This ParallelLoopGenerator subclass handles the generation of parallelized code, utilizing the LLVM OpenMP library. More...

#include <LoopGeneratorsKMP.h>

Inheritance diagram for polly::ParallelLoopGeneratorKMP:

[legend]

Public Member Functions
	ParallelLoopGeneratorKMP (PollyIRBuilder &Builder, const DataLayout &DL)
	Create a parallel loop generator for the current function.

void	createCallSpawnThreads (Value SubFn, Value SubFnParam, Value LB, Value UB, Value *Stride)
	Create a runtime library call to spawn the worker threads.

void	deployParallelExecution (Function SubFn, Value SubFnParam, Value LB, Value UB, Value *Stride) override
	Create the runtime library calls for spawn and join of the worker threads.

Function *	prepareSubFnDefinition (Function *F) const override
	Prepare the definition of the parallel subfunction.

std::tuple< Value , Function >	createSubFn (Value Stride, AllocaInst Struct, SetVector< Value * > UsedValues, ValueMapT &VMap) override
	Create the parallel subfunction.

Value *	createCallGlobalThreadNum ()
	Create a runtime library call to get the current global thread number.

void	createCallPushNumThreads (Value GlobalThreadID, Value NumThreads)
	Create a runtime library call to request a number of threads.

void	createCallDispatchInit (Value GlobalThreadID, Value LB, Value UB, Value Inc, Value *ChunkSize)
	Create a runtime library call to prepare the OpenMP runtime.

Value *	createCallDispatchNext (Value GlobalThreadID, Value IsLastPtr, Value LBPtr, Value UBPtr, Value *StridePtr)
	Create a runtime library call to retrieve the next (dynamically) allocated chunk of work for this thread.

void	createCallStaticInit (Value GlobalThreadID, Value IsLastPtr, Value LBPtr, Value UBPtr, Value StridePtr, Value ChunkSize)
	Create a runtime library call to prepare the OpenMP runtime.

void	createCallStaticFini (Value *GlobalThreadID)
	Create a runtime library call to mark the end of a statically scheduled loop.

GlobalVariable *	createSourceLocation ()
	Create the current source location.

Public Member Functions inherited from polly::ParallelLoopGenerator
	ParallelLoopGenerator (PollyIRBuilder &Builder, const DataLayout &DL)
	Create a parallel loop generator for the current function.

virtual	~ParallelLoopGenerator ()

Value *	createParallelLoop (Value LB, Value UB, Value Stride, SetVector< Value > &Values, ValueMapT &VMap, BasicBlock::iterator *LoopBody)
	Create a parallel loop.

DominatorTree *	getCalleeDominatorTree () const
	Returns the DominatorTree for the generated subfunction.

LoopInfo *	getCalleeLoopInfo () const
	Returns the LoopInfo for the generated subfunction.

AllocaInst *	storeValuesIntoStruct (SetVector< Value * > &Values)
	Create a struct for all `Values` and store them in there.

void	extractValuesFromStruct (SetVector< Value * > Values, Type Ty, Value Struct, ValueMapT &VMap)
	Extract all values from the `Struct` and construct the mapping.

Function *	createSubFnDefinition ()
	Create the definition of the parallel subfunction.

virtual void	deployParallelExecution (Function SubFn, Value SubFnParam, Value LB, Value UB, Value *Stride)=0
	Create the runtime library calls for spawn and join of the worker threads.

virtual Function *	prepareSubFnDefinition (Function *F) const =0
	Prepare the definition of the parallel subfunction.

virtual std::tuple< Value , Function >	createSubFn (Value Stride, AllocaInst Struct, SetVector< Value * > UsedValues, ValueMapT &VMap)=0
	Create the parallel subfunction.

Protected Member Functions
OMPGeneralSchedulingType	getSchedType (int ChunkSize, OMPGeneralSchedulingType Scheduling) const
	Convert the combination of given chunk size and scheduling type (which might have been set via the command line) into the corresponding scheduling type.

bool	is64BitArch ()
	Returns True if 'LongType' is 64bit wide, otherwise: False.

Protected Attributes
GlobalValue *	SourceLocationInfo
	The source location struct of this loop.

Protected Attributes inherited from polly::ParallelLoopGenerator
PollyIRBuilder &	Builder
	The IR builder we use to create instructions.

std::unique_ptr< LoopInfo >	SubFnLI
	The loop info for the generated subfunction.

std::unique_ptr< DominatorTree >	SubFnDT
	The dominance tree for the generated subfunction.

