPrologue(*this, Args, LocalAddrs,
+ VLASizes, CXXThisValue, FO);
for (const auto &LocalAddrPair : LocalAddrs) {
if (LocalAddrPair.second.first) {
setAddrOfLocalVar(LocalAddrPair.second.first,
@@ -445,28 +517,28 @@ CodeGenFunction::GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S) {
PGO.assignRegionCounters(GlobalDecl(CD), F);
CapturedStmtInfo->EmitBody(*this, CD->getBody());
FinishFunction(CD->getBodyRBrace());
- if (!NeedWrapperFunction || !HasUIntPtrArgs)
+ if (!NeedWrapperFunction)
return F;
FunctionOptions WrapperFO(&S, /*UIntPtrCastRequired=*/true,
/*RegisterCastedArgsOnly=*/true,
- ".nondebug_wrapper.");
+ CapturedStmtInfo->getHelperName());
CodeGenFunction WrapperCGF(CGM, /*suppressNewContext=*/true);
- WrapperCGF.disableDebugInfo();
Args.clear();
LocalAddrs.clear();
VLASizes.clear();
llvm::Function *WrapperF =
emitOutlinedFunctionPrologue(WrapperCGF, Args, LocalAddrs, VLASizes,
- WrapperCGF.CXXThisValue, WrapperFO).first;
- LValueBaseInfo BaseInfo(AlignmentSource::Decl, false);
+ WrapperCGF.CXXThisValue, WrapperFO);
llvm::SmallVector<llvm::Value *, 4> CallArgs;
for (const auto *Arg : Args) {
llvm::Value *CallArg;
auto I = LocalAddrs.find(Arg);
if (I != LocalAddrs.end()) {
- LValue LV =
- WrapperCGF.MakeAddrLValue(I->second.second, Arg->getType(), BaseInfo);
+ LValue LV = WrapperCGF.MakeAddrLValue(
+ I->second.second,
+ I->second.first ? I->second.first->getType() : Arg->getType(),
+ AlignmentSource::Decl);
CallArg = WrapperCGF.EmitLoadOfScalar(LV, SourceLocation());
} else {
auto EI = VLASizes.find(Arg);
@@ -474,13 +546,15 @@ CodeGenFunction::GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S) {
CallArg = EI->second.second;
else {
LValue LV = WrapperCGF.MakeAddrLValue(WrapperCGF.GetAddrOfLocalVar(Arg),
- Arg->getType(), BaseInfo);
+ Arg->getType(),
+ AlignmentSource::Decl);
CallArg = WrapperCGF.EmitLoadOfScalar(LV, SourceLocation());
}
}
- CallArgs.emplace_back(CallArg);
+ CallArgs.emplace_back(WrapperCGF.EmitFromMemory(CallArg, Arg->getType()));
}
- WrapperCGF.Builder.CreateCall(F, CallArgs);
+ CGM.getOpenMPRuntime().emitOutlinedFunctionCall(WrapperCGF, S.getLocStart(),
+ F, CallArgs);
WrapperCGF.FinishFunction();
return WrapperF;
}
@@ -956,7 +1030,9 @@ void CodeGenFunction::EmitOMPReductionClauseInit(
auto *LHSVD = cast<VarDecl>(cast<DeclRefExpr>(*ILHS)->getDecl());
auto *RHSVD = cast<VarDecl>(cast<DeclRefExpr>(*IRHS)->getDecl());
- if (isa<OMPArraySectionExpr>(IRef)) {
+ QualType Type = PrivateVD->getType();
+ bool isaOMPArraySectionExpr = isa<OMPArraySectionExpr>(IRef);
+ if (isaOMPArraySectionExpr && Type->isVariablyModifiedType()) {
// Store the address of the original variable associated with the LHS
// implicit variable.
PrivateScope.addPrivate(LHSVD, [&RedCG, Count]() -> Address {
@@ -965,7 +1041,8 @@ void CodeGenFunction::EmitOMPReductionClauseInit(
PrivateScope.addPrivate(RHSVD, [this, PrivateVD]() -> Address {
return GetAddrOfLocalVar(PrivateVD);
});
- } else if (isa<ArraySubscriptExpr>(IRef)) {
+ } else if ((isaOMPArraySectionExpr && Type->isScalarType()) ||
+ isa<ArraySubscriptExpr>(IRef)) {
// Store the address of the original variable associated with the LHS
// implicit variable.
PrivateScope.addPrivate(LHSVD, [&RedCG, Count]() -> Address {
@@ -1024,7 +1101,8 @@ void CodeGenFunction::EmitOMPReductionClauseFinal(
bool WithNowait = D.getSingleClause<OMPNowaitClause>() ||
isOpenMPParallelDirective(D.getDirectiveKind()) ||
D.getDirectiveKind() == OMPD_simd;
- bool SimpleReduction = D.getDirectiveKind() == OMPD_simd;
+ bool SimpleReduction = D.getDirectiveKind() == OMPD_simd ||
+ D.getDirectiveKind() == OMPD_distribute_simd;
// Emit nowait reduction if nowait clause is present or directive is a
// parallel directive (it always has implicit barrier).
CGM.getOpenMPRuntime().emitReduction(
@@ -1146,9 +1224,13 @@ void CodeGenFunction::EmitOMPLoopBody(const OMPLoopDirective &D,
EmitIgnoredExpr(I);
}
// Update the linear variables.
- for (const auto *C : D.getClausesOfKind<OMPLinearClause>()) {
- for (auto *U : C->updates())
- EmitIgnoredExpr(U);
+ // In distribute directives only loop counters may be marked as linear, no
+ // need to generate the code for them.
+ if (!isOpenMPDistributeDirective(D.getDirectiveKind())) {
+ for (const auto *C : D.getClausesOfKind<OMPLinearClause>()) {
+ for (auto *U : C->updates())
+ EmitIgnoredExpr(U);
+ }
}
// On a continue in the body, jump to the end.
@@ -1488,83 +1570,90 @@ static void emitOMPLoopBodyWithStopPoint(CodeGenFunction &CGF,
CGF.EmitStopPoint(&S);
}
-void CodeGenFunction::EmitOMPSimdDirective(const OMPSimdDirective &S) {
- auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
- OMPLoopScope PreInitScope(CGF, S);
- // if (PreCond) {
- // for (IV in 0..LastIteration) BODY;
- // <Final counter/linear vars updates>;
- // }
- //
+static void emitOMPSimdRegion(CodeGenFunction &CGF, const OMPLoopDirective &S,
+ PrePostActionTy &Action) {
+ Action.Enter(CGF);
+ assert(isOpenMPSimdDirective(S.getDirectiveKind()) &&
+ "Expected simd directive");
+ OMPLoopScope PreInitScope(CGF, S);
+ // if (PreCond) {
+ // for (IV in 0..LastIteration) BODY;
+ // <Final counter/linear vars updates>;
+ // }
+ //
- // Emit: if (PreCond) - begin.
- // If the condition constant folds and can be elided, avoid emitting the
- // whole loop.
- bool CondConstant;
- llvm::BasicBlock *ContBlock = nullptr;
- if (CGF.ConstantFoldsToSimpleInteger(S.getPreCond(), CondConstant)) {
- if (!CondConstant)
- return;
- } else {
- auto *ThenBlock = CGF.createBasicBlock("simd.if.then");
- ContBlock = CGF.createBasicBlock("simd.if.end");
- emitPreCond(CGF, S, S.getPreCond(), ThenBlock, ContBlock,
- CGF.getProfileCount(&S));
- CGF.EmitBlock(ThenBlock);
- CGF.incrementProfileCounter(&S);
- }
+ // Emit: if (PreCond) - begin.
