summaryrefslogtreecommitdiffstats
path: root/clang/lib
diff options
context:
space:
mode:
authorAlexey Bataev <a.bataev@hotmail.com>2018-01-15 19:06:12 +0000
committerAlexey Bataev <a.bataev@hotmail.com>2018-01-15 19:06:12 +0000
commit8451efad898cd00370a26c957953f4fb1acd7252 (patch)
tree837e570397d6888acb7fc4720010ccf923524e46 /clang/lib
parent62875fcd6ce7b3028b468e0d847f76bbdee79945 (diff)
downloadbcm5719-llvm-8451efad898cd00370a26c957953f4fb1acd7252.tar.gz
bcm5719-llvm-8451efad898cd00370a26c957953f4fb1acd7252.zip
[OPENMP] Add codegen for `depend` clauses on `target` directive.
Added basic support for codegen of `depend` clauses on `target` directive. llvm-svn: 322501
Diffstat (limited to 'clang/lib')
-rw-r--r--clang/lib/Basic/OpenMPKinds.cpp4
-rw-r--r--clang/lib/CodeGen/CGOpenMPRuntime.cpp276
-rw-r--r--clang/lib/CodeGen/CGOpenMPRuntime.h8
-rw-r--r--clang/lib/CodeGen/CGStmtOpenMP.cpp22
-rw-r--r--clang/lib/Sema/SemaOpenMP.cpp91
5 files changed, 260 insertions, 141 deletions
diff --git a/clang/lib/Basic/OpenMPKinds.cpp b/clang/lib/Basic/OpenMPKinds.cpp
index 5bb50e594ac..739fc914a31 100644
--- a/clang/lib/Basic/OpenMPKinds.cpp
+++ b/clang/lib/Basic/OpenMPKinds.cpp
@@ -891,6 +891,7 @@ void clang::getOpenMPCaptureRegions(
case OMPD_target_teams:
case OMPD_target_teams_distribute:
case OMPD_target_teams_distribute_simd:
+ CaptureRegions.push_back(OMPD_task);
CaptureRegions.push_back(OMPD_target);
CaptureRegions.push_back(OMPD_teams);
break;
@@ -901,6 +902,7 @@ void clang::getOpenMPCaptureRegions(
break;
case OMPD_target:
case OMPD_target_simd:
+ CaptureRegions.push_back(OMPD_task);
CaptureRegions.push_back(OMPD_target);
break;
case OMPD_teams_distribute_parallel_for:
@@ -911,6 +913,7 @@ void clang::getOpenMPCaptureRegions(
case OMPD_target_parallel:
case OMPD_target_parallel_for:
case OMPD_target_parallel_for_simd:
+ CaptureRegions.push_back(OMPD_task);
CaptureRegions.push_back(OMPD_target);
CaptureRegions.push_back(OMPD_parallel);
break;
@@ -925,6 +928,7 @@ void clang::getOpenMPCaptureRegions(
CaptureRegions.push_back(OMPD_taskloop);
break;
case OMPD_target_teams_distribute_parallel_for:
+ CaptureRegions.push_back(OMPD_task);
CaptureRegions.push_back(OMPD_target);
CaptureRegions.push_back(OMPD_teams);
CaptureRegions.push_back(OMPD_parallel);
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.cpp b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
index 9cff2a35784..88ff4120026 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -4187,6 +4187,11 @@ static void emitPrivatesInit(CodeGenFunction &CGF,
auto &C = CGF.getContext();
auto FI = std::next(KmpTaskTWithPrivatesQTyRD->field_begin());
LValue PrivatesBase = CGF.EmitLValueForField(TDBase, *FI);
+ OpenMPDirectiveKind Kind = isOpenMPTaskLoopDirective(D.getDirectiveKind())
+ ? OMPD_taskloop
+ : OMPD_task;
+ const CapturedStmt &CS = *D.getCapturedStmt(Kind);
+ CodeGenFunction::CGCapturedStmtInfo CapturesInfo(CS);
LValue SrcBase;
bool IsTargetTask =
isOpenMPTargetDataManagementDirective(D.getDirectiveKind()) ||
@@ -4195,16 +4200,12 @@ static void emitPrivatesInit(CodeGenFunction &CGF,
// PointersArray and SizesArray. The original variables for these arrays are
// not captured and we get their addresses explicitly.