Type *	LongType
	The type of a "long" on this hardware used for backend calls.

Module *	M
	The current module.

llvm::DebugLoc	DLGenerated
	Debug location for generated code without direct link to any specific line.

Detailed Description

This ParallelLoopGenerator subclass handles the generation of parallelized code, utilizing the LLVM OpenMP library.

Definition at line 27 of file LoopGeneratorsKMP.h.

Constructor & Destructor Documentation

◆ ParallelLoopGeneratorKMP()

polly::ParallelLoopGeneratorKMP::ParallelLoopGeneratorKMP	(	PollyIRBuilder &	Builder,
		const DataLayout &	DL
	)

inline

Create a parallel loop generator for the current function.

Definition at line 30 of file LoopGeneratorsKMP.h.

References createSourceLocation(), and SourceLocationInfo.

Member Function Documentation

◆ createCallDispatchInit()

void ParallelLoopGeneratorKMP::createCallDispatchInit	(	Value *	GlobalThreadID,
		Value *	LB,
		Value *	UB,
		Value *	Inc,
		Value *	ChunkSize
	)

Create a runtime library call to prepare the OpenMP runtime.

For dynamically scheduled loops, saving the loop arguments.

Parameters

GlobalThreadID	The global thread ID.
LB	The loop's lower bound.
UB	The loop's upper bound.
Inc	The loop increment.
ChunkSize	The chunk size of the parallel loop.

Definition at line 428 of file LoopGeneratorsKMP.cpp.

References polly::ParallelLoopGenerator::Builder, polly::ParallelLoopGenerator::DLGenerated, Function, getSchedType(), is64BitArch(), polly::ParallelLoopGenerator::LongType, polly::ParallelLoopGenerator::M, polly::PollyChunkSize, polly::PollyScheduling, SourceLocationInfo, and polly::Value.

Referenced by createSubFn().

◆ createCallDispatchNext()

Value * ParallelLoopGeneratorKMP::createCallDispatchNext	(	Value *	GlobalThreadID,
		Value *	IsLastPtr,
		Value *	LBPtr,
		Value *	UBPtr,
		Value *	StridePtr
	)

Create a runtime library call to retrieve the next (dynamically) allocated chunk of work for this thread.

Parameters

GlobalThreadID	The global thread ID.
IsLastPtr	Pointer to a flag, which is set to 1 if this is the last chunk of work, or 0 otherwise.
LBPtr	Pointer to the lower bound for the next chunk.
UBPtr	Pointer to the upper bound for the next chunk.
StridePtr	Pointer to the stride for the next chunk.

Returns: A Value which holds 1 if there is work to be done, 0 otherwise.

Definition at line 469 of file LoopGeneratorsKMP.cpp.

References polly::ParallelLoopGenerator::Builder, polly::ParallelLoopGenerator::DLGenerated, Function, is64BitArch(), polly::ParallelLoopGenerator::LongType, polly::ParallelLoopGenerator::M, SourceLocationInfo, and polly::Value.

Referenced by createSubFn().

◆ createCallGlobalThreadNum()

Value * ParallelLoopGeneratorKMP::createCallGlobalThreadNum ( )

Create a runtime library call to get the current global thread number.

Returns: A Value ref which holds the current global thread number.

Definition at line 317 of file LoopGeneratorsKMP.cpp.

References polly::ParallelLoopGenerator::Builder, polly::ParallelLoopGenerator::DLGenerated, Function, polly::ParallelLoopGenerator::M, SourceLocationInfo, and polly::Value.

Referenced by deployParallelExecution().

◆ createCallPushNumThreads()

void ParallelLoopGeneratorKMP::createCallPushNumThreads	(	Value *	GlobalThreadID,
		Value *	NumThreads
	)

Create a runtime library call to request a number of threads.

Which will be used in the next OpenMP section (by the next fork).

Parameters

GlobalThreadID	The global thread ID.
NumThreads	The number of threads to use.

Definition at line 338 of file LoopGeneratorsKMP.cpp.

References polly::ParallelLoopGenerator::Builder, polly::ParallelLoopGenerator::DLGenerated, Function, polly::ParallelLoopGenerator::M, SourceLocationInfo, and polly::Value.