+ // If the condition constant folds and can be elided, avoid emitting the
+ // whole loop.
+ bool CondConstant;
+ llvm::BasicBlock *ContBlock = nullptr;
+ if (CGF.ConstantFoldsToSimpleInteger(S.getPreCond(), CondConstant)) {
+ if (!CondConstant)
+ return;
+ } else {
+ auto *ThenBlock = CGF.createBasicBlock("simd.if.then");
+ ContBlock = CGF.createBasicBlock("simd.if.end");
+ emitPreCond(CGF, S, S.getPreCond(), ThenBlock, ContBlock,
+ CGF.getProfileCount(&S));
+ CGF.EmitBlock(ThenBlock);
+ CGF.incrementProfileCounter(&S);
+ }
- // Emit the loop iteration variable.
- const Expr *IVExpr = S.getIterationVariable();
- const VarDecl *IVDecl = cast<VarDecl>(cast<DeclRefExpr>(IVExpr)->getDecl());
- CGF.EmitVarDecl(*IVDecl);
- CGF.EmitIgnoredExpr(S.getInit());
+ // Emit the loop iteration variable.
+ const Expr *IVExpr = S.getIterationVariable();
+ const VarDecl *IVDecl = cast<VarDecl>(cast<DeclRefExpr>(IVExpr)->getDecl());
+ CGF.EmitVarDecl(*IVDecl);
+ CGF.EmitIgnoredExpr(S.getInit());
- // Emit the iterations count variable.
- // If it is not a variable, Sema decided to calculate iterations count on
- // each iteration (e.g., it is foldable into a constant).
- if (auto LIExpr = dyn_cast<DeclRefExpr>(S.getLastIteration())) {
- CGF.EmitVarDecl(*cast<VarDecl>(LIExpr->getDecl()));
- // Emit calculation of the iterations count.
- CGF.EmitIgnoredExpr(S.getCalcLastIteration());
- }
+ // Emit the iterations count variable.
+ // If it is not a variable, Sema decided to calculate iterations count on
+ // each iteration (e.g., it is foldable into a constant).
+ if (auto LIExpr = dyn_cast<DeclRefExpr>(S.getLastIteration())) {
+ CGF.EmitVarDecl(*cast<VarDecl>(LIExpr->getDecl()));
+ // Emit calculation of the iterations count.
+ CGF.EmitIgnoredExpr(S.getCalcLastIteration());
+ }
- CGF.EmitOMPSimdInit(S);
+ CGF.EmitOMPSimdInit(S);
- emitAlignedClause(CGF, S);
- (void)CGF.EmitOMPLinearClauseInit(S);
- {
- OMPPrivateScope LoopScope(CGF);
- CGF.EmitOMPPrivateLoopCounters(S, LoopScope);
- CGF.EmitOMPLinearClause(S, LoopScope);
- CGF.EmitOMPPrivateClause(S, LoopScope);
- CGF.EmitOMPReductionClauseInit(S, LoopScope);
- bool HasLastprivateClause =
- CGF.EmitOMPLastprivateClauseInit(S, LoopScope);
- (void)LoopScope.Privatize();
- CGF.EmitOMPInnerLoop(S, LoopScope.requiresCleanups(), S.getCond(),
- S.getInc(),
- [&S](CodeGenFunction &CGF) {
- CGF.EmitOMPLoopBody(S, JumpDest());
- CGF.EmitStopPoint(&S);
- },
- [](CodeGenFunction &) {});
- CGF.EmitOMPSimdFinal(
- S, [](CodeGenFunction &) -> llvm::Value * { return nullptr; });
- // Emit final copy of the lastprivate variables at the end of loops.
- if (HasLastprivateClause)
- CGF.EmitOMPLastprivateClauseFinal(S, /*NoFinals=*/true);
- CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_simd);
- emitPostUpdateForReductionClause(
- CGF, S, [](CodeGenFunction &) -> llvm::Value * { return nullptr; });
- }
- CGF.EmitOMPLinearClauseFinal(
+ emitAlignedClause(CGF, S);
+ (void)CGF.EmitOMPLinearClauseInit(S);
+ {
+ CodeGenFunction::OMPPrivateScope LoopScope(CGF);
+ CGF.EmitOMPPrivateLoopCounters(S, LoopScope);
+ CGF.EmitOMPLinearClause(S, LoopScope);
+ CGF.EmitOMPPrivateClause(S, LoopScope);
+ CGF.EmitOMPReductionClauseInit(S, LoopScope);
+ bool HasLastprivateClause = CGF.EmitOMPLastprivateClauseInit(S, LoopScope);
+ (void)LoopScope.Privatize();
+ CGF.EmitOMPInnerLoop(S, LoopScope.requiresCleanups(), S.getCond(),
+ S.getInc(),
+ [&S](CodeGenFunction &CGF) {
+ CGF.EmitOMPLoopBody(S, CodeGenFunction::JumpDest());
+ CGF.EmitStopPoint(&S);
+ },
+ [](CodeGenFunction &) {});
+ CGF.EmitOMPSimdFinal(
S, [](CodeGenFunction &) -> llvm::Value * { return nullptr; });
- // Emit: if (PreCond) - end.
- if (ContBlock) {
- CGF.EmitBranch(ContBlock);
- CGF.EmitBlock(ContBlock, true);
- }
+ // Emit final copy of the lastprivate variables at the end of loops.
+ if (HasLastprivateClause)
+ CGF.EmitOMPLastprivateClauseFinal(S, /*NoFinals=*/true);
+ CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_simd);
+ emitPostUpdateForReductionClause(
+ CGF, S, [](CodeGenFunction &) -> llvm::Value * { return nullptr; });
+ }
+ CGF.EmitOMPLinearClauseFinal(
+ S, [](CodeGenFunction &) -> llvm::Value * { return nullptr; });
+ // Emit: if (PreCond) - end.
+ if (ContBlock) {
+ CGF.EmitBranch(ContBlock);
+ CGF.EmitBlock(ContBlock, true);
+ }
+}
+
+void CodeGenFunction::EmitOMPSimdDirective(const OMPSimdDirective &S) {
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ emitOMPSimdRegion(CGF, S, Action);
};
OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_simd, CodeGen);
@@ -1669,7 +1758,8 @@ void CodeGenFunction::EmitOMPOuterLoop(
// Tell the runtime we are done.