if ((!IsTargetTask && !Data.FirstprivateVars.empty()) ||
- (IsTargetTask && Data.FirstprivateVars.size() > 3)) {
+ (IsTargetTask && KmpTaskSharedsPtr.isValid())) {
SrcBase = CGF.MakeAddrLValue(
CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
KmpTaskSharedsPtr, CGF.ConvertTypeForMem(SharedsPtrTy)),
SharedsTy);
}
- OpenMPDirectiveKind Kind = isOpenMPTaskLoopDirective(D.getDirectiveKind())
- ? OMPD_taskloop
- : OMPD_task;
- CodeGenFunction::CGCapturedStmtInfo CapturesInfo(*D.getCapturedStmt(Kind));
FI = cast<RecordDecl>(FI->getType()->getAsTagDecl())->field_begin();
for (auto &&Pair : Privates) {
auto *VD = Pair.second.PrivateCopy;
@@ -4218,17 +4219,19 @@ static void emitPrivatesInit(CodeGenFunction &CGF,
// PointersArray or SizesArray.
LValue SharedRefLValue;
QualType Type = OriginalVD->getType();
- if (IsTargetTask && isa<ImplicitParamDecl>(OriginalVD) &&
- isa<CapturedDecl>(OriginalVD->getDeclContext()) &&
- cast<CapturedDecl>(OriginalVD->getDeclContext())->getNumParams() ==
- 0 &&
- isa<TranslationUnitDecl>(
- cast<CapturedDecl>(OriginalVD->getDeclContext())
- ->getDeclContext())) {
+ auto *SharedField = CapturesInfo.lookup(OriginalVD);
+ if (IsTargetTask && !SharedField) {
+ assert(isa<ImplicitParamDecl>(OriginalVD) &&
+ isa<CapturedDecl>(OriginalVD->getDeclContext()) &&
+ cast<CapturedDecl>(OriginalVD->getDeclContext())
+ ->getNumParams() == 0 &&
+ isa<TranslationUnitDecl>(
+ cast<CapturedDecl>(OriginalVD->getDeclContext())
+ ->getDeclContext()) &&
+ "Expected artificial target data variable.");
SharedRefLValue =
CGF.MakeAddrLValue(CGF.GetAddrOfLocalVar(OriginalVD), Type);
} else {
- auto *SharedField = CapturesInfo.lookup(OriginalVD);
SharedRefLValue = CGF.EmitLValueForField(SrcBase, SharedField);
SharedRefLValue = CGF.MakeAddrLValue(
Address(SharedRefLValue.getPointer(), C.getDeclAlign(OriginalVD)),
@@ -7040,86 +7043,27 @@ void CGOpenMPRuntime::emitTargetCall(CodeGenFunction &CGF,
const OMPExecutableDirective &D,
llvm::Value *OutlinedFn,
llvm::Value *OutlinedFnID,
- const Expr *IfCond, const Expr *Device,
- ArrayRef<llvm::Value *> CapturedVars) {
+ const Expr *IfCond, const Expr *Device) {
if (!CGF.HaveInsertPoint())
return;
assert(OutlinedFn && "Invalid outlined function!");
- // Fill up the arrays with all the captured variables.
- MappableExprsHandler::MapValuesArrayTy KernelArgs;
- MappableExprsHandler::MapBaseValuesArrayTy BasePointers;
- MappableExprsHandler::MapValuesArrayTy Pointers;
- MappableExprsHandler::MapValuesArrayTy Sizes;
- MappableExprsHandler::MapFlagsArrayTy MapTypes;
-
- MappableExprsHandler::MapBaseValuesArrayTy CurBasePointers;
- MappableExprsHandler::MapValuesArrayTy CurPointers;
- MappableExprsHandler::MapValuesArrayTy CurSizes;
- MappableExprsHandler::MapFlagsArrayTy CurMapTypes;
-
- // Get mappable expression information.
- MappableExprsHandler MEHandler(D, CGF);
-
+ const bool RequiresOuterTask = D.hasClausesOfKind<OMPDependClause>();
+ llvm::SmallVector<llvm::Value *, 16> CapturedVars;
const CapturedStmt &CS = *D.getCapturedStmt(OMPD_target);
- auto RI = CS.getCapturedRecordDecl()->field_begin();
- auto CV = CapturedVars.begin();
- for (CapturedStmt::const_capture_iterator CI = CS.capture_begin(),
- CE = CS.capture_end();
- CI != CE; ++CI, ++RI, ++CV) {
- CurBasePointers.clear();
- CurPointers.clear();
- CurSizes.clear();
- CurMapTypes.clear();
-
- // VLA sizes are passed to the outlined region by copy and do not have map
- // information associated.
- if (CI->capturesVariableArrayType()) {
- CurBasePointers.push_back(*CV);
- CurPointers.push_back(*CV);
- CurSizes.push_back(CGF.getTypeSize(RI->getType()));
- // Copy to the device as an argument. No need to retrieve it.
