clang
10.0.0svn
|
#include "CGOpenMPRuntimeNVPTX.h"
#include "CodeGenFunction.h"
#include "clang/AST/DeclOpenMP.h"
#include "clang/AST/StmtOpenMP.h"
#include "clang/AST/StmtVisitor.h"
#include "clang/Basic/Cuda.h"
#include "llvm/ADT/SmallPtrSet.h"
Go to the source code of this file.
Classes | |
struct | CopyOptionsTy |
Enumerations | |
enum | OpenMPRTLFunctionNVPTX |
enum | MachineConfiguration : unsigned |
GPU Configuration: This information can be derived from cuda registers, however, providing compile time constants helps generate more efficient code. More... | |
enum | ModeFlagsTy : unsigned |
Enum for accesseing the reserved_2 field of the ident_t struct. More... | |
enum | CopyAction : unsigned |
Functions | |
static llvm::Value * | getNVPTXWarpSize (CodeGenFunction &CGF) |
Get the GPU warp size. More... | |
static llvm::Value * | getNVPTXThreadID (CodeGenFunction &CGF) |
Get the id of the current thread on the GPU. More... | |
static llvm::Value * | getNVPTXWarpID (CodeGenFunction &CGF) |
Get the id of the warp in the block. More... | |
static llvm::Value * | getNVPTXLaneID (CodeGenFunction &CGF) |
Get the id of the current lane in the Warp. More... | |
static llvm::Value * | getNVPTXNumThreads (CodeGenFunction &CGF) |
Get the maximum number of threads in a block of the GPU. More... | |
static llvm::Value * | getThreadLimit (CodeGenFunction &CGF, bool IsInSPMDExecutionMode=false) |
Get the value of the thread_limit clause in the teams directive. More... | |
static llvm::Value * | getMasterThreadID (CodeGenFunction &CGF) |
Get the thread id of the OMP master thread. More... | |
static CGOpenMPRuntimeNVPTX::DataSharingMode | getDataSharingMode (CodeGenModule &CGM) |
static bool | hasNestedSPMDDirective (ASTContext &Ctx, const OMPExecutableDirective &D) |
Check for inner (nested) SPMD construct, if any. More... | |
static bool | supportsSPMDExecutionMode (ASTContext &Ctx, const OMPExecutableDirective &D) |
static bool | hasStaticScheduling (const OMPExecutableDirective &D) |
Check if the directive is loops based and has schedule clause at all or has static scheduling. More... | |
static bool | hasNestedLightweightDirective (ASTContext &Ctx, const OMPExecutableDirective &D) |
Check for inner (nested) lightweight runtime construct, if any. More... | |
static bool | supportsLightweightRuntime (ASTContext &Ctx, const OMPExecutableDirective &D) |
Checks if the construct supports lightweight runtime. More... | |
static void | setPropertyExecutionMode (CodeGenModule &CGM, StringRef Name, bool Mode) |
static void | getDistributeLastprivateVars (ASTContext &Ctx, const OMPExecutableDirective &D, llvm::SmallVectorImpl< const ValueDecl *> &Vars) |
Get list of lastprivate variables from the teams distribute ... More... | |
static void | getTeamsReductionVars (ASTContext &Ctx, const OMPExecutableDirective &D, llvm::SmallVectorImpl< const ValueDecl *> &Vars) |
Get list of reduction variables from the teams ... directives. More... | |
static llvm::Value * | castValueToType (CodeGenFunction &CGF, llvm::Value *Val, QualType ValTy, QualType CastTy, SourceLocation Loc) |
Cast value to the specified type. More... | |
static llvm::Value * | createRuntimeShuffleFunction (CodeGenFunction &CGF, llvm::Value *Elem, QualType ElemType, llvm::Value *Offset, SourceLocation Loc) |
This function creates calls to one of two shuffle functions to copy variables between lanes in a warp. More... | |
static void | shuffleAndStore (CodeGenFunction &CGF, Address SrcAddr, Address DestAddr, QualType ElemType, llvm::Value *Offset, SourceLocation Loc) |
static void | emitReductionListCopy (CopyAction Action, CodeGenFunction &CGF, QualType ReductionArrayTy, ArrayRef< const Expr *> Privates, Address SrcBase, Address DestBase, CopyOptionsTy CopyOptions={nullptr, nullptr, nullptr}) |
Emit instructions to copy a Reduce list, which contains partially aggregated values, in the specified direction. More... | |
static llvm::Value * | emitInterWarpCopyFunction (CodeGenModule &CGM, ArrayRef< const Expr *> Privates, QualType ReductionArrayTy, SourceLocation Loc) |
This function emits a helper that gathers Reduce lists from the first lane of every active warp to lanes in the first warp. More... | |
static llvm::Function * | emitShuffleAndReduceFunction (CodeGenModule &CGM, ArrayRef< const Expr *> Privates, QualType ReductionArrayTy, llvm::Function *ReduceFn, SourceLocation Loc) |