auto &&CodeGen = [DynamicOrOrdered, &S](CodeGenFunction &CGF) {
if (!DynamicOrOrdered)
- CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd());
+ CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd(),
+ S.getDirectiveKind());
};
OMPCancelStack.emitExit(*this, S.getDirectiveKind(), CodeGen);
}
@@ -1753,9 +1843,11 @@ void CodeGenFunction::EmitOMPForOuterLoop(
RT.emitForDispatchInit(*this, S.getLocStart(), ScheduleKind, IVSize,
IVSigned, Ordered, DipatchRTInputValues);
} else {
- RT.emitForStaticInit(*this, S.getLocStart(), ScheduleKind, IVSize, IVSigned,
- Ordered, LoopArgs.IL, LoopArgs.LB, LoopArgs.UB,
- LoopArgs.ST, LoopArgs.Chunk);
+ CGOpenMPRuntime::StaticRTInput StaticInit(
+ IVSize, IVSigned, Ordered, LoopArgs.IL, LoopArgs.LB, LoopArgs.UB,
+ LoopArgs.ST, LoopArgs.Chunk);
+ RT.emitForStaticInit(*this, S.getLocStart(), S.getDirectiveKind(),
+ ScheduleKind, StaticInit);
}
auto &&CodeGenOrdered = [Ordered](CodeGenFunction &CGF, SourceLocation Loc,
@@ -1797,10 +1889,10 @@ void CodeGenFunction::EmitOMPDistributeOuterLoop(
const unsigned IVSize = getContext().getTypeSize(IVExpr->getType());
const bool IVSigned = IVExpr->getType()->hasSignedIntegerRepresentation();
- RT.emitDistributeStaticInit(*this, S.getLocStart(), ScheduleKind, IVSize,
- IVSigned, /* Ordered = */ false, LoopArgs.IL,
- LoopArgs.LB, LoopArgs.UB, LoopArgs.ST,
- LoopArgs.Chunk);
+ CGOpenMPRuntime::StaticRTInput StaticInit(
+ IVSize, IVSigned, /* Ordered = */ false, LoopArgs.IL, LoopArgs.LB,
+ LoopArgs.UB, LoopArgs.ST, LoopArgs.Chunk);
+ RT.emitDistributeStaticInit(*this, S.getLocStart(), ScheduleKind, StaticInit);
// for combined 'distribute' and 'for' the increment expression of distribute
// is store in DistInc. For 'distribute' alone, it is in Inc.
@@ -1929,13 +2021,27 @@ emitInnerParallelForWhenCombined(CodeGenFunction &CGF,
CodeGenFunction::JumpDest LoopExit) {
auto &&CGInlinedWorksharingLoop = [&S](CodeGenFunction &CGF,
PrePostActionTy &) {
+ bool HasCancel = false;
+ if (!isOpenMPSimdDirective(S.getDirectiveKind())) {
+ if (const auto *D = dyn_cast<OMPTeamsDistributeParallelForDirective>(&S))
+ HasCancel = D->hasCancel();
+ else if (const auto *D = dyn_cast<OMPDistributeParallelForDirective>(&S))
+ HasCancel = D->hasCancel();
+ else if (const auto *D =
+ dyn_cast<OMPTargetTeamsDistributeParallelForDirective>(&S))
+ HasCancel = D->hasCancel();
+ }
+ CodeGenFunction::OMPCancelStackRAII CancelRegion(CGF, S.getDirectiveKind(),
+ HasCancel);
CGF.EmitOMPWorksharingLoop(S, S.getPrevEnsureUpperBound(),
emitDistributeParallelForInnerBounds,
emitDistributeParallelForDispatchBounds);
};
emitCommonOMPParallelDirective(
- CGF, S, OMPD_for, CGInlinedWorksharingLoop,
+ CGF, S,
+ isOpenMPSimdDirective(S.getDirectiveKind()) ? OMPD_for_simd : OMPD_for,
+ CGInlinedWorksharingLoop,
emitDistributeParallelForDistributeInnerBoundParams);
}
@@ -1946,119 +2052,53 @@ void CodeGenFunction::EmitOMPDistributeParallelForDirective(
S.getDistInc());
};
OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
- OMPCancelStackRAII CancelRegion(*this, OMPD_distribute_parallel_for,
- /*HasCancel=*/false);
- CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_distribute, CodeGen,
- /*HasCancel=*/false);
+ CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_distribute, CodeGen);
}
void CodeGenFunction::EmitOMPDistributeParallelForSimdDirective(
const OMPDistributeParallelForSimdDirective &S) {
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+ CGF.EmitOMPDistributeLoop(S, emitInnerParallelForWhenCombined,
+ S.getDistInc());
+ };
OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
- CGM.getOpenMPRuntime().emitInlinedDirective(
- *this, OMPD_distribute_parallel_for_simd,
- [&S](CodeGenFunction &CGF, PrePostActionTy &) {
- OMPLoopScope PreInitScope(CGF, S);
- CGF.EmitStmt(
- cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
- });
+ CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_distribute, CodeGen);
}
void CodeGenFunction::EmitOMPDistributeSimdDirective(
const OMPDistributeSimdDirective &S) {
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+ CGF.EmitOMPDistributeLoop(S, emitOMPLoopBodyWithStopPoint, S.getInc());
+ };
OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
- CGM.getOpenMPRuntime().emitInlinedDirective(
- *this, OMPD_distribute_simd,
- [&S](CodeGenFunction &CGF, PrePostActionTy &) {
- OMPLoopScope PreInitScope(CGF, S);
- CGF.EmitStmt(
- cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
- });
+ CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_simd, CodeGen);
}
-void CodeGenFunction::EmitOMPTargetParallelForSimdDirective(
- const OMPTargetParallelForSimdDirective &S) {
- OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
- CGM.getOpenMPRuntime().emitInlinedDirective(
- *this, OMPD_target_parallel_for_simd,
- [&S](CodeGenFunction &CGF, PrePostActionTy &) {
- OMPLoopScope PreInitScope(CGF, S);
- CGF.EmitStmt(
- cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
- });
+void CodeGenFunction::EmitOMPTargetSimdDeviceFunction(
+ CodeGenModule &CGM, StringRef ParentName, const OMPTargetSimdDirective &S) {
+ // Emit SPMD target parallel for region as a standalone region.
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ emitOMPSimdRegion(CGF, S, Action);
+ };
+ llvm::Function *Fn;
+ llvm::Constant *Addr;
+ // Emit target region as a standalone region.