- CurMapTypes.push_back(MappableExprsHandler::OMP_MAP_LITERAL |
- MappableExprsHandler::OMP_MAP_TARGET_PARAM);
- } else {
- // If we have any information in the map clause, we use it, otherwise we
- // just do a default mapping.
- MEHandler.generateInfoForCapture(CI, *CV, CurBasePointers, CurPointers,
- CurSizes, CurMapTypes);
- if (CurBasePointers.empty())
- MEHandler.generateDefaultMapInfo(*CI, **RI, *CV, CurBasePointers,
- CurPointers, CurSizes, CurMapTypes);
- }
- // We expect to have at least an element of information for this capture.
- assert(!CurBasePointers.empty() && "Non-existing map pointer for capture!");
- assert(CurBasePointers.size() == CurPointers.size() &&
- CurBasePointers.size() == CurSizes.size() &&
- CurBasePointers.size() == CurMapTypes.size() &&
- "Inconsistent map information sizes!");
-
- // The kernel args are always the first elements of the base pointers
- // associated with a capture.
- KernelArgs.push_back(*CurBasePointers.front());
- // We need to append the results of this capture to what we already have.
- BasePointers.append(CurBasePointers.begin(), CurBasePointers.end());
- Pointers.append(CurPointers.begin(), CurPointers.end());
- Sizes.append(CurSizes.begin(), CurSizes.end());
- MapTypes.append(CurMapTypes.begin(), CurMapTypes.end());
- }
+ auto &&ArgsCodegen = [&CS, &CapturedVars](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ CGF.GenerateOpenMPCapturedVars(CS, CapturedVars);
+ };
+ emitInlinedDirective(CGF, OMPD_unknown, ArgsCodegen);
+ CodeGenFunction::OMPTargetDataInfo InputInfo;
+ llvm::Value *MapTypesArray = nullptr;
// Fill up the pointer arrays and transfer execution to the device.
- auto &&ThenGen = [this, &BasePointers, &Pointers, &Sizes, &MapTypes, Device,
- OutlinedFn, OutlinedFnID, &D,
- &KernelArgs](CodeGenFunction &CGF, PrePostActionTy &) {
- auto &RT = CGF.CGM.getOpenMPRuntime();
- // Emit the offloading arrays.
- TargetDataInfo Info;
- emitOffloadingArrays(CGF, BasePointers, Pointers, Sizes, MapTypes, Info);
- emitOffloadingArraysArgument(CGF, Info.BasePointersArray,
- Info.PointersArray, Info.SizesArray,
- Info.MapTypesArray, Info);
-
+ auto &&ThenGen = [this, Device, OutlinedFn, OutlinedFnID, &D, &InputInfo,
+ &MapTypesArray, &CS, RequiresOuterTask,
+ &CapturedVars](CodeGenFunction &CGF, PrePostActionTy &) {
// On top of the arrays that were filled up, the target offloading call
// takes as arguments the device id as well as the host pointer. The host
// pointer is used by the runtime library to identify the current target
@@ -7142,13 +7086,14 @@ void CGOpenMPRuntime::emitTargetCall(CodeGenFunction &CGF,
}
// Emit the number of elements in the offloading arrays.
- llvm::Value *PointerNum = CGF.Builder.getInt32(BasePointers.size());
+ llvm::Value *PointerNum =
+ CGF.Builder.getInt32(InputInfo.NumberOfTargetItems);
// Return value of the runtime offloading call.
llvm::Value *Return;
- auto *NumTeams = emitNumTeamsForTargetDirective(RT, CGF, D);
- auto *NumThreads = emitNumThreadsForTargetDirective(RT, CGF, D);
+ auto *NumTeams = emitNumTeamsForTargetDirective(*this, CGF, D);
+ auto *NumThreads = emitNumThreadsForTargetDirective(*this, CGF, D);
bool HasNowait = D.hasClausesOfKind<OMPNowaitClause>();
// The target region is an outlined function launched by the runtime
@@ -7186,25 +7131,30 @@ void CGOpenMPRuntime::emitTargetCall(CodeGenFunction &CGF,
// passed to the runtime library - a 32-bit integer with the value zero.