Emit a helper that reduces data across two OpenMP threads (lanes) in the same warp. More... | |
static llvm::Value * | emitListToGlobalCopyFunction (CodeGenModule &CGM, ArrayRef< const Expr *> Privates, QualType ReductionArrayTy, SourceLocation Loc, const RecordDecl *TeamReductionRec, const llvm::SmallDenseMap< const ValueDecl *, const FieldDecl *> &VarFieldMap) |
This function emits a helper that copies all the reduction variables from the team into the provided global buffer for the reduction variables. More... | |
static llvm::Value * | emitListToGlobalReduceFunction (CodeGenModule &CGM, ArrayRef< const Expr *> Privates, QualType ReductionArrayTy, SourceLocation Loc, const RecordDecl *TeamReductionRec, const llvm::SmallDenseMap< const ValueDecl *, const FieldDecl *> &VarFieldMap, llvm::Function *ReduceFn) |
This function emits a helper that reduces all the reduction variables from the team into the provided global buffer for the reduction variables. More... | |
static llvm::Value * | emitGlobalToListCopyFunction (CodeGenModule &CGM, ArrayRef< const Expr *> Privates, QualType ReductionArrayTy, SourceLocation Loc, const RecordDecl *TeamReductionRec, const llvm::SmallDenseMap< const ValueDecl *, const FieldDecl *> &VarFieldMap) |
This function emits a helper that copies all the reduction variables from the team into the provided global buffer for the reduction variables. More... | |
static llvm::Value * | emitGlobalToListReduceFunction (CodeGenModule &CGM, ArrayRef< const Expr *> Privates, QualType ReductionArrayTy, SourceLocation Loc, const RecordDecl *TeamReductionRec, const llvm::SmallDenseMap< const ValueDecl *, const FieldDecl *> &VarFieldMap, llvm::Function *ReduceFn) |
This function emits a helper that reduces all the reduction variables from the team into the provided global buffer for the reduction variables. More... | |
static CudaArch | getCudaArch (CodeGenModule &CGM) |
static std::pair< unsigned, unsigned > | getSMsBlocksPerSM (CodeGenModule &CGM) |
Get number of SMs and number of blocks per SM. More... | |
enum CopyAction : unsigned |
Definition at line 2942 of file CGOpenMPRuntimeNVPTX.cpp.
enum MachineConfiguration : unsigned |
GPU Configuration: This information can be derived from cuda registers, however, providing compile time constants helps generate more efficient code.
For all practical purposes this is fine because the configuration is the same for all known NVPTX architectures.
Definition at line 192 of file CGOpenMPRuntimeNVPTX.cpp.
enum ModeFlagsTy : unsigned |
Enum for accesseing the reserved_2 field of the ident_t struct.
Definition at line 1882 of file CGOpenMPRuntimeNVPTX.cpp.
Definition at line 26 of file CGOpenMPRuntimeNVPTX.cpp.
|
static |
Cast value to the specified type.
Definition at line 2813 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenFunction::Builder, clang::CodeGen::CodeGenFunction::ConvertTypeForMem(), clang::CodeGen::CGBuilderTy::CreateBitCast(), clang::CodeGen::CodeGenFunction::CreateMemTemp(), clang::CodeGen::CGBuilderTy::CreatePointerBitCastOrAddrSpaceCast(), clang::CodeGen::CodeGenFunction::EmitLoadOfScalar(), clang::CodeGen::CodeGenFunction::EmitStoreOfScalar(), clang::CodeGen::Address::getAddressSpace(), clang::CodeGen::CodeGenFunction::getContext(), clang::CodeGen::Address::getType(), clang::ASTContext::getTypeSizeInChars(), clang::Type::hasSignedIntegerRepresentation(), clang::Type::isIntegerType(), clang::CharUnits::isZero(), and clang::CodeGen::Type.
Referenced by createRuntimeShuffleFunction(), and clang::CodeGen::CGOpenMPRuntimeNVPTX::emitOutlinedFunctionCall().
|
static |
This function creates calls to one of two shuffle functions to copy variables between lanes in a warp.
Definition at line 2838 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenFunction::Builder, castValueToType(), clang::CodeGen::CodeGenFunction::CGM, clang::CodeGen::CGOpenMPRuntimeNVPTX::createNVPTXRuntimeFunction(), clang::CodeGen::CodeGenFunction::EmitRuntimeCall(), clang::CodeGen::CodeGenFunction::getContext(), clang::ASTContext::getIntTypeForBitwidth(), getNVPTXWarpSize(), clang::CodeGen::CodeGenModule::getOpenMPRuntime(), clang::CharUnits::getQuantity(), clang::ASTContext::getTypeSizeInChars(), clang::CodeGen::CodeGenTypeCache::Int16Ty, and Offset.