+ CGM.getOpenMPRuntime().emitTargetOutlinedFunction(
+ S, ParentName, Fn, Addr, /*IsOffloadEntry=*/true, CodeGen);
+ assert(Fn && Addr && "Target device function emission failed.");
}
void CodeGenFunction::EmitOMPTargetSimdDirective(
const OMPTargetSimdDirective &S) {
- OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
- CGM.getOpenMPRuntime().emitInlinedDirective(
- *this, OMPD_target_simd, [&S](CodeGenFunction &CGF, PrePostActionTy &) {
- OMPLoopScope PreInitScope(CGF, S);
- CGF.EmitStmt(
- cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
- });
-}
-
-void CodeGenFunction::EmitOMPTeamsDistributeDirective(
- const OMPTeamsDistributeDirective &S) {
- OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
- CGM.getOpenMPRuntime().emitInlinedDirective(
- *this, OMPD_teams_distribute,
- [&S](CodeGenFunction &CGF, PrePostActionTy &) {
- OMPLoopScope PreInitScope(CGF, S);
- CGF.EmitStmt(
- cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
- });
-}
-
-void CodeGenFunction::EmitOMPTeamsDistributeSimdDirective(
- const OMPTeamsDistributeSimdDirective &S) {
- OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
- CGM.getOpenMPRuntime().emitInlinedDirective(
- *this, OMPD_teams_distribute_simd,
- [&S](CodeGenFunction &CGF, PrePostActionTy &) {
- OMPLoopScope PreInitScope(CGF, S);
- CGF.EmitStmt(
- cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
- });
-}
-
-void CodeGenFunction::EmitOMPTeamsDistributeParallelForSimdDirective(
- const OMPTeamsDistributeParallelForSimdDirective &S) {
- OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
- CGM.getOpenMPRuntime().emitInlinedDirective(
- *this, OMPD_teams_distribute_parallel_for_simd,
- [&S](CodeGenFunction &CGF, PrePostActionTy &) {
- OMPLoopScope PreInitScope(CGF, S);
- CGF.EmitStmt(
- cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
- });
-}
-
-void CodeGenFunction::EmitOMPTeamsDistributeParallelForDirective(
- const OMPTeamsDistributeParallelForDirective &S) {
- OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
- CGM.getOpenMPRuntime().emitInlinedDirective(
- *this, OMPD_teams_distribute_parallel_for,
- [&S](CodeGenFunction &CGF, PrePostActionTy &) {
- OMPLoopScope PreInitScope(CGF, S);
- CGF.EmitStmt(
- cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
- });
-}
-
-void CodeGenFunction::EmitOMPTargetTeamsDistributeDirective(
- const OMPTargetTeamsDistributeDirective &S) {
- CGM.getOpenMPRuntime().emitInlinedDirective(
- *this, OMPD_target_teams_distribute,
- [&S](CodeGenFunction &CGF, PrePostActionTy &) {
- CGF.EmitStmt(
- cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
- });
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ emitOMPSimdRegion(CGF, S, Action);
+ };
+ emitCommonOMPTargetDirective(*this, S, CodeGen);
}
void CodeGenFunction::EmitOMPTargetTeamsDistributeParallelForDirective(
const OMPTargetTeamsDistributeParallelForDirective &S) {
+ OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
CGM.getOpenMPRuntime().emitInlinedDirective(
*this, OMPD_target_teams_distribute_parallel_for,
[&S](CodeGenFunction &CGF, PrePostActionTy &) {
@@ -2069,6 +2109,7 @@ void CodeGenFunction::EmitOMPTargetTeamsDistributeParallelForDirective(
void CodeGenFunction::EmitOMPTargetTeamsDistributeParallelForSimdDirective(
const OMPTargetTeamsDistributeParallelForSimdDirective &S) {
+ OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
CGM.getOpenMPRuntime().emitInlinedDirective(
*this, OMPD_target_teams_distribute_parallel_for_simd,
[&S](CodeGenFunction &CGF, PrePostActionTy &) {
@@ -2077,16 +2118,6 @@ void CodeGenFunction::EmitOMPTargetTeamsDistributeParallelForSimdDirective(
});
}
-void CodeGenFunction::EmitOMPTargetTeamsDistributeSimdDirective(
- const OMPTargetTeamsDistributeSimdDirective &S) {
- CGM.getOpenMPRuntime().emitInlinedDirective(
- *this, OMPD_target_teams_distribute_simd,
- [&S](CodeGenFunction &CGF, PrePostActionTy &) {
- CGF.EmitStmt(
- cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
- });
-}
-
namespace {
struct ScheduleKindModifiersTy {
OpenMPScheduleClauseKind Kind;
@@ -2209,10 +2240,11 @@ bool CodeGenFunction::EmitOMPWorksharingLoop(
// chunks that are approximately equal in size, and at most one chunk is
// distributed to each thread. Note that the size of the chunks is
// unspecified in this case.
- RT.emitForStaticInit(*this, S.getLocStart(), ScheduleKind,
- IVSize, IVSigned, Ordered,
- IL.getAddress(), LB.getAddress(),
- UB.getAddress(), ST.getAddress());
+ CGOpenMPRuntime::StaticRTInput StaticInit(
+ IVSize, IVSigned, Ordered, IL.getAddress(), LB.getAddress(),
+ UB.getAddress(), ST.getAddress());
+ RT.emitForStaticInit(*this, S.getLocStart(), S.getDirectiveKind(),
+ ScheduleKind, StaticInit);
auto LoopExit =
getJumpDestInCurrentScope(createBasicBlock("omp.loop.exit"));
// UB = min(UB, GlobalUB);
@@ -2230,7 +2262,8 @@ bool CodeGenFunction::EmitOMPWorksharingLoop(
EmitBlock(LoopExit.getBlock());
// Tell the runtime we are done.
auto &&CodeGen = [&S](CodeGenFunction &CGF) {
- CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd());
+ CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd(),
+ S.getDirectiveKind());
};
OMPCancelStack.emitExit(*this, S.getDirectiveKind(), CodeGen);
} else {
@@ -2444,10 +2477,11 @@ void CodeGenFunction::EmitSections(const OMPExecutableDirective &S) {
// Emit static non-chunked loop.
OpenMPScheduleTy ScheduleKind;
ScheduleKind.Schedule = OMPC_SCHEDULE_static;
+ CGOpenMPRuntime::StaticRTInput StaticInit(
+ /*IVSize=*/32, /*IVSigned=*/true, /*Ordered=*/false, IL.getAddress(),
+ LB.getAddress(), UB.getAddress(), ST.getAddress());
CGF.CGM.getOpenMPRuntime().emitForStaticInit(
- CGF, S.getLocStart(), ScheduleKind, /*IVSize=*/32,
- /*IVSigned=*/true, /*Ordered=*/false, IL.getAddress(), LB.getAddress(),
- UB.getAddress(), ST.getAddress());
+ CGF, S.getLocStart(), S.getDirectiveKind(), ScheduleKind, StaticInit);
// UB = min(UB, GlobalUB);
auto *UBVal = CGF.EmitLoadOfScalar(UB, S.getLocStart());
auto *MinUBGlobalUB = CGF.Builder.CreateSelect(
@@ -2460,7 +2494,8 @@ void CodeGenFunction::EmitSections(const OMPExecutableDirective &S) {
[](CodeGenFunction &) {});
// Tell the runtime we are done.
auto &&CodeGen = [&S](CodeGenFunction &CGF) {
- CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd());
+ CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd(),
+ S.getDirectiveKind());
};
CGF.OMPCancelStack.emitExit(CGF, S.getDirectiveKind(), CodeGen);
CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_parallel);
@@ -2731,6 +2766,7 @@ void CodeGenFunction::EmitOMPTaskBasedDirective(const OMPExecutableDirective &S,
OMPPrivateScope Scope(CGF);
if (!Data.PrivateVars.empty() || !Data.FirstprivateVars.empty() ||
!Data.LastprivateVars.empty()) {
+ enum { PrivatesParam = 2, CopyFnParam = 3 };
auto *CopyFn = CGF.Builder.CreateLoad(
CGF.GetAddrOfLocalVar(CS->getCapturedDecl()->getParam(3)));
auto *PrivatesPtr = CGF.Builder.CreateLoad(
@@ -2762,7 +2798,8 @@ void CodeGenFunction::EmitOMPTaskBasedDirective(const OMPExecutableDirective &S,
PrivatePtrs.push_back(std::make_pair(VD, PrivatePtr));
CallArgs.push_back(PrivatePtr.getPointer());
}
- CGF.EmitRuntimeCall(CopyFn, CallArgs);
+ CGF.CGM.getOpenMPRuntime().emitOutlinedFunctionCall(CGF, S.getLocStart(),
+ CopyFn, CallArgs);
for (auto &&Pair : LastprivateDstsOrigs) {
auto *OrigVD = cast<VarDecl>(Pair.second->getDecl());
DeclRefExpr DRE(
@@ -2808,7 +2845,57 @@ void CodeGenFunction::EmitOMPTaskBasedDirective(const OMPExecutableDirective &S,
RedCG, Cnt);
}
}
+ // Privatize all private variables except for in_reduction items.