assert(NumThreads && "Thread limit expression should be available along "
"with number of teams.");
- llvm::Value *OffloadingArgs[] = {
- DeviceID, OutlinedFnID,
- PointerNum, Info.BasePointersArray,
- Info.PointersArray, Info.SizesArray,
- Info.MapTypesArray, NumTeams,
- NumThreads};
+ llvm::Value *OffloadingArgs[] = {DeviceID,
+ OutlinedFnID,
+ PointerNum,
+ InputInfo.BasePointersArray.getPointer(),
+ InputInfo.PointersArray.getPointer(),
+ InputInfo.SizesArray.getPointer(),
+ MapTypesArray,
+ NumTeams,
+ NumThreads};
Return = CGF.EmitRuntimeCall(
- RT.createRuntimeFunction(HasNowait ? OMPRTL__tgt_target_teams_nowait
- : OMPRTL__tgt_target_teams),
+ createRuntimeFunction(HasNowait ? OMPRTL__tgt_target_teams_nowait
+ : OMPRTL__tgt_target_teams),
OffloadingArgs);
} else {
- llvm::Value *OffloadingArgs[] = {
- DeviceID, OutlinedFnID,
- PointerNum, Info.BasePointersArray,
- Info.PointersArray, Info.SizesArray,
- Info.MapTypesArray};
+ llvm::Value *OffloadingArgs[] = {DeviceID,
+ OutlinedFnID,
+ PointerNum,
+ InputInfo.BasePointersArray.getPointer(),
+ InputInfo.PointersArray.getPointer(),
+ InputInfo.SizesArray.getPointer(),
+ MapTypesArray};
Return = CGF.EmitRuntimeCall(
- RT.createRuntimeFunction(HasNowait ? OMPRTL__tgt_target_nowait
- : OMPRTL__tgt_target),
+ createRuntimeFunction(HasNowait ? OMPRTL__tgt_target_nowait
+ : OMPRTL__tgt_target),
OffloadingArgs);
}
@@ -7217,17 +7167,114 @@ void CGOpenMPRuntime::emitTargetCall(CodeGenFunction &CGF,
CGF.Builder.CreateCondBr(Failed, OffloadFailedBlock, OffloadContBlock);
CGF.EmitBlock(OffloadFailedBlock);
- emitOutlinedFunctionCall(CGF, D.getLocStart(), OutlinedFn, KernelArgs);
+ if (RequiresOuterTask) {
+ CapturedVars.clear();
+ CGF.GenerateOpenMPCapturedVars(CS, CapturedVars);
+ }
+ emitOutlinedFunctionCall(CGF, D.getLocStart(), OutlinedFn, CapturedVars);
CGF.EmitBranch(OffloadContBlock);
CGF.EmitBlock(OffloadContBlock, /*IsFinished=*/true);
};
// Notify that the host version must be executed.
- auto &&ElseGen = [this, &D, OutlinedFn, &KernelArgs](CodeGenFunction &CGF,
- PrePostActionTy &) {
- emitOutlinedFunctionCall(CGF, D.getLocStart(), OutlinedFn,
- KernelArgs);
+ auto &&ElseGen = [this, &D, OutlinedFn, &CS, &CapturedVars,
+ RequiresOuterTask](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ if (RequiresOuterTask) {
+ CapturedVars.clear();
+ CGF.GenerateOpenMPCapturedVars(CS, CapturedVars);
+ }
+ emitOutlinedFunctionCall(CGF, D.getLocStart(), OutlinedFn, CapturedVars);
+ };
+
+ auto &&TargetThenGen = [this, &ThenGen, &D, &InputInfo, &MapTypesArray,
+ &CapturedVars, RequiresOuterTask,
+ &CS](CodeGenFunction &CGF, PrePostActionTy &) {
+ // Fill up the arrays with all the captured variables.
+ MappableExprsHandler::MapBaseValuesArrayTy BasePointers;
+ MappableExprsHandler::MapValuesArrayTy Pointers;
+ MappableExprsHandler::MapValuesArrayTy Sizes;
+ MappableExprsHandler::MapFlagsArrayTy MapTypes;
+
+ MappableExprsHandler::MapBaseValuesArrayTy CurBasePointers;
+ MappableExprsHandler::MapValuesArrayTy CurPointers;
+ MappableExprsHandler::MapValuesArrayTy CurSizes;
+ MappableExprsHandler::MapFlagsArrayTy CurMapTypes;
+
+ // Get mappable expression information.
+ MappableExprsHandler MEHandler(D, CGF);
+
+ auto RI = CS.getCapturedRecordDecl()->field_begin();
+ auto CV = CapturedVars.begin();
+ for (CapturedStmt::const_capture_iterator CI = CS.capture_begin(),
+ CE = CS.capture_end();
+ CI != CE; ++CI, ++RI, ++CV) {
+ CurBasePointers.clear();
+ CurPointers.clear();
+ CurSizes.clear();
+ CurMapTypes.clear();
+
+ // VLA sizes are passed to the outlined region by copy and do not have map
+ // information associated.