Referenced by shuffleAndStore().
|
static |
This function emits a helper that copies all the reduction variables from the team into the provided global buffer for the reduction variables.
void list_to_global_copy_func(void *buffer, int Idx, void *reduce_data) For all data entries D in reduce_data: Copy buffer.D[Idx] to local D;
Definition at line 3824 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenTypes::arrangeBuiltinFunctionDeclaration(), clang::CodeGen::CodeGenFunction::Builder, clang::CodeGen::CodeGenTypes::ConvertTypeForMem(), clang::CodeGen::CodeGenFunction::ConvertTypeForMem(), clang::Create(), clang::CodeGen::CGBuilderTy::CreateConstArrayGEP(), clang::CodeGen::CGBuilderTy::CreatePointerBitCastOrAddrSpaceCast(), clang::CodeGen::AggValueSlot::DoesNotOverlap, clang::CodeGen::CodeGenFunction::EmitAggregateCopy(), clang::CodeGen::CodeGenFunction::EmitLoadOfComplex(), clang::CodeGen::CodeGenFunction::EmitLoadOfScalar(), clang::CodeGen::CodeGenFunction::EmitLValueForField(), clang::CodeGen::CodeGenFunction::EmitStoreOfComplex(), clang::CodeGen::CodeGenFunction::EmitStoreOfScalar(), clang::CodeGen::CodeGenFunction::FinishFunction(), clang::CodeGen::CodeGenFunction::GetAddrOfLocalVar(), clang::CodeGen::LValue::getAlignment(), clang::CodeGen::CodeGenModule::getContext(), clang::CodeGen::CodeGenFunction::getEvaluationKind(), clang::CodeGen::CodeGenTypes::GetFunctionType(), clang::CodeGen::CodeGenModule::getModule(), clang::CodeGen::LValue::getPointer(), clang::CodeGen::CodeGenTypeCache::getPointerAlign(), clang::ASTContext::getRecordType(), clang::CodeGen::Address::getType(), clang::CodeGen::LValue::getType(), clang::ASTContext::getTypeAlignInChars(), clang::CodeGen::CodeGenModule::getTypes(), clang::CodeGen::CodeGenTypeCache::Int32Ty, clang::InternalLinkage, clang::ASTContext::IntTy, clang::CodeGen::CodeGenFunction::MakeAddrLValue(), clang::CodeGen::CodeGenFunction::MakeNaturalAlignAddrLValue(), clang::ImplicitParamDecl::Other, clang::CodeGen::LValue::setAddress(), clang::CodeGen::CodeGenModule::SetInternalFunctionAttributes(), clang::CodeGen::CodeGenFunction::StartFunction(), clang::CodeGen::TEK_Aggregate, clang::CodeGen::TEK_Complex, clang::CodeGen::TEK_Scalar, clang::CodeGen::Type, V, clang::ASTContext::VoidPtrTy, and clang::ASTContext::VoidTy.
|
static |
This function emits a helper that reduces all the reduction variables from the team into the provided global buffer for the reduction variables.
void global_to_list_reduce_func(void <em>buffer, int Idx, void *reduce_data) void *GlobPtrs[]; GlobPtrs[0] = (void)&buffer.D0[Idx]; ... GlobPtrs[N] = (void*)&buffer.DN[Idx]; reduce_function(reduce_data, GlobPtrs);
Definition at line 3929 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenTypes::arrangeBuiltinFunctionDeclaration(), clang::CodeGen::CodeGenFunction::Builder, clang::CodeGen::CodeGenTypes::ConvertTypeForMem(), clang::Create(), clang::CodeGen::CGBuilderTy::CreateConstArrayGEP(), clang::CodeGen::CodeGenFunction::CreateMemTemp(), clang::CodeGen::CGBuilderTy::CreatePointerBitCastOrAddrSpaceCast(), clang::CodeGen::CGBuilderTy::CreateStore(), clang::CodeGen::CodeGenFunction::EmitCastToVoidPtr(), clang::CodeGen::CodeGenFunction::EmitLoadOfScalar(), clang::CodeGen::CodeGenFunction::EmitLValueForField(), clang::CodeGen::CodeGenFunction::EmitStoreOfScalar(), clang::CodeGen::CodeGenFunction::FinishFunction(), clang::CodeGen::CodeGenFunction::GetAddrOfLocalVar(), clang::ASTContext::getAsVariableArrayType(), clang::CodeGen::CodeGenModule::getContext(), clang::CodeGen::CodeGenFunction::getContext(), clang::CodeGen::CodeGenTypes::GetFunctionType(), clang::CodeGen::CodeGenModule::getModule(), clang::CodeGen::CodeGenModule::getOpenMPRuntime(), clang::CodeGen::Address::getPointer(), clang::CodeGen::LValue::getPointer(), clang::ASTContext::getRecordType(), clang::CodeGen::CodeGenModule::getTypes(), clang::CodeGen::CodeGenFunction::getVLASize(), clang::CodeGen::CodeGenTypeCache::Int32Ty, clang::InternalLinkage, clang::ASTContext::IntTy, clang::CodeGen::CodeGenFunction::MakeNaturalAlignAddrLValue(), clang::CodeGen::CodeGenFunction::VlaSizePair::NumElts, clang::ImplicitParamDecl::Other, clang::CodeGen::CodeGenModule::SetInternalFunctionAttributes(), clang::CodeGen::CodeGenTypeCache::SizeTy, clang::CodeGen::CodeGenFunction::StartFunction(), clang::CodeGen::Type, clang::CodeGen::CodeGenTypeCache::VoidPtrTy, clang::ASTContext::VoidPtrTy, and clang::ASTContext::VoidTy.