(void)Scope.Privatize();
+ SmallVector<const Expr *, 4> InRedVars;
+ SmallVector<const Expr *, 4> InRedPrivs;
+ SmallVector<const Expr *, 4> InRedOps;
+ SmallVector<const Expr *, 4> TaskgroupDescriptors;
+ for (const auto *C : S.getClausesOfKind<OMPInReductionClause>()) {
+ auto IPriv = C->privates().begin();
+ auto IRed = C->reduction_ops().begin();
+ auto ITD = C->taskgroup_descriptors().begin();
+ for (const auto *Ref : C->varlists()) {
+ InRedVars.emplace_back(Ref);
+ InRedPrivs.emplace_back(*IPriv);
+ InRedOps.emplace_back(*IRed);
+ TaskgroupDescriptors.emplace_back(*ITD);
+ std::advance(IPriv, 1);
+ std::advance(IRed, 1);
+ std::advance(ITD, 1);
+ }
+ }
+ // Privatize in_reduction items here, because taskgroup descriptors must be
+ // privatized earlier.
+ OMPPrivateScope InRedScope(CGF);
+ if (!InRedVars.empty()) {
+ ReductionCodeGen RedCG(InRedVars, InRedPrivs, InRedOps);
+ for (unsigned Cnt = 0, E = InRedVars.size(); Cnt < E; ++Cnt) {
+ RedCG.emitSharedLValue(CGF, Cnt);
+ RedCG.emitAggregateType(CGF, Cnt);
+ // The taskgroup descriptor variable is always implicit firstprivate and
+ // privatized already during procoessing of the firstprivates.
+ llvm::Value *ReductionsPtr = CGF.EmitLoadOfScalar(
+ CGF.EmitLValue(TaskgroupDescriptors[Cnt]), SourceLocation());
+ Address Replacement = CGF.CGM.getOpenMPRuntime().getTaskReductionItem(
+ CGF, S.getLocStart(), ReductionsPtr, RedCG.getSharedLValue(Cnt));
+ Replacement = Address(
+ CGF.EmitScalarConversion(
+ Replacement.getPointer(), CGF.getContext().VoidPtrTy,
+ CGF.getContext().getPointerType(InRedPrivs[Cnt]->getType()),
+ SourceLocation()),
+ Replacement.getAlignment());
+ Replacement = RedCG.adjustPrivateAddress(CGF, Cnt, Replacement);
+ InRedScope.addPrivate(RedCG.getBaseDecl(Cnt),
+ [Replacement]() { return Replacement; });
+ // FIXME: This must removed once the runtime library is fixed.
+ // Emit required threadprivate variables for
+ // initilizer/combiner/finalizer.
+ CGF.CGM.getOpenMPRuntime().emitTaskReductionFixups(CGF, S.getLocStart(),
+ RedCG, Cnt);
+ }
+ }
+ (void)InRedScope.Privatize();
Action.Enter(CGF);
BodyGen(CGF);
@@ -2867,6 +2954,35 @@ void CodeGenFunction::EmitOMPTaskgroupDirective(
const OMPTaskgroupDirective &S) {
auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
Action.Enter(CGF);
+ if (const Expr *E = S.getReductionRef()) {
+ SmallVector<const Expr *, 4> LHSs;
+ SmallVector<const Expr *, 4> RHSs;
+ OMPTaskDataTy Data;
+ for (const auto *C : S.getClausesOfKind<OMPTaskReductionClause>()) {
+ auto IPriv = C->privates().begin();
+ auto IRed = C->reduction_ops().begin();
+ auto ILHS = C->lhs_exprs().begin();
+ auto IRHS = C->rhs_exprs().begin();
+ for (const auto *Ref : C->varlists()) {
+ Data.ReductionVars.emplace_back(Ref);
+ Data.ReductionCopies.emplace_back(*IPriv);
+ Data.ReductionOps.emplace_back(*IRed);
+ LHSs.emplace_back(*ILHS);
+ RHSs.emplace_back(*IRHS);
+ std::advance(IPriv, 1);
+ std::advance(IRed, 1);
+ std::advance(ILHS, 1);
+ std::advance(IRHS, 1);
+ }
+ }
+ llvm::Value *ReductionDesc =
+ CGF.CGM.getOpenMPRuntime().emitTaskReductionInit(CGF, S.getLocStart(),
+ LHSs, RHSs, Data);
+ const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
+ CGF.EmitVarDecl(*VD);
+ CGF.EmitStoreOfScalar(ReductionDesc, CGF.GetAddrOfLocalVar(VD),
+ /*Volatile=*/false, E->getType());
+ }
CGF.EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
};
OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
@@ -2923,6 +3039,7 @@ void CodeGenFunction::EmitOMPDistributeLoop(const OMPLoopDirective &S,
incrementProfileCounter(&S);
}
+ emitAlignedClause(*this, S);
// Emit 'then' code.
{
// Emit helper vars inits.
@@ -2944,14 +3061,18 @@ void CodeGenFunction::EmitOMPDistributeLoop(const OMPLoopDirective &S,
OMPPrivateScope LoopScope(*this);
if (EmitOMPFirstprivateClause(S, LoopScope)) {
- // Emit implicit barrier to synchronize threads and avoid data races on
- // initialization of firstprivate variables and post-update of
+ // Emit implicit barrier to synchronize threads and avoid data races
+ // on initialization of firstprivate variables and post-update of
// lastprivate variables.
CGM.getOpenMPRuntime().emitBarrierCall(
- *this, S.getLocStart(), OMPD_unknown, /*EmitChecks=*/false,
- /*ForceSimpleCall=*/true);
+ *this, S.getLocStart(), OMPD_unknown, /*EmitChecks=*/false,
+ /*ForceSimpleCall=*/true);
}
EmitOMPPrivateClause(S, LoopScope);
+ if (isOpenMPSimdDirective(S.getDirectiveKind()) &&
+ !isOpenMPParallelDirective(S.getDirectiveKind()) &&
+ !isOpenMPTeamsDirective(S.getDirectiveKind()))
+ EmitOMPReductionClauseInit(S, LoopScope);
HasLastprivateClause = EmitOMPLastprivateClauseInit(S, LoopScope);
EmitOMPPrivateLoopCounters(S, LoopScope);
(void)LoopScope.Privatize();
@@ -2964,8 +3085,8 @@ void CodeGenFunction::EmitOMPDistributeLoop(const OMPLoopDirective &S,
if (const auto *Ch = C->getChunkSize()) {
Chunk = EmitScalarExpr(Ch);
Chunk = EmitScalarConversion(Chunk, Ch->getType(),
- S.getIterationVariable()->getType(),
- S.getLocStart());
+ S.getIterationVariable()->getType(),
+ S.getLocStart());
}
}
const unsigned IVSize = getContext().getTypeSize(IVExpr->getType());
@@ -2981,10 +3102,13 @@ void CodeGenFunction::EmitOMPDistributeLoop(const OMPLoopDirective &S,
// league. The size of the chunks is unspecified in this case.
if (RT.isStaticNonchunked(ScheduleKind,
/* Chunked */ Chunk != nullptr)) {
+ if (isOpenMPSimdDirective(S.getDirectiveKind()))
+ EmitOMPSimdInit(S, /*IsMonotonic=*/true);
+ CGOpenMPRuntime::StaticRTInput StaticInit(
+ IVSize, IVSigned, /* Ordered = */ false, IL.getAddress(),
+ LB.getAddress(), UB.getAddress(), ST.getAddress());
RT.emitDistributeStaticInit(*this, S.getLocStart(), ScheduleKind,
- IVSize, IVSigned, /* Ordered = */ false,
- IL.getAddress(), LB.getAddress(),
- UB.getAddress(), ST.getAddress());
+ StaticInit);
auto LoopExit =
getJumpDestInCurrentScope(createBasicBlock("omp.loop.exit"));
// UB = min(UB, GlobalUB);
@@ -3011,7 +3135,7 @@ void CodeGenFunction::EmitOMPDistributeLoop(const OMPLoopDirective &S,
[](CodeGenFunction &) {});
EmitBlock(LoopExit.getBlock());
// Tell the runtime we are done.