+ if (CI->capturesVariableArrayType()) {
+ CurBasePointers.push_back(*CV);
+ CurPointers.push_back(*CV);
+ CurSizes.push_back(CGF.getTypeSize(RI->getType()));
+ // Copy to the device as an argument. No need to retrieve it.
+ CurMapTypes.push_back(MappableExprsHandler::OMP_MAP_LITERAL |
+ MappableExprsHandler::OMP_MAP_TARGET_PARAM);
+ } else {
+ // If we have any information in the map clause, we use it, otherwise we
+ // just do a default mapping.
+ MEHandler.generateInfoForCapture(CI, *CV, CurBasePointers, CurPointers,
+ CurSizes, CurMapTypes);
+ if (CurBasePointers.empty())
+ MEHandler.generateDefaultMapInfo(*CI, **RI, *CV, CurBasePointers,
+ CurPointers, CurSizes, CurMapTypes);
+ }
+ // We expect to have at least an element of information for this capture.
+ assert(!CurBasePointers.empty() &&
+ "Non-existing map pointer for capture!");
+ assert(CurBasePointers.size() == CurPointers.size() &&
+ CurBasePointers.size() == CurSizes.size() &&
+ CurBasePointers.size() == CurMapTypes.size() &&
+ "Inconsistent map information sizes!");
+
+ // We need to append the results of this capture to what we already have.
+ BasePointers.append(CurBasePointers.begin(), CurBasePointers.end());
+ Pointers.append(CurPointers.begin(), CurPointers.end());
+ Sizes.append(CurSizes.begin(), CurSizes.end());
+ MapTypes.append(CurMapTypes.begin(), CurMapTypes.end());
+ }
+
+ TargetDataInfo Info;
+ // Fill up the arrays and create the arguments.
+ emitOffloadingArrays(CGF, BasePointers, Pointers, Sizes, MapTypes, Info);
+ emitOffloadingArraysArgument(CGF, Info.BasePointersArray,
+ Info.PointersArray, Info.SizesArray,
+ Info.MapTypesArray, Info);
+ InputInfo.NumberOfTargetItems = Info.NumberOfPtrs;
+ InputInfo.BasePointersArray =
+ Address(Info.BasePointersArray, CGM.getPointerAlign());
+ InputInfo.PointersArray =
+ Address(Info.PointersArray, CGM.getPointerAlign());
+ InputInfo.SizesArray = Address(Info.SizesArray, CGM.getPointerAlign());
+ MapTypesArray = Info.MapTypesArray;
+ if (RequiresOuterTask)
+ CGF.EmitOMPTargetTaskBasedDirective(D, ThenGen, InputInfo);
+ else
+ emitInlinedDirective(CGF, D.getDirectiveKind(), ThenGen);
+ };
+
+ auto &&TargetElseGen = [this, &ElseGen, &D, RequiresOuterTask](
+ CodeGenFunction &CGF, PrePostActionTy &) {
+ if (RequiresOuterTask) {
+ CodeGenFunction::OMPTargetDataInfo InputInfo;
+ CGF.EmitOMPTargetTaskBasedDirective(D, ElseGen, InputInfo);
+ } else {
+ emitInlinedDirective(CGF, D.getDirectiveKind(), ElseGen);
+ }
};
// If we have a target function ID it means that we need to support
@@ -7235,14 +7282,14 @@ void CGOpenMPRuntime::emitTargetCall(CodeGenFunction &CGF,
// regardless of the conditional in the if clause if, e.g., the user do not
// specify target triples.
if (OutlinedFnID) {
- if (IfCond)
- emitOMPIfClause(CGF, IfCond, ThenGen, ElseGen);
- else {
- RegionCodeGenTy ThenRCG(ThenGen);
+ if (IfCond) {
+ emitOMPIfClause(CGF, IfCond, TargetThenGen, TargetElseGen);
+ } else {
+ RegionCodeGenTy ThenRCG(TargetThenGen);
ThenRCG(CGF);
}
} else {
- RegionCodeGenTy ElseRCG(ElseGen);
+ RegionCodeGenTy ElseRCG(TargetElseGen);
ElseRCG(CGF);
}
}
@@ -8260,8 +8307,7 @@ void CGOpenMPSIMDRuntime::emitTargetCall(CodeGenFunction &CGF,
const OMPExecutableDirective &D,
llvm::Value *OutlinedFn,
llvm::Value *OutlinedFnID,
- const Expr *IfCond, const Expr *Device,
- ArrayRef<llvm::Value *> CapturedVars) {
+ const Expr *IfCond, const Expr *Device) {
llvm_unreachable("Not supported in SIMD-only mode");
}
diff --git a/clang/lib/CodeGen/CGOpenMPRuntime.h b/clang/lib/CodeGen/CGOpenMPRuntime.h
index 706b48c235d..e1ba8334368 100644
--- a/clang/lib/CodeGen/CGOpenMPRuntime.h
+++ b/clang/lib/CodeGen/CGOpenMPRuntime.h
@@ -1207,13 +1207,11 @@ public:
/// directive, or null if no if clause is used.