|
static |
This function emits a helper that gathers Reduce lists from the first lane of every active warp to lanes in the first warp.
void inter_warp_copy_func(void* reduce_data, num_warps) shared smem[warp_size]; For all data entries D in reduce_data: sync If (I am the first lane in each warp) Copy my local D to smem[warp_id] sync if (I am the first warp) Copy smem[thread_id] to my local D
Definition at line 3170 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenModule::addCompilerUsedGlobal(), clang::CodeGen::CodeGenTypes::arrangeBuiltinFunctionDeclaration(), clang::CodeGen::CodeGenFunction::Builder, clang::CodeGen::CodeGenFunction::ConvertTypeForMem(), clang::Create(), clang::CodeGen::CodeGenFunction::createBasicBlock(), clang::CodeGen::ApplyDebugLocation::CreateEmpty(), clang::CodeGen::CodeGenFunction::CreateMemTemp(), clang::CodeGen::CGBuilderTy::CreatePointerBitCastOrAddrSpaceCast(), clang::cuda_shared, clang::CodeGen::CodeGenFunction::EmitBlock(), clang::CodeGen::CodeGenFunction::EmitLoadOfScalar(), clang::CodeGen::CodeGenFunction::EmitStoreOfScalar(), clang::CharUnits::fromQuantity(), clang::CodeGen::CodeGenFunction::GetAddrOfLocalVar(), clang::CodeGen::CodeGenModule::getContext(), clang::CodeGen::CodeGenTypes::GetFunctionType(), clang::ASTContext::getIntTypeForBitwidth(), clang::CodeGen::CodeGenModule::getModule(), getNVPTXLaneID(), getNVPTXThreadID(), getNVPTXWarpID(), clang::CodeGen::CodeGenTypeCache::getPointerAlign(), clang::CharUnits::getQuantity(), clang::ASTContext::getTargetAddressSpace(), clang::ASTContext::getTypeAlignInChars(), clang::CodeGen::CodeGenModule::getTypes(), clang::ASTContext::getTypeSizeInChars(), clang::CodeGen::CodeGenTypeCache::Int32Ty, clang::InternalLinkage, clang::CodeGen::CodeGenTypeCache::IntTy, clang::ASTContext::IntTy, clang::CodeGen::Address::invalid(), clang::ImplicitParamDecl::Other, clang::CodeGen::CodeGenModule::SetInternalFunctionAttributes(), clang::CodeGen::CodeGenFunction::StartFunction(), clang::ASTContext::toBits(), clang::CodeGen::Type, clang::ASTContext::VoidPtrTy, and clang::ASTContext::VoidTy.
|
static |
This function emits a helper that copies all the reduction variables from the team into the provided global buffer for the reduction variables.
void list_to_global_copy_func(void *buffer, int Idx, void *reduce_data) For all data entries D in reduce_data: Copy local D to buffer.D[Idx]
Definition at line 3624 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenTypes::arrangeBuiltinFunctionDeclaration(), clang::CodeGen::CodeGenFunction::Builder, clang::CodeGen::CodeGenTypes::ConvertTypeForMem(), clang::CodeGen::CodeGenFunction::ConvertTypeForMem(), clang::Create(), clang::CodeGen::CGBuilderTy::CreateConstArrayGEP(), clang::CodeGen::CGBuilderTy::CreatePointerBitCastOrAddrSpaceCast(), clang::CodeGen::AggValueSlot::DoesNotOverlap, clang::CodeGen::CodeGenFunction::EmitAggregateCopy(), clang::CodeGen::CodeGenFunction::EmitLoadOfComplex(), clang::CodeGen::CodeGenFunction::EmitLoadOfScalar(), clang::CodeGen::CodeGenFunction::EmitLValueForField(), clang::CodeGen::CodeGenFunction::EmitStoreOfComplex(), clang::CodeGen::CodeGenFunction::EmitStoreOfScalar(), clang::CodeGen::CodeGenFunction::FinishFunction(), clang::CodeGen::CodeGenFunction::GetAddrOfLocalVar(), clang::CodeGen::LValue::getAlignment(), clang::CodeGen::CodeGenModule::getContext(), clang::CodeGen::CodeGenFunction::getEvaluationKind(), clang::CodeGen::CodeGenTypes::GetFunctionType(), clang::CodeGen::CodeGenModule::getModule(), clang::CodeGen::LValue::getPointer(), clang::CodeGen::CodeGenTypeCache::getPointerAlign(), clang::ASTContext::getRecordType(), clang::CodeGen::Address::getType(), clang::CodeGen::LValue::getType(), clang::ASTContext::getTypeAlignInChars(), clang::CodeGen::CodeGenModule::getTypes(), clang::CodeGen::CodeGenTypeCache::Int32Ty, clang::InternalLinkage, clang::ASTContext::IntTy, clang::CodeGen::CodeGenFunction::MakeAddrLValue(), clang::CodeGen::CodeGenFunction::MakeNaturalAlignAddrLValue(), clang::ImplicitParamDecl::Other, clang::CodeGen::LValue::setAddress(), clang::CodeGen::CodeGenModule::SetInternalFunctionAttributes(), clang::CodeGen::CodeGenFunction::StartFunction(), clang::CodeGen::TEK_Aggregate, clang::CodeGen::TEK_Complex, clang::CodeGen::TEK_Scalar, clang::CodeGen::Type, V, clang::ASTContext::VoidPtrTy, and clang::ASTContext::VoidTy.