Referenced by deployParallelExecution().

◆ createCallSpawnThreads()

void ParallelLoopGeneratorKMP::createCallSpawnThreads	(	Value *	SubFn,
		Value *	SubFnParam,
		Value *	LB,
		Value *	UB,
		Value *	Stride
	)

Create a runtime library call to spawn the worker threads.

Parameters

SubFn	The subfunction which holds the loop body.
SubFnParam	The parameter for the subfunction (basically the struct filled with the outside values).
LB	The lower bound for the loop we parallelize.
UB	The upper bound for the loop we parallelize.
Stride	The stride of the loop we parallelize.

Definition at line 21 of file LoopGeneratorsKMP.cpp.

References polly::ParallelLoopGenerator::Builder, polly::ParallelLoopGenerator::DLGenerated, Function, polly::ParallelLoopGenerator::M, SourceLocationInfo, and polly::Value.

Referenced by deployParallelExecution().

◆ createCallStaticFini()

void ParallelLoopGeneratorKMP::createCallStaticFini ( Value * GlobalThreadID )

Create a runtime library call to mark the end of a statically scheduled loop.

Parameters

GlobalThreadID The global thread ID.

Definition at line 408 of file LoopGeneratorsKMP.cpp.

References polly::ParallelLoopGenerator::Builder, polly::ParallelLoopGenerator::DLGenerated, Function, polly::ParallelLoopGenerator::M, SourceLocationInfo, and polly::Value.

Referenced by createSubFn().

◆ createCallStaticInit()

void ParallelLoopGeneratorKMP::createCallStaticInit	(	Value *	GlobalThreadID,
		Value *	IsLastPtr,
		Value *	LBPtr,
		Value *	UBPtr,
		Value *	StridePtr,
		Value *	ChunkSize
	)

Create a runtime library call to prepare the OpenMP runtime.

For statically scheduled loops, saving the loop arguments.

Parameters

GlobalThreadID	The global thread ID.
IsLastPtr	Pointer to a flag, which is set to 1 if this is the last chunk of work, or 0 otherwise.
LBPtr	Pointer to the lower bound for the next chunk.
UBPtr	Pointer to the upper bound for the next chunk.
StridePtr	Pointer to the stride for the next chunk.
ChunkSize	The chunk size of the parallel loop.

Definition at line 362 of file LoopGeneratorsKMP.cpp.

References polly::ParallelLoopGenerator::Builder, polly::ParallelLoopGenerator::DLGenerated, Function, getSchedType(), is64BitArch(), polly::ParallelLoopGenerator::LongType, polly::ParallelLoopGenerator::M, polly::PollyChunkSize, polly::PollyScheduling, SourceLocationInfo, and polly::Value.

Referenced by createSubFn().

◆ createSourceLocation()

GlobalVariable * ParallelLoopGeneratorKMP::createSourceLocation ( )

Create the current source location.

TODO: Generates only(!) dummy values.

Definition at line 505 of file LoopGeneratorsKMP.cpp.

References polly::ParallelLoopGenerator::Builder, and polly::ParallelLoopGenerator::M.

Referenced by ParallelLoopGeneratorKMP().

◆ createSubFn()

std::tuple< Value , Function > ParallelLoopGeneratorKMP::createSubFn	(	Value *	Stride,
		AllocaInst *	Struct,
		SetVector< Value * >	UsedValues,
		ValueMapT &	VMap
	)

overridevirtual

Create the parallel subfunction.

Parameters

Stride	The induction variable increment.
Struct	A struct holding all values in `Values`.
Values	A set of LLVM-IR Values that should be available in the new loop body.
VMap	A map to allow outside access to the new versions of the values in `Values`.
SubFn	The newly created subfunction is returned here.

Returns: The newly created induction variable.

Implements polly::ParallelLoopGenerator.

Definition at line 133 of file LoopGeneratorsKMP.cpp.

References polly::ParallelLoopGenerator::Builder, createCallDispatchInit(), createCallDispatchNext(), createCallStaticFini(), createCallStaticInit(), polly::createLoop(), polly::ParallelLoopGenerator::createSubFnDefinition(), polly::Dynamic, polly::ParallelLoopGenerator::extractValuesFromStruct(), Function, getSchedType(), polly::Guided, is64BitArch(), polly::ParallelLoopGenerator::LongType, polly::PollyChunkSize, polly::PollyScheduling, polly::Runtime, polly::StaticChunked, polly::StaticNonChunked, polly::ParallelLoopGenerator::SubFnDT, polly::ParallelLoopGenerator::SubFnLI, and polly::Value.