/// \param Device Expression evaluated in device clause associated with the
/// target directive, or null if no device clause is used.
- /// \param CapturedVars Values captured in the current region.
virtual void emitTargetCall(CodeGenFunction &CGF,
const OMPExecutableDirective &D,
llvm::Value *OutlinedFn,
llvm::Value *OutlinedFnID, const Expr *IfCond,
- const Expr *Device,
- ArrayRef<llvm::Value *> CapturedVars);
+ const Expr *Device);
/// \brief Emit the target regions enclosed in \a GD function definition or
/// the function itself in case it is a valid device function. Returns true if
@@ -1833,11 +1831,9 @@ public:
/// directive, or null if no if clause is used.
/// \param Device Expression evaluated in device clause associated with the
/// target directive, or null if no device clause is used.
- /// \param CapturedVars Values captured in the current region.
void emitTargetCall(CodeGenFunction &CGF, const OMPExecutableDirective &D,
llvm::Value *OutlinedFn, llvm::Value *OutlinedFnID,
- const Expr *IfCond, const Expr *Device,
- ArrayRef<llvm::Value *> CapturedVars) override;
+ const Expr *IfCond, const Expr *Device) override;
/// \brief Emit the target regions enclosed in \a GD function definition or
/// the function itself in case it is a valid device function. Returns true if
diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp
index 116647d7cfa..c6a0cdb5a54 100644
--- a/clang/lib/CodeGen/CGStmtOpenMP.cpp
+++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp
@@ -3091,12 +3091,14 @@ void CodeGenFunction::EmitOMPTargetTaskBasedDirective(
}
// Privatize all private variables except for in_reduction items.
(void)Scope.Privatize();
- InputInfo.BasePointersArray = CGF.Builder.CreateConstArrayGEP(
- CGF.GetAddrOfLocalVar(BPVD), /*Index=*/0, CGF.getPointerSize());
- InputInfo.PointersArray = CGF.Builder.CreateConstArrayGEP(
- CGF.GetAddrOfLocalVar(PVD), /*Index=*/0, CGF.getPointerSize());
- InputInfo.SizesArray = CGF.Builder.CreateConstArrayGEP(
- CGF.GetAddrOfLocalVar(SVD), /*Index=*/0, CGF.getSizeSize());
+ if (InputInfo.NumberOfTargetItems > 0) {
+ InputInfo.BasePointersArray = CGF.Builder.CreateConstArrayGEP(
+ CGF.GetAddrOfLocalVar(BPVD), /*Index=*/0, CGF.getPointerSize());
+ InputInfo.PointersArray = CGF.Builder.CreateConstArrayGEP(
+ CGF.GetAddrOfLocalVar(PVD), /*Index=*/0, CGF.getPointerSize());
+ InputInfo.SizesArray = CGF.Builder.CreateConstArrayGEP(
+ CGF.GetAddrOfLocalVar(SVD), /*Index=*/0, CGF.getSizeSize());
+ }
Action.Enter(CGF);
OMPLexicalScope LexScope(CGF, S, OMPD_task, /*EmitPreInitStmt=*/false);
@@ -3910,7 +3912,6 @@ static void emitCommonOMPTargetDirective(CodeGenFunction &CGF,
const RegionCodeGenTy &CodeGen) {
assert(isOpenMPTargetExecutionDirective(S.getDirectiveKind()));
CodeGenModule &CGM = CGF.CGM;
- const CapturedStmt &CS = *S.getCapturedStmt(OMPD_target);
llvm::Function *Fn = nullptr;
llvm::Constant *FnID = nullptr;
@@ -3958,11 +3959,8 @@ static void emitCommonOMPTargetDirective(CodeGenFunction &CGF,
// Emit target region as a standalone region.