|
static |
This function emits a helper that reduces all the reduction variables from the team into the provided global buffer for the reduction variables.
void list_to_global_reduce_func(void <em>buffer, int Idx, void *reduce_data) void *GlobPtrs[]; GlobPtrs[0] = (void)&buffer.D0[Idx]; ... GlobPtrs[N] = (void*)&buffer.DN[Idx]; reduce_function(GlobPtrs, reduce_data);
Definition at line 3729 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenTypes::arrangeBuiltinFunctionDeclaration(), clang::CodeGen::CodeGenFunction::Builder, clang::CodeGen::CodeGenTypes::ConvertTypeForMem(), clang::Create(), clang::CodeGen::CGBuilderTy::CreateConstArrayGEP(), clang::CodeGen::CodeGenFunction::CreateMemTemp(), clang::CodeGen::CGBuilderTy::CreatePointerBitCastOrAddrSpaceCast(), clang::CodeGen::CGBuilderTy::CreateStore(), clang::CodeGen::CodeGenFunction::EmitCastToVoidPtr(), clang::CodeGen::CodeGenFunction::EmitLoadOfScalar(), clang::CodeGen::CodeGenFunction::EmitLValueForField(), clang::CodeGen::CodeGenFunction::EmitStoreOfScalar(), clang::CodeGen::CodeGenFunction::FinishFunction(), clang::CodeGen::CodeGenFunction::GetAddrOfLocalVar(), clang::ASTContext::getAsVariableArrayType(), clang::CodeGen::CodeGenModule::getContext(), clang::CodeGen::CodeGenFunction::getContext(), clang::CodeGen::CodeGenTypes::GetFunctionType(), clang::CodeGen::CodeGenModule::getModule(), clang::CodeGen::CodeGenModule::getOpenMPRuntime(), clang::CodeGen::Address::getPointer(), clang::CodeGen::LValue::getPointer(), clang::ASTContext::getRecordType(), clang::CodeGen::CodeGenModule::getTypes(), clang::CodeGen::CodeGenFunction::getVLASize(), clang::CodeGen::CodeGenTypeCache::Int32Ty, clang::InternalLinkage, clang::ASTContext::IntTy, clang::CodeGen::CodeGenFunction::MakeNaturalAlignAddrLValue(), clang::CodeGen::CodeGenFunction::VlaSizePair::NumElts, clang::ImplicitParamDecl::Other, clang::CodeGen::CodeGenModule::SetInternalFunctionAttributes(), clang::CodeGen::CodeGenTypeCache::SizeTy, clang::CodeGen::CodeGenFunction::StartFunction(), clang::CodeGen::Type, clang::CodeGen::CodeGenTypeCache::VoidPtrTy, clang::ASTContext::VoidPtrTy, and clang::ASTContext::VoidTy.
|
static |
Emit instructions to copy a Reduce list, which contains partially aggregated values, in the specified direction.
Definition at line 2964 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenFunction::Builder, clang::CodeGen::CodeGenFunction::CGM, clang::CodeGen::CodeGenFunction::ConvertTypeForMem(), clang::CodeGen::CGBuilderTy::CreateConstArrayGEP(), clang::CodeGen::CGBuilderTy::CreateElementBitCast(), clang::CodeGen::CodeGenFunction::CreateMemTemp(), clang::CodeGen::CGBuilderTy::CreatePointerBitCastOrAddrSpaceCast(), clang::CodeGen::AggValueSlot::DoesNotOverlap, clang::CodeGen::CodeGenFunction::EmitAggregateCopy(), clang::CodeGen::CodeGenFunction::EmitLoadOfComplex(), clang::CodeGen::CodeGenFunction::EmitLoadOfPointer(), clang::CodeGen::CodeGenFunction::EmitLoadOfScalar(), clang::CodeGen::CodeGenFunction::EmitStoreOfComplex(), clang::CodeGen::CodeGenFunction::EmitStoreOfScalar(), clang::CodeGen::CodeGenModule::getContext(), clang::CodeGen::Address::getElementType(), clang::CodeGen::CodeGenFunction::getEvaluationKind(), clang::CodeGen::Address::getPointer(), clang::CodeGen::CodeGenTypeCache::getPointerAlign(), clang::ASTContext::getPointerType(), clang::CodeGen::Address::getType(), clang::CodeGen::LValue::getType(), clang::ASTContext::getTypeAlignInChars(), clang::CodeGen::CodeGenFunction::getTypeSize(), clang::CodeGen::Address::invalid(), clang::CodeGen::CodeGenFunction::MakeAddrLValue(), shuffleAndStore(), clang::CodeGen::CodeGenTypeCache::SizeTy, clang::CodeGen::TEK_Aggregate, clang::CodeGen::TEK_Complex, clang::CodeGen::TEK_Scalar, clang::CodeGen::CodeGenTypeCache::VoidPtrTy, and clang::ASTContext::VoidPtrTy.