- RT.emitForStaticFinish(*this, S.getLocStart());
+ RT.emitForStaticFinish(*this, S.getLocStart(), S.getDirectiveKind());
} else {
// Emit the outer loop, which requests its work chunk [LB..UB] from
// runtime and runs the inner loop to process it.
@@ -3021,13 +3145,38 @@ void CodeGenFunction::EmitOMPDistributeLoop(const OMPLoopDirective &S,
EmitOMPDistributeOuterLoop(ScheduleKind, S, LoopScope, LoopArguments,
CodeGenLoop);
}
-
+ if (isOpenMPSimdDirective(S.getDirectiveKind())) {
+ EmitOMPSimdFinal(S, [&](CodeGenFunction &CGF) -> llvm::Value * {
+ return CGF.Builder.CreateIsNotNull(
+ CGF.EmitLoadOfScalar(IL, S.getLocStart()));
+ });
+ }
+ OpenMPDirectiveKind ReductionKind = OMPD_unknown;
+ if (isOpenMPParallelDirective(S.getDirectiveKind()) &&
+ isOpenMPSimdDirective(S.getDirectiveKind())) {
+ ReductionKind = OMPD_parallel_for_simd;
+ } else if (isOpenMPParallelDirective(S.getDirectiveKind())) {
+ ReductionKind = OMPD_parallel_for;
+ } else if (isOpenMPSimdDirective(S.getDirectiveKind())) {
+ ReductionKind = OMPD_simd;
+ } else if (!isOpenMPTeamsDirective(S.getDirectiveKind()) &&
+ S.hasClausesOfKind<OMPReductionClause>()) {
+ llvm_unreachable(
+ "No reduction clauses is allowed in distribute directive.");
+ }
+ EmitOMPReductionClauseFinal(S, ReductionKind);
+ // Emit post-update of the reduction variables if IsLastIter != 0.
+ emitPostUpdateForReductionClause(
+ *this, S, [&](CodeGenFunction &CGF) -> llvm::Value * {
+ return CGF.Builder.CreateIsNotNull(
+ CGF.EmitLoadOfScalar(IL, S.getLocStart()));
+ });
// Emit final copy of the lastprivate variables if IsLastIter != 0.
- if (HasLastprivateClause)
+ if (HasLastprivateClause) {
EmitOMPLastprivateClauseFinal(
S, /*NoFinals=*/false,
- Builder.CreateIsNotNull(
- EmitLoadOfScalar(IL, S.getLocStart())));
+ Builder.CreateIsNotNull(EmitLoadOfScalar(IL, S.getLocStart())));
+ }
}
// We're now done with the loop, so jump to the continuation block.
@@ -3045,8 +3194,7 @@ void CodeGenFunction::EmitOMPDistributeDirective(
CGF.EmitOMPDistributeLoop(S, emitOMPLoopBodyWithStopPoint, S.getInc());
};
OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
- CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_distribute, CodeGen,
- false);
+ CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_distribute, CodeGen);
}
static llvm::Function *emitOutlinedOrderedFunction(CodeGenModule &CGM,
@@ -3073,7 +3221,8 @@ void CodeGenFunction::EmitOMPOrderedDirective(const OMPOrderedDirective &S) {
llvm::SmallVector<llvm::Value *, 16> CapturedVars;
CGF.GenerateOpenMPCapturedVars(*CS, CapturedVars);
auto *OutlinedFn = emitOutlinedOrderedFunction(CGM, CS);
- CGF.EmitNounwindRuntimeCall(OutlinedFn, CapturedVars);
+ CGM.getOpenMPRuntime().emitOutlinedFunctionCall(CGF, S.getLocStart(),
+ OutlinedFn, CapturedVars);
} else {
Action.Enter(CGF);
CGF.EmitStmt(
@@ -3259,6 +3408,7 @@ static std::pair<bool, RValue> emitOMPAtomicRMW(CodeGenFunction &CGF, LValue X,
case BO_GE:
case BO_EQ:
case BO_NE:
+ case BO_Cmp:
case BO_AddAssign:
case BO_SubAssign:
case BO_AndAssign:
@@ -3470,6 +3620,7 @@ static void EmitOMPAtomicExpr(CodeGenFunction &CGF, OpenMPClauseKind Kind,
case OMPC_lastprivate:
case OMPC_reduction:
case OMPC_task_reduction:
+ case OMPC_in_reduction:
case OMPC_safelen:
case OMPC_simdlen:
case OMPC_collapse:
@@ -3552,7 +3703,7 @@ static void emitCommonOMPTargetDirective(CodeGenFunction &CGF,
const RegionCodeGenTy &CodeGen) {
assert(isOpenMPTargetExecutionDirective(S.getDirectiveKind()));
CodeGenModule &CGM = CGF.CGM;
- const CapturedStmt &CS = *cast<CapturedStmt>(S.getAssociatedStmt());
+ const CapturedStmt &CS = *S.getCapturedStmt(OMPD_target);
llvm::Function *Fn = nullptr;
llvm::Constant *FnID = nullptr;
@@ -3675,7 +3826,7 @@ void CodeGenFunction::EmitOMPTeamsDirective(const OMPTeamsDirective &S) {
CGF.EmitStmt(cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt());
CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_teams);
};
- emitCommonOMPTeamsDirective(*this, S, OMPD_teams, CodeGen);
+ emitCommonOMPTeamsDirective(*this, S, OMPD_distribute, CodeGen);
emitPostUpdateForReductionClause(
*this, S, [](CodeGenFunction &) -> llvm::Value * { return nullptr; });
}
@@ -3684,11 +3835,20 @@ static void emitTargetTeamsRegion(CodeGenFunction &CGF, PrePostActionTy &Action,
const OMPTargetTeamsDirective &S) {
auto *CS = S.getCapturedStmt(OMPD_teams);
Action.Enter(CGF);
- auto &&CodeGen = [CS](CodeGenFunction &CGF, PrePostActionTy &) {
- // TODO: Add support for clauses.
+ // Emit teams region as a standalone region.
+ auto &&CodeGen = [&S, CS](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ CodeGenFunction::OMPPrivateScope PrivateScope(CGF);
+ (void)CGF.EmitOMPFirstprivateClause(S, PrivateScope);
+ CGF.EmitOMPPrivateClause(S, PrivateScope);
+ CGF.EmitOMPReductionClauseInit(S, PrivateScope);
+ (void)PrivateScope.Privatize();
+ Action.Enter(CGF);
CGF.EmitStmt(CS->getCapturedStmt());
+ CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_teams);
};
emitCommonOMPTeamsDirective(CGF, S, OMPD_teams, CodeGen);
+ emitPostUpdateForReductionClause(
+ CGF, S, [](CodeGenFunction &) -> llvm::Value * { return nullptr; });
}
void CodeGenFunction::EmitOMPTargetTeamsDeviceFunction(
@@ -3713,6 +3873,183 @@ void CodeGenFunction::EmitOMPTargetTeamsDirective(
emitCommonOMPTargetDirective(*this, S, CodeGen);
}
+static void
+emitTargetTeamsDistributeRegion(CodeGenFunction &CGF, PrePostActionTy &Action,
+ const OMPTargetTeamsDistributeDirective &S) {
+ Action.Enter(CGF);
+ auto &&CodeGenDistribute = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+ CGF.EmitOMPDistributeLoop(S, emitOMPLoopBodyWithStopPoint, S.getInc());
+ };
+
+ // Emit teams region as a standalone region.