CGM.getOpenMPRuntime().emitTargetOutlinedFunction(S, ParentName, Fn, FnID,
IsOffloadEntry, CodeGen);
- OMPLexicalScope Scope(CGF, S);
- llvm::SmallVector<llvm::Value *, 16> CapturedVars;
- CGF.GenerateOpenMPCapturedVars(CS, CapturedVars);
- CGM.getOpenMPRuntime().emitTargetCall(CGF, S, Fn, FnID, IfCond, Device,
- CapturedVars);
+ OMPLexicalScope Scope(CGF, S, OMPD_task);
+ CGM.getOpenMPRuntime().emitTargetCall(CGF, S, Fn, FnID, IfCond, Device);
}
static void emitTargetRegion(CodeGenFunction &CGF, const OMPTargetDirective &S,
diff --git a/clang/lib/Sema/SemaOpenMP.cpp b/clang/lib/Sema/SemaOpenMP.cpp
index 8c7111fe646..4f45d6d1a3b 100644
--- a/clang/lib/Sema/SemaOpenMP.cpp
+++ b/clang/lib/Sema/SemaOpenMP.cpp
@@ -2121,13 +2121,28 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
case OMPD_target_parallel_for_simd:
case OMPD_target_teams_distribute:
case OMPD_target_teams_distribute_simd: {
+ QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1);
+ QualType Args[] = {Context.VoidPtrTy.withConst().withRestrict()};
+ FunctionProtoType::ExtProtoInfo EPI;
+ EPI.Variadic = true;
+ QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
+ Sema::CapturedParamNameType Params[] = {
+ std::make_pair(".global_tid.", KmpInt32Ty),
+ std::make_pair(".part_id.", Context.getPointerType(KmpInt32Ty)),
+ std::make_pair(".privates.", Context.VoidPtrTy.withConst()),
+ std::make_pair(".copy_fn.",
+ Context.getPointerType(CopyFnType).withConst()),
+ std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
+ std::make_pair(StringRef(), QualType()) // __context with shared vars
+ };
+ ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
+ Params);
Sema::CapturedParamNameType ParamsTarget[] = {
std::make_pair(StringRef(), QualType()) // __context with shared vars
};
// Start a captured region for 'target' with no implicit parameters.
ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
ParamsTarget);
- QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1);
QualType KmpInt32PtrTy =
Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
Sema::CapturedParamNameType ParamsTeamsOrParallel[] = {
@@ -2141,6 +2156,33 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
ParamsTeamsOrParallel);
break;
}
+ case OMPD_target:
+ case OMPD_target_simd: {
+ QualType KmpInt32Ty = Context.getIntTypeForBitwidth(32, 1);
+ QualType Args[] = {Context.VoidPtrTy.withConst().withRestrict()};
+ FunctionProtoType::ExtProtoInfo EPI;
+ EPI.Variadic = true;
+ QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
+ Sema::CapturedParamNameType Params[] = {
+ std::make_pair(".global_tid.", KmpInt32Ty),
+ std::make_pair(".part_id.", Context.getPointerType(KmpInt32Ty)),
+ std::make_pair(".privates.", Context.VoidPtrTy.withConst()),
+ std::make_pair(".copy_fn.",
+ Context.getPointerType(CopyFnType).withConst()),
+ std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
+ std::make_pair(StringRef(), QualType()) // __context with shared vars
+ };
+ ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
+ Params);
+ // Mark this captured region as inlined, because we don't use outlined
+ // function directly.
+ getCurCapturedRegion()->TheCapturedDecl->addAttr(
+ AlwaysInlineAttr::CreateImplicit(
+ Context, AlwaysInlineAttr::Keyword_forceinline, SourceRange()));
+ ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
+ std::make_pair(StringRef(), QualType()));
+ break;
+ }
case OMPD_simd:
case OMPD_for:
case OMPD_for_simd:
@@ -2154,9 +2196,7 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
case OMPD_distribute_simd:
case OMPD_ordered:
case OMPD_atomic:
- case OMPD_target_data:
- case OMPD_target:
- case OMPD_target_simd: {
+ case OMPD_target_data: {
Sema::CapturedParamNameType Params[] = {
std::make_pair(StringRef(), QualType()) // __context with shared vars
};
@@ -2247,6 +2287,21 @@ void Sema::ActOnOpenMPRegionStart(OpenMPDirectiveKind DKind, Scope *CurScope) {
QualType KmpInt32PtrTy =
Context.getPointerType(KmpInt32Ty).withConst().withRestrict();
+ QualType Args[] = {Context.VoidPtrTy.withConst().withRestrict()};
+ FunctionProtoType::ExtProtoInfo EPI;
+ EPI.Variadic = true;
+ QualType CopyFnType = Context.getFunctionType(Context.VoidTy, Args, EPI);
+ Sema::CapturedParamNameType Params[] = {
+ std::make_pair(".global_tid.", KmpInt32Ty),
+ std::make_pair(".part_id.", Context.getPointerType(KmpInt32Ty)),
+ std::make_pair(".privates.", Context.VoidPtrTy.withConst()),
+ std::make_pair(".copy_fn.",
+ Context.getPointerType(CopyFnType).withConst()),
+ std::make_pair(".task_t.", Context.VoidPtrTy.withConst()),
+ std::make_pair(StringRef(), QualType()) // __context with shared vars
+ };
+ ActOnCapturedRegionStart(DSAStack->getConstructLoc(), CurScope, CR_OpenMP,
+ Params);
Sema::CapturedParamNameType ParamsTarget[] = {
std::make_pair(StringRef(), QualType()) // __context with shared vars
};
@@ -6354,13 +6409,23 @@ StmtResult Sema::ActOnOpenMPTargetDirective(ArrayRef<OMPClause *> Clauses,
// The point of exit cannot be a branch out of the structured block.