Referenced by emitShuffleAndReduceFunction().
|
static |
Emit a helper that reduces data across two OpenMP threads (lanes) in the same warp.
It uses shuffle instructions to copy over data from a remote lane's stack. The reduction algorithm performed is specified by the fourth parameter.
Algorithm Versions. Full Warp Reduce (argument value 0): This algorithm assumes that all 32 lanes are active and gathers data from these 32 lanes, producing a single resultant value. Contiguous Partial Warp Reduce (argument value 1): This algorithm assumes that only a contiguous subset of lanes are active. This happens for the last warp in a parallel region when the user specified num_threads is not an integer multiple of
Terminology Reduce element: Reduce element refers to the individual data field with primitive data types to be combined and reduced across threads. Reduce list: Reduce list refers to a collection of local, thread-private reduce elements. Remote Reduce list: Remote Reduce list refers to a collection of remote (relative to the current thread) reduce elements.
We distinguish between three states of threads that are important to the implementation of this function. Alive threads: Threads in a warp executing the SIMT instruction, as distinguished from threads that are inactive due to divergent control flow. Active threads: The minimal set of threads that has to be alive upon entry to this function. The computation is correct iff active threads are alive. Some threads are alive but they are not active because they do not contribute to the computation in any useful manner. Turning them off may introduce control flow overheads without any tangible benefits. Effective threads: In order to comply with the argument requirements of the shuffle function, we must keep all lanes holding data alive. But at most half of them perform value aggregation; we refer to this half of threads as effective. The other half is simply handing off their data.
Procedure Value shuffle: In this step active threads transfer data from higher lane positions in the warp to lower lane positions, creating Remote Reduce list. Value aggregation: In this step, effective threads combine their thread local Reduce list with Remote Reduce list and store the result in the thread local Reduce list. Value copy: In this step, we deal with the assumption made by algorithm 2 (i.e. contiguity assumption). When we have an odd number of lanes active, say 2k+1, only k threads will be effective and therefore k new values will be produced. However, the Reduce list owned by the (2k+1)th thread is ignored in the value aggregation. Therefore we copy the Reduce list from the (2k+1)th lane to (k+1)th lane so that the contiguity assumption still holds.
Definition at line 3462 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenTypes::arrangeBuiltinFunctionDeclaration(), clang::CodeGen::CodeGenFunction::Builder, clang::CodeGen::CodeGenFunction::ConvertTypeForMem(), clang::Create(), clang::CodeGen::CodeGenFunction::createBasicBlock(), clang::CodeGen::CodeGenFunction::CreateMemTemp(), clang::CodeGen::CGBuilderTy::CreatePointerBitCastOrAddrSpaceCast(), clang::CodeGen::CodeGenFunction::EmitBlock(), clang::CodeGen::CodeGenFunction::EmitLoadOfScalar(), emitReductionListCopy(), clang::CodeGen::CodeGenFunction::FinishFunction(), clang::CodeGen::CodeGenFunction::GetAddrOfLocalVar(), clang::CodeGen::CodeGenModule::getContext(), clang::CodeGen::CodeGenTypes::GetFunctionType(), clang::CodeGen::CodeGenModule::getLangOpts(), clang::CodeGen::CodeGenModule::getModule(), clang::CodeGen::CodeGenModule::getOpenMPRuntime(), clang::CodeGen::Address::getPointer(), clang::CodeGen::CodeGenTypeCache::getPointerAlign(), clang::CodeGen::CodeGenModule::getTypes(), clang::InternalLinkage, clang::ImplicitParamDecl::Other, clang::CodeGen::CodeGenModule::SetInternalFunctionAttributes(), clang::ASTContext::ShortTy, clang::CodeGen::CodeGenFunction::StartFunction(), clang::CodeGen::CodeGenTypeCache::VoidPtrTy, clang::ASTContext::VoidPtrTy, and clang::ASTContext::VoidTy.
|
static |
Definition at line 4933 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::TargetOptions::CPU, clang::TargetOptions::Features, clang::CodeGen::CodeGenModule::getDiags(), clang::CodeGen::CodeGenModule::getTarget(), clang::TargetInfo::getTargetOpts(), clang::TargetInfo::hasFeature(), clang::TargetInfo::initFeatureMap(), clang::StringToCudaArch(), and clang::UNKNOWN.
Referenced by clang::CodeGen::CGOpenMPRuntimeNVPTX::checkArchForUnifiedAddressing(), and getSMsBlocksPerSM().
|
static |
Definition at line 717 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CGOpenMPRuntimeNVPTX::CUDA, clang::CodeGen::CGOpenMPRuntimeNVPTX::Generic, and clang::CodeGen::CodeGenModule::getLangOpts().
Referenced by clang::CodeGen::CGOpenMPRuntimeNVPTX::emitFunctionProlog(), clang::CodeGen::CGOpenMPRuntimeNVPTX::emitTeamsOutlinedFunction(), and clang::CodeGen::CGOpenMPRuntimeNVPTX::getAddressOfLocalVariable().
|
static |
Get list of lastprivate variables from the teams distribute ...
or teams {distribute ...} directives.
