Deploy all tasks to GPU. More...

#include <AllOnGPU.h>

Inheritance diagram for tarch::multicore::orchestration::AllOnGPU:

Collaboration diagram for tarch::multicore::orchestration::AllOnGPU:

Public Member Functions
	AllOnGPU (int device, FusionImplementation FusionImplementation=FusionImplementation::LazyEvaluation)
	Initialise strategy.

virtual void	startBSPSection (int nestedParallelismLevel) override
	Notifies the strategy that we enter a BSP section.

virtual void	endBSPSection (int nestedParallelismLevel) override
	Notifies the strategy that we leave a BSP (fork-join) section.

virtual FuseInstruction	fuse (int taskType) override
	Return how many tasks to fuse at least, at most and to which device to deploy them.

virtual ExecutionPolicy	paralleliseForkJoinSection (int nestedParallelismLevel, int numberOfTasks, int taskType) override
	No if we encounter nested parallelism.

virtual void	updateFuseStatistics (int taskType, int queueSize, double averageNumberOfPendingReadyTasksPerProcessTaskExecution, double probabilityThatTaskIsCompleted) override
	Update task fusion status.

virtual FusionImplementation	getFusionImplementation () const override
	Hardcoded to tarch::multicore::orchestration::FusionImplementation::LazyEvaluation.

Public Member Functions inherited from tarch::multicore::orchestration::Strategy
virtual	~Strategy ()=default

Static Public Attributes
static int constexpr	RoundRobin = -1

Static Public Attributes inherited from tarch::multicore::orchestration::Strategy
static constexpr int	EndOfBSPSection = -1

Private Member Functions
void	toggleDevice ()

Private Attributes
const int	_device

int	_deviceToUse

bool	_isInBspSection

const FusionImplementation	_fusionImplementation

Additional Inherited Members
Public Types inherited from tarch::multicore::orchestration::Strategy
enum class	ExecutionPolicy { RunSerially , RunParallel }
	Provide hint of execution policy. More...

Detailed Description

Deploy all tasks to GPU.

This is a very simple orchestration which tries to get all tasks over to a GPU. For this, it relies on tarch::multicore::orchestration::FusionImplementation::LazyEvaluation as orchestration strategy. The class puts no restrictions on the number of tasks that it can fuse. Depending on how the object is initialised, all tasks either go on one GPU, or the routine runs round robin through all accelerators.

Definition at line 26 of file AllOnGPU.h.

Constructor & Destructor Documentation

◆ AllOnGPU()

tarch::multicore::orchestration::AllOnGPU::AllOnGPU	(	int	device,
		FusionImplementation	FusionImplementation = `FusionImplementation::LazyEvaluation`
	)

Initialise strategy.

Parameters

device Either a nonnegative integer picking one device or RoundRobin.

Member Function Documentation

◆ endBSPSection()

virtual void tarch::multicore::orchestration::AllOnGPU::endBSPSection ( int nestedParallelismLevel )

overridevirtual

Notifies the strategy that we leave a BSP (fork-join) section.

Implements tarch::multicore::orchestration::Strategy.

◆ fuse()

virtual FuseInstruction tarch::multicore::orchestration::AllOnGPU::fuse ( int taskType )

overridevirtual

Return how many tasks to fuse at least, at most and to which device to deploy them.

Returns: (-1,std::numeric_limits<int>::max(),0) within the BSP section, such that the lower limit is never exceeded. Return the device otherwise with a min size of 1.

Implements tarch::multicore::orchestration::Strategy.

◆ getFusionImplementation()

virtual FusionImplementation tarch::multicore::orchestration::AllOnGPU::getFusionImplementation ( ) const

overridevirtual

Hardcoded to tarch::multicore::orchestration::FusionImplementation::LazyEvaluation.

Implements tarch::multicore::orchestration::Strategy.

◆ paralleliseForkJoinSection()

virtual ExecutionPolicy tarch::multicore::orchestration::AllOnGPU::paralleliseForkJoinSection	(	int	nestedParallelismLevel,
		int	numberOfTasks,
		int	taskType
	)

overridevirtual

No if we encounter nested parallelism.

Implements tarch::multicore::orchestration::Strategy.

◆ startBSPSection()

virtual void tarch::multicore::orchestration::AllOnGPU::startBSPSection ( int nestedParallelismLevel )

overridevirtual

Notifies the strategy that we enter a BSP section.

Implements tarch::multicore::orchestration::Strategy.

◆ toggleDevice()

void tarch::multicore::orchestration::AllOnGPU::toggleDevice ( )

private

◆ updateFuseStatistics()

virtual void tarch::multicore::orchestration::AllOnGPU::updateFuseStatistics	(	int	taskType,
		int	queueSize,
		double	averageNumberOfReadyTasksPerProcessTaskExecution,
		double	probabilityThatTaskIsCompleted
	)

overridevirtual

Update task fusion status.

Feed orchestration with some stats around the task queue of interest. This routine is not called every time a process read tasks kicks off, but only every now and then. Furthermore, please note that this routine is invoked iff the orchestration beforehand had decided to fuse tasks.

Parameters

taskType	Number of the task that we are studying
queueSize	Size of the queue at this observation point
averageNumberOfReadyTasksPerProcessTaskExecution	average over number of tasks in the queue whenever a process task has kicked off and the underlying real task has not completed yet.
probabilityThatTaskIsCompleted	Probability that task result has been there already when the corresponding instance of ProcessReadyTask fires up. This is a biased value: If we try to look up if a task is already completed and the underlying dictionary is locked, then this counts as a "no hit" as well.

Implements tarch::multicore::orchestration::Strategy.

Field Documentation

◆ _device

const int tarch::multicore::orchestration::AllOnGPU::_device

private

Definition at line 28 of file AllOnGPU.h.

◆ _deviceToUse

int tarch::multicore::orchestration::AllOnGPU::_deviceToUse

private

Definition at line 29 of file AllOnGPU.h.

◆ _fusionImplementation

const FusionImplementation tarch::multicore::orchestration::AllOnGPU::_fusionImplementation

private

Definition at line 31 of file AllOnGPU.h.

◆ _isInBspSection

bool tarch::multicore::orchestration::AllOnGPU::_isInBspSection

private

Definition at line 30 of file AllOnGPU.h.

◆ RoundRobin

int constexpr tarch::multicore::orchestration::AllOnGPU::RoundRobin = -1

staticconstexpr

Definition at line 35 of file AllOnGPU.h.

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

src/tarch/multicore/orchestration/AllOnGPU.h

Public Member Functions

Static Public Attributes

Private Member Functions

Private Attributes

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ AllOnGPU()

Member Function Documentation

◆ endBSPSection()

◆ fuse()

◆ getFusionImplementation()

◆ paralleliseForkJoinSection()

◆ startBSPSection()

◆ toggleDevice()

◆ updateFuseStatistics()

Field Documentation

◆ _device

◆ _deviceToUse

◆ _fusionImplementation

◆ _isInBspSection

◆ RoundRobin