// longjmp() and throw() must not violate the entry/exit criteria.
CS->getCapturedDecl()->setNothrow();
+ for (int ThisCaptureLevel = getOpenMPCaptureLevels(OMPD_target);
+ ThisCaptureLevel > 1; --ThisCaptureLevel) {
+ CS = cast<CapturedStmt>(CS->getCapturedStmt());
+ // 1.2.2 OpenMP Language Terminology
+ // Structured block - An executable statement with a single entry at the
+ // top and a single exit at the bottom.
+ // The point of exit cannot be a branch out of the structured block.
+ // longjmp() and throw() must not violate the entry/exit criteria.
+ CS->getCapturedDecl()->setNothrow();
+ }
// OpenMP [2.16, Nesting of Regions]
// If specified, a teams construct must be contained within a target
// construct. That target construct must contain no statements or directives
// outside of the teams construct.
if (DSAStack->hasInnerTeamsRegion()) {
- auto S = AStmt->IgnoreContainers(/*IgnoreCaptured*/ true);
+ Stmt *S = CS->IgnoreContainers(/*IgnoreCaptured=*/true);
bool OMPTeamsFound = true;
if (auto *CS = dyn_cast<CompoundStmt>(S)) {
auto I = CS->body_begin();
@@ -6407,6 +6472,16 @@ Sema::ActOnOpenMPTargetParallelDirective(ArrayRef<OMPClause *> Clauses,
// The point of exit cannot be a branch out of the structured block.
// longjmp() and throw() must not violate the entry/exit criteria.
CS->getCapturedDecl()->setNothrow();
+ for (int ThisCaptureLevel = getOpenMPCaptureLevels(OMPD_target_parallel);
+ ThisCaptureLevel > 1; --ThisCaptureLevel) {
+ CS = cast<CapturedStmt>(CS->getCapturedStmt());
+ // 1.2.2 OpenMP Language Terminology
+ // Structured block - An executable statement with a single entry at the
+ // top and a single exit at the bottom.
+ // The point of exit cannot be a branch out of the structured block.
+ // longjmp() and throw() must not violate the entry/exit criteria.
+ CS->getCapturedDecl()->setNothrow();
+ }
getCurFunction()->setHasBranchProtectedScope();
@@ -8049,6 +8124,7 @@ static OpenMPDirectiveKind getOpenMPCaptureRegionForClause(
case OMPD_target_update:
case OMPD_target_enter_data:
case OMPD_target_exit_data:
+ case OMPD_target:
CaptureRegion = OMPD_task;
break;
case OMPD_target_teams:
@@ -8057,7 +8133,6 @@ static OpenMPDirectiveKind getOpenMPCaptureRegionForClause(
case OMPD_target_teams_distribute_parallel_for:
case OMPD_target_teams_distribute_parallel_for_simd:
case OMPD_target_data:
- case OMPD_target:
case OMPD_target_simd:
case OMPD_target_parallel:
case OMPD_target_parallel_for:
@@ -11419,8 +11494,8 @@ OMPClause *Sema::ActOnOpenMPDeviceClause(Expr *Device, SourceLocation StartLoc,
HelperValStmt = buildPreInits(Context, Captures);
}
- return new (Context)
- OMPDeviceClause(ValExpr, HelperValStmt, StartLoc, LParenLoc, EndLoc);
+ return new (Context) OMPDeviceClause(ValExpr, HelperValStmt, CaptureRegion,
+ StartLoc, LParenLoc, EndLoc);
}
static bool CheckTypeMappable(SourceLocation SL, SourceRange SR, Sema &SemaRef,
OpenPOWER on IntegriCloud