Definition at line 2001 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::OMPExecutableDirective::getDirectiveKind(), clang::isOpenMPDistributeDirective(), and clang::isOpenMPTeamsDirective().
Referenced by clang::CodeGen::CGOpenMPRuntimeNVPTX::emitTeamsOutlinedFunction().
|
static |
Get the thread id of the OMP master thread.
The master thread id is the first thread (lane) of the last warp in the GPU block. Warp size is assumed to be some power of 2. Thread id is 0 indexed. E.g: If NumThreads is 33, master id is 32. If NumThreads is 64, master id is 32. If NumThreads is 1024, master id is 992.
Definition at line 682 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenTypes::arrangeNullaryFunction(), clang::CodeGen::CodeGenFunction::Builder, clang::Create(), clang::CodeGen::CodeGenTypes::GetFunctionType(), clang::CodeGen::CodeGenModule::getModule(), getNVPTXNumThreads(), getNVPTXWarpSize(), clang::CodeGen::CodeGenModule::getTypes(), clang::InternalLinkage, and clang::CodeGen::CodeGenModule::SetInternalFunctionAttributes().
Referenced by supportsLightweightRuntime().
|
static |
Get the id of the current lane in the Warp.
We assume that the warp size is 32, which is always the case on the NVPTX device, to generate more efficient code.
Definition at line 647 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenFunction::Builder, and getNVPTXThreadID().
Referenced by emitInterWarpCopyFunction(), and clang::CodeGen::CGOpenMPRuntimeNVPTX::emitTeamsOutlinedFunction().
|
static |
Get the maximum number of threads in a block of the GPU.
Definition at line 654 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenFunction::CGM, clang::CodeGen::CodeGenFunction::EmitRuntimeCall(), and clang::CodeGen::CodeGenModule::getModule().
Referenced by clang::CodeGen::CGOpenMPRuntimeNVPTX::emitCriticalRegion(), getMasterThreadID(), and getThreadLimit().
|
static |
Get the id of the current thread on the GPU.
Definition at line 629 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenFunction::CGM, clang::CodeGen::CodeGenFunction::EmitRuntimeCall(), and clang::CodeGen::CodeGenModule::getModule().
Referenced by clang::CodeGen::CGOpenMPRuntimeNVPTX::emitCriticalRegion(), emitInterWarpCopyFunction(), getNVPTXLaneID(), getNVPTXWarpID(), and supportsLightweightRuntime().
|
static |
Get the id of the warp in the block.
We assume that the warp size is 32, which is always the case on the NVPTX device, to generate more efficient code.
Definition at line 639 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenFunction::Builder, and getNVPTXThreadID().
Referenced by emitInterWarpCopyFunction().
|
static |
Get the GPU warp size.
Definition at line 621 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenFunction::CGM, clang::CodeGen::CodeGenFunction::EmitRuntimeCall(), and clang::CodeGen::CodeGenModule::getModule().
Referenced by createRuntimeShuffleFunction(), getMasterThreadID(), and getThreadLimit().
|
static |
Get number of SMs and number of blocks per SM.
Definition at line 5006 of file CGOpenMPRuntimeNVPTX.cpp.
References getCudaArch(), clang::CodeGen::CodeGenModule::getLangOpts(), clang::GFX1010, clang::GFX1011, clang::GFX1012, clang::GFX600, clang::GFX601, clang::GFX700, clang::GFX701, clang::GFX702, clang::GFX703, clang::GFX704, clang::GFX801, clang::GFX802, clang::GFX803, clang::GFX810, clang::GFX900, clang::GFX902, clang::GFX904, clang::GFX906, clang::GFX908, clang::GFX909, clang::LAST, clang::SM_20, clang::SM_21, clang::SM_30, clang::SM_32, clang::SM_35, clang::SM_37, clang::SM_50, clang::SM_52, clang::SM_53, clang::SM_60, clang::SM_61, clang::SM_62, clang::SM_70, clang::SM_72, clang::SM_75, and clang::UNKNOWN.
Referenced by clang::CodeGen::CGOpenMPRuntimeNVPTX::clear().
|
static |
Get list of reduction variables from the teams ... directives.
Definition at line 2026 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::OMPExecutableDirective::getClausesOfKind(), clang::OMPExecutableDirective::getDirectiveKind(), getPrivateItem(), and clang::isOpenMPTeamsDirective().
Referenced by clang::CodeGen::CGOpenMPRuntimeNVPTX::emitTeamsOutlinedFunction().
|
static |
Get the value of the thread_limit clause in the teams directive.
For the 'generic' execution mode, the runtime encodes thread_limit in the launch parameters, always starting thread_limit+warpSize threads per CTA. The threads in the last warp are reserved for master execution. For the 'spmd' execution mode, all threads in a CTA are part of the team.
Definition at line 666 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenFunction::Builder, getNVPTXNumThreads(), and getNVPTXWarpSize().