+ auto &&CodeGen = [&S, &CodeGenDistribute](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ CodeGenFunction::OMPPrivateScope PrivateScope(CGF);
+ CGF.EmitOMPReductionClauseInit(S, PrivateScope);
+ (void)PrivateScope.Privatize();
+ CGF.CGM.getOpenMPRuntime().emitInlinedDirective(CGF, OMPD_distribute,
+ CodeGenDistribute);
+ CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_teams);
+ };
+ emitCommonOMPTeamsDirective(CGF, S, OMPD_distribute, CodeGen);
+ emitPostUpdateForReductionClause(CGF, S,
+ [](CodeGenFunction &) { return nullptr; });
+}
+
+void CodeGenFunction::EmitOMPTargetTeamsDistributeDeviceFunction(
+ CodeGenModule &CGM, StringRef ParentName,
+ const OMPTargetTeamsDistributeDirective &S) {
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ emitTargetTeamsDistributeRegion(CGF, Action, S);
+ };
+ llvm::Function *Fn;
+ llvm::Constant *Addr;
+ // Emit target region as a standalone region.
+ CGM.getOpenMPRuntime().emitTargetOutlinedFunction(
+ S, ParentName, Fn, Addr, /*IsOffloadEntry=*/true, CodeGen);
+ assert(Fn && Addr && "Target device function emission failed.");
+}
+
+void CodeGenFunction::EmitOMPTargetTeamsDistributeDirective(
+ const OMPTargetTeamsDistributeDirective &S) {
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ emitTargetTeamsDistributeRegion(CGF, Action, S);
+ };
+ emitCommonOMPTargetDirective(*this, S, CodeGen);
+}
+
+static void emitTargetTeamsDistributeSimdRegion(
+ CodeGenFunction &CGF, PrePostActionTy &Action,
+ const OMPTargetTeamsDistributeSimdDirective &S) {
+ Action.Enter(CGF);
+ auto &&CodeGenDistribute = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+ CGF.EmitOMPDistributeLoop(S, emitOMPLoopBodyWithStopPoint, S.getInc());
+ };
+
+ // Emit teams region as a standalone region.
+ auto &&CodeGen = [&S, &CodeGenDistribute](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ CodeGenFunction::OMPPrivateScope PrivateScope(CGF);
+ CGF.EmitOMPReductionClauseInit(S, PrivateScope);
+ (void)PrivateScope.Privatize();
+ CGF.CGM.getOpenMPRuntime().emitInlinedDirective(CGF, OMPD_distribute,
+ CodeGenDistribute);
+ CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_teams);
+ };
+ emitCommonOMPTeamsDirective(CGF, S, OMPD_distribute_simd, CodeGen);
+ emitPostUpdateForReductionClause(CGF, S,
+ [](CodeGenFunction &) { return nullptr; });
+}
+
+void CodeGenFunction::EmitOMPTargetTeamsDistributeSimdDeviceFunction(
+ CodeGenModule &CGM, StringRef ParentName,
+ const OMPTargetTeamsDistributeSimdDirective &S) {
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ emitTargetTeamsDistributeSimdRegion(CGF, Action, S);
+ };
+ llvm::Function *Fn;
+ llvm::Constant *Addr;
+ // Emit target region as a standalone region.
+ CGM.getOpenMPRuntime().emitTargetOutlinedFunction(
+ S, ParentName, Fn, Addr, /*IsOffloadEntry=*/true, CodeGen);
+ assert(Fn && Addr && "Target device function emission failed.");
+}
+
+void CodeGenFunction::EmitOMPTargetTeamsDistributeSimdDirective(
+ const OMPTargetTeamsDistributeSimdDirective &S) {
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ emitTargetTeamsDistributeSimdRegion(CGF, Action, S);
+ };
+ emitCommonOMPTargetDirective(*this, S, CodeGen);
+}
+
+void CodeGenFunction::EmitOMPTeamsDistributeDirective(
+ const OMPTeamsDistributeDirective &S) {
+
+ auto &&CodeGenDistribute = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+ CGF.EmitOMPDistributeLoop(S, emitOMPLoopBodyWithStopPoint, S.getInc());
+ };
+
+ // Emit teams region as a standalone region.
+ auto &&CodeGen = [&S, &CodeGenDistribute](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ OMPPrivateScope PrivateScope(CGF);
+ CGF.EmitOMPReductionClauseInit(S, PrivateScope);
+ (void)PrivateScope.Privatize();
+ CGF.CGM.getOpenMPRuntime().emitInlinedDirective(CGF, OMPD_distribute,
+ CodeGenDistribute);
+ CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_teams);
+ };
+ emitCommonOMPTeamsDirective(*this, S, OMPD_distribute, CodeGen);
+ emitPostUpdateForReductionClause(*this, S,
+ [](CodeGenFunction &) { return nullptr; });
+}
+
+void CodeGenFunction::EmitOMPTeamsDistributeSimdDirective(
+ const OMPTeamsDistributeSimdDirective &S) {
+ auto &&CodeGenDistribute = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+ CGF.EmitOMPDistributeLoop(S, emitOMPLoopBodyWithStopPoint, S.getInc());
+ };
+
+ // Emit teams region as a standalone region.
+ auto &&CodeGen = [&S, &CodeGenDistribute](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ OMPPrivateScope PrivateScope(CGF);
+ CGF.EmitOMPReductionClauseInit(S, PrivateScope);
+ (void)PrivateScope.Privatize();
+ CGF.CGM.getOpenMPRuntime().emitInlinedDirective(CGF, OMPD_simd,
+ CodeGenDistribute);
+ CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_teams);
+ };
+ emitCommonOMPTeamsDirective(*this, S, OMPD_distribute_simd, CodeGen);
+ emitPostUpdateForReductionClause(*this, S,
+ [](CodeGenFunction &) { return nullptr; });
+}
+
+void CodeGenFunction::EmitOMPTeamsDistributeParallelForDirective(
+ const OMPTeamsDistributeParallelForDirective &S) {
+ auto &&CodeGenDistribute = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+ CGF.EmitOMPDistributeLoop(S, emitInnerParallelForWhenCombined,
+ S.getDistInc());
+ };
+
+ // Emit teams region as a standalone region.
+ auto &&CodeGen = [&S, &CodeGenDistribute](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ OMPPrivateScope PrivateScope(CGF);
+ CGF.EmitOMPReductionClauseInit(S, PrivateScope);
+ (void)PrivateScope.Privatize();
+ CGF.CGM.getOpenMPRuntime().emitInlinedDirective(CGF, OMPD_distribute,
+ CodeGenDistribute);
+ CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_teams);
+ };
+ emitCommonOMPTeamsDirective(*this, S, OMPD_distribute_parallel_for, CodeGen);
+ emitPostUpdateForReductionClause(*this, S,
+ [](CodeGenFunction &) { return nullptr; });
+}
+
+void CodeGenFunction::EmitOMPTeamsDistributeParallelForSimdDirective(
+ const OMPTeamsDistributeParallelForSimdDirective &S) {
+ auto &&CodeGenDistribute = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+ CGF.EmitOMPDistributeLoop(S, emitInnerParallelForWhenCombined,
+ S.getDistInc());
+ };
+
+ // Emit teams region as a standalone region.