◆ deployParallelExecution()

void ParallelLoopGeneratorKMP::deployParallelExecution	(	Function *	SubFn,
		Value *	SubFnParam,
		Value *	LB,
		Value *	UB,
		Value *	Stride
	)

overridevirtual

Create the runtime library calls for spawn and join of the worker threads.

Additionally, places a call to the specified subfunction.

Parameters

SubFn	The subfunction which holds the loop body.
SubFnParam	The parameter for the subfunction (basically the struct filled with the outside values).
LB	The lower bound for the loop we parallelize.
UB	The upper bound for the loop we parallelize.
Stride	The stride of the loop we parallelize.

Implements polly::ParallelLoopGenerator.

Definition at line 65 of file LoopGeneratorsKMP.cpp.

References polly::ParallelLoopGenerator::Builder, createCallGlobalThreadNum(), createCallPushNumThreads(), createCallSpawnThreads(), polly::PollyNumThreads, and polly::Value.

◆ getSchedType()

OMPGeneralSchedulingType ParallelLoopGeneratorKMP::getSchedType	(	int	ChunkSize,
		OMPGeneralSchedulingType	Scheduling
	)		const

protected

Convert the combination of given chunk size and scheduling type (which might have been set via the command line) into the corresponding scheduling type.

This may result (e.g.) in a 'change' from "static chunked" scheduling to "static non-chunked" (regarding the provided and returned scheduling types).

Parameters

ChunkSize	The chunk size, set via command line or its default.
Scheduling	The scheduling, set via command line or its default.

Returns: The corresponding OMPGeneralSchedulingType.

Definition at line 561 of file LoopGeneratorsKMP.cpp.

References polly::StaticChunked, and polly::StaticNonChunked.

Referenced by createCallDispatchInit(), createCallStaticInit(), and createSubFn().

◆ is64BitArch()

bool ParallelLoopGeneratorKMP::is64BitArch ( )

protected

Returns True if 'LongType' is 64bit wide, otherwise: False.

Definition at line 557 of file LoopGeneratorsKMP.cpp.

References polly::ParallelLoopGenerator::LongType.

Referenced by createCallDispatchInit(), createCallDispatchNext(), createCallStaticInit(), and createSubFn().

◆ prepareSubFnDefinition()

Function * ParallelLoopGeneratorKMP::prepareSubFnDefinition ( Function * F ) const

overridevirtual

Prepare the definition of the parallel subfunction.

Creates the argument list and names them (as well as the subfunction).

Parameters

F	A pointer to the (parallel) subfunction's parent function.

Returns: The pointer to the (parallel) subfunction.

Implements polly::ParallelLoopGenerator.

Definition at line 79 of file LoopGeneratorsKMP.cpp.

References polly::ParallelLoopGenerator::Builder, Function, polly::ParallelLoopGenerator::LongType, and polly::ParallelLoopGenerator::M.

Member Data Documentation

◆ SourceLocationInfo

GlobalValue* polly::ParallelLoopGeneratorKMP::SourceLocationInfo

protected

The source location struct of this loop.

ident_t = type { i32, i32, i32, i32, i8* }

Definition at line 38 of file LoopGeneratorsKMP.h.

Referenced by createCallDispatchInit(), createCallDispatchNext(), createCallGlobalThreadNum(), createCallPushNumThreads(), createCallSpawnThreads(), createCallStaticFini(), createCallStaticInit(), and ParallelLoopGeneratorKMP().

The documentation for this class was generated from the following files:

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

Constructor & Destructor Documentation

◆ ParallelLoopGeneratorKMP()

Member Function Documentation

◆ createCallDispatchInit()

◆ createCallDispatchNext()

◆ createCallGlobalThreadNum()

◆ createCallPushNumThreads()

◆ createCallSpawnThreads()

◆ createCallStaticFini()

◆ createCallStaticInit()

◆ createSourceLocation()

◆ createSubFn()

◆ deployParallelExecution()

◆ getSchedType()

◆ is64BitArch()

◆ prepareSubFnDefinition()

Member Data Documentation

◆ SourceLocationInfo