Referenced by supportsLightweightRuntime().
|
static |
Check for inner (nested) lightweight runtime construct, if any.
Definition at line 906 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CapturedStmt::getCapturedStmt(), clang::OMPExecutableDirective::getDirectiveKind(), clang::OMPExecutableDirective::getInnermostCapturedStmt(), clang::CodeGen::getSingleCompoundChild(), hasStaticScheduling(), clang::Stmt::IgnoreContainers(), clang::isOpenMPLoopDirective(), clang::isOpenMPParallelDirective(), clang::isOpenMPWorksharingDirective(), clang::OMPD_unknown, and supportsSPMDExecutionMode().
Referenced by supportsLightweightRuntime().
|
static |
Check for inner (nested) SPMD construct, if any.
Definition at line 723 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CapturedStmt::getCapturedStmt(), clang::OMPExecutableDirective::getDirectiveKind(), clang::OMPExecutableDirective::getInnermostCapturedStmt(), clang::CodeGen::getSingleCompoundChild(), clang::Stmt::IgnoreContainers(), clang::isOpenMPParallelDirective(), and clang::OMPD_unknown.
Referenced by supportsSPMDExecutionMode().
|
static |
Check if the directive is loops based and has schedule clause at all or has static scheduling.
Definition at line 893 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::OMPExecutableDirective::getClausesOfKind(), clang::OMPExecutableDirective::getDirectiveKind(), clang::OMPExecutableDirective::hasClausesOfKind(), clang::isOpenMPLoopDirective(), and clang::isOpenMPWorksharingDirective().
Referenced by hasNestedLightweightDirective(), and supportsLightweightRuntime().
|
static |
Definition at line 1381 of file CGOpenMPRuntimeNVPTX.cpp.
Referenced by clang::CodeGen::CGOpenMPRuntimeNVPTX::createNVPTXRuntimeFunction().
|
static |
Definition at line 2869 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenFunction::Builder, clang::CodeGen::CodeGenFunction::ConvertTypeForMem(), clang::CodeGen::CodeGenFunction::createBasicBlock(), clang::CodeGen::CGBuilderTy::CreateConstGEP(), clang::CodeGen::CGBuilderTy::CreatePointerBitCastOrAddrSpaceCast(), createRuntimeShuffleFunction(), clang::CodeGen::CodeGenFunction::EmitBlock(), clang::CodeGen::CodeGenFunction::EmitBranch(), clang::CodeGen::CodeGenFunction::EmitLoadOfScalar(), clang::CodeGen::CodeGenFunction::EmitStoreOfScalar(), clang::CharUnits::fromQuantity(), clang::CodeGen::Address::getAlignment(), clang::CodeGen::CodeGenFunction::getContext(), clang::ASTContext::getIntTypeForBitwidth(), clang::CodeGen::Address::getPointer(), clang::CharUnits::getQuantity(), clang::CodeGen::Address::getType(), clang::ASTContext::getTypeSizeInChars(), Offset, clang::ASTContext::toBits(), clang::CodeGen::Type, and clang::CodeGen::CodeGenTypeCache::VoidPtrTy.
Referenced by emitReductionListCopy().
|
static |
Checks if the construct supports lightweight runtime.
It must be SPMD construct + inner loop-based construct with static scheduling.
Definition at line 1061 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::CodeGen::CodeGenFunction::Builder, clang::CodeGen::CodeGenFunction::CGM, clang::CodeGen::CGOpenMPRuntime::clearLocThreadIdInsertPt(), clang::CodeGen::CodeGenFunction::createBasicBlock(), clang::cuda_shared, clang::CodeGen::CodeGenFunction::EmitBlock(), clang::CodeGen::CodeGenFunction::EmitBranch(), clang::CodeGen::CodeGenFunction::EmitRuntimeCall(), clang::OMPExecutableDirective::getBeginLoc(), clang::CodeGen::CodeGenModule::getContext(), clang::OMPExecutableDirective::getDirectiveKind(), clang::CodeGen::CodeGenModule::getLangOpts(), getMasterThreadID(), clang::CodeGen::CodeGenModule::getModule(), getNVPTXThreadID(), clang::CodeGen::CodeGenModule::getOpenMPRuntime(), clang::ASTContext::getTargetAddressSpace(), getThreadLimit(), hasNestedLightweightDirective(), hasStaticScheduling(), clang::CodeGen::CodeGenFunction::HaveInsertPoint(), clang::InternalLinkage, clang::OMPD_unknown, clang::CodeGen::RegionCodeGenTy::setAction(), clang::CodeGen::CGOpenMPRuntime::setLocThreadIdInsertPt(), supportsSPMDExecutionMode(), and clang::CodeGen::CodeGenTypeCache::VoidPtrTy.
|
static |
Definition at line 818 of file CGOpenMPRuntimeNVPTX.cpp.
References clang::OMPExecutableDirective::getDirectiveKind(), hasNestedSPMDDirective(), and clang::OMPD_unknown.
Referenced by clang::CodeGen::CGOpenMPRuntimeNVPTX::createNVPTXRuntimeFunction(), hasNestedLightweightDirective(), and supportsLightweightRuntime().