+ auto &&CodeGen = [&S, &CodeGenDistribute](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ OMPPrivateScope PrivateScope(CGF);
+ CGF.EmitOMPReductionClauseInit(S, PrivateScope);
+ (void)PrivateScope.Privatize();
+ CGF.CGM.getOpenMPRuntime().emitInlinedDirective(
+ CGF, OMPD_distribute, CodeGenDistribute, /*HasCancel=*/false);
+ CGF.EmitOMPReductionClauseFinal(S, /*ReductionKind=*/OMPD_teams);
+ };
+ emitCommonOMPTeamsDirective(*this, S, OMPD_distribute_parallel_for, CodeGen);
+ emitPostUpdateForReductionClause(*this, S,
+ [](CodeGenFunction &) { return nullptr; });
+}
+
void CodeGenFunction::EmitOMPCancellationPointDirective(
const OMPCancellationPointDirective &S) {
CGM.getOpenMPRuntime().emitCancellationPointCall(*this, S.getLocStart(),
@@ -3740,7 +4077,9 @@ CodeGenFunction::getOMPCancelDestination(OpenMPDirectiveKind Kind) {
assert(Kind == OMPD_for || Kind == OMPD_section || Kind == OMPD_sections ||
Kind == OMPD_parallel_sections || Kind == OMPD_parallel_for ||
Kind == OMPD_distribute_parallel_for ||
- Kind == OMPD_target_parallel_for);
+ Kind == OMPD_target_parallel_for ||
+ Kind == OMPD_teams_distribute_parallel_for ||
+ Kind == OMPD_target_teams_distribute_parallel_for);
return OMPCancelStack.getExitBlock();
}
@@ -3913,7 +4252,14 @@ void CodeGenFunction::EmitOMPTargetEnterDataDirective(
if (auto *C = S.getSingleClause<OMPDeviceClause>())
Device = C->getDevice();
- CGM.getOpenMPRuntime().emitTargetDataStandAloneCall(*this, S, IfCond, Device);
+ auto &&CodeGen = [&S, IfCond, Device](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ CGF.CGM.getOpenMPRuntime().emitTargetDataStandAloneCall(CGF, S, IfCond,
+ Device);
+ };
+ OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
+ CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_target_enter_data,
+ CodeGen);
}
void CodeGenFunction::EmitOMPTargetExitDataDirective(
@@ -3933,7 +4279,14 @@ void CodeGenFunction::EmitOMPTargetExitDataDirective(
if (auto *C = S.getSingleClause<OMPDeviceClause>())
Device = C->getDevice();
- CGM.getOpenMPRuntime().emitTargetDataStandAloneCall(*this, S, IfCond, Device);
+ auto &&CodeGen = [&S, IfCond, Device](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ CGF.CGM.getOpenMPRuntime().emitTargetDataStandAloneCall(CGF, S, IfCond,
+ Device);
+ };
+ OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
+ CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_target_exit_data,
+ CodeGen);
}
static void emitTargetParallelRegion(CodeGenFunction &CGF,
@@ -3980,9 +4333,81 @@ void CodeGenFunction::EmitOMPTargetParallelDirective(
emitCommonOMPTargetDirective(*this, S, CodeGen);
}
+static void emitTargetParallelForRegion(CodeGenFunction &CGF,
+ const OMPTargetParallelForDirective &S,
+ PrePostActionTy &Action) {
+ Action.Enter(CGF);
+ // Emit directive as a combined directive that consists of two implicit
+ // directives: 'parallel' with 'for' directive.
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+ CodeGenFunction::OMPCancelStackRAII CancelRegion(
+ CGF, OMPD_target_parallel_for, S.hasCancel());
+ CGF.EmitOMPWorksharingLoop(S, S.getEnsureUpperBound(), emitForLoopBounds,
+ emitDispatchForLoopBounds);
+ };
+ emitCommonOMPParallelDirective(CGF, S, OMPD_for, CodeGen,
+ emitEmptyBoundParameters);
+}
+
+void CodeGenFunction::EmitOMPTargetParallelForDeviceFunction(
+ CodeGenModule &CGM, StringRef ParentName,
+ const OMPTargetParallelForDirective &S) {
+ // Emit SPMD target parallel for region as a standalone region.
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ emitTargetParallelForRegion(CGF, S, Action);
+ };
+ llvm::Function *Fn;
+ llvm::Constant *Addr;
+ // Emit target region as a standalone region.
+ CGM.getOpenMPRuntime().emitTargetOutlinedFunction(
+ S, ParentName, Fn, Addr, /*IsOffloadEntry=*/true, CodeGen);
+ assert(Fn && Addr && "Target device function emission failed.");
+}
+
void CodeGenFunction::EmitOMPTargetParallelForDirective(
const OMPTargetParallelForDirective &S) {
- // TODO: codegen for target parallel for.
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ emitTargetParallelForRegion(CGF, S, Action);
+ };
+ emitCommonOMPTargetDirective(*this, S, CodeGen);
+}
+
+static void
+emitTargetParallelForSimdRegion(CodeGenFunction &CGF,
+ const OMPTargetParallelForSimdDirective &S,
+ PrePostActionTy &Action) {
+ Action.Enter(CGF);
+ // Emit directive as a combined directive that consists of two implicit
+ // directives: 'parallel' with 'for' directive.
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &) {
+ CGF.EmitOMPWorksharingLoop(S, S.getEnsureUpperBound(), emitForLoopBounds,
+ emitDispatchForLoopBounds);
+ };
+ emitCommonOMPParallelDirective(CGF, S, OMPD_simd, CodeGen,
+ emitEmptyBoundParameters);
+}
+
+void CodeGenFunction::EmitOMPTargetParallelForSimdDeviceFunction(
+ CodeGenModule &CGM, StringRef ParentName,
+ const OMPTargetParallelForSimdDirective &S) {
+ // Emit SPMD target parallel for region as a standalone region.
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ emitTargetParallelForSimdRegion(CGF, S, Action);
+ };
+ llvm::Function *Fn;
+ llvm::Constant *Addr;
+ // Emit target region as a standalone region.
+ CGM.getOpenMPRuntime().emitTargetOutlinedFunction(
+ S, ParentName, Fn, Addr, /*IsOffloadEntry=*/true, CodeGen);
+ assert(Fn && Addr && "Target device function emission failed.");
+}
+
+void CodeGenFunction::EmitOMPTargetParallelForSimdDirective(
+ const OMPTargetParallelForSimdDirective &S) {
+ auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &Action) {
+ emitTargetParallelForSimdRegion(CGF, S, Action);
+ };
+ emitCommonOMPTargetDirective(*this, S, CodeGen);
}
/// Emit a helper variable and return corresponding lvalue.
@@ -4160,5 +4585,12 @@ void CodeGenFunction::EmitOMPTargetUpdateDirective(
if (auto *C = S.getSingleClause<OMPDeviceClause>())
Device = C->getDevice();
- CGM.getOpenMPRuntime().emitTargetDataStandAloneCall(*this, S, IfCond, Device);
+ auto &&CodeGen = [&S, IfCond, Device](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ CGF.CGM.getOpenMPRuntime().emitTargetDataStandAloneCall(CGF, S, IfCond,
+ Device);
+ };
+ OMPLexicalScope Scope(*this, S, /*AsInlined=*/true);
+ CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_target_update,
+ CodeGen);
}
generated by cgit v1.2.3 (git 2.25.1) at 2025-06-27 21:31:27 +0000