TissueForge::cuda::EngineConfig Struct Reference

CUDA runtime control interface for Tissue Forge engine. More...

#include <tfEngineConfig.h>

Public Member Functions
bool	onDevice ()
	Check whether the engine is currently on a device.
int	getDevice ()
	Get the id of the device running the engine.
HRESULT	setDevice (int deviceId=0)
	Set the id of the device for running the engine.
HRESULT	clearDevice ()
	Clear configured device for the engine.
HRESULT	toDevice ()
	Send engine to device. If engine is already on device, then the call is ignored.
HRESULT	fromDevice ()
	Pull engine from device. If engine is not on a device, then the call is ignored.
HRESULT	setBlocks (unsigned int numBlocks, int deviceId=-1)
	Set the number of blocks of the CUDA configuration for a CUDA device.
HRESULT	setThreads (unsigned int numThreads, int deviceId=-1)
	Set the number of threads of the CUDA configuration for a CUDA device.
HRESULT	refreshPotentials ()
	Update potentials on a CUDA device.
HRESULT	refreshFluxes ()
	Update fluxes on a CUDA device.
HRESULT	refreshBoundaryConditions ()
	Update boundary conditions on a CUDA device.
HRESULT	refresh ()
	Update the image of the engine on a CUDA device.

Detailed Description

CUDA runtime control interface for Tissue Forge engine.

This object provides control for configuring engine calculations on CUDA devices. Associated calculations include nonbonded particle interactions and sorting, fluxes and space partitioning.

At any time during a simulation, supported engine calculations can be sent to a particular CUDA device, or brought back to the CPU when deployed on a CUDA device. CUDA dynamic parallelism can also be specified before deploying engine calculations to a CUDA device. Future Tissue Forge versions will support deployment on multiple devices.

Member Function Documentation

clearDevice()

HRESULT TissueForge::cuda::EngineConfig::clearDevice

(

)

Clear configured device for the engine.

Fails if engine is currently on a device.

Returns

HRESULT

fromDevice()

HRESULT TissueForge::cuda::EngineConfig::fromDevice

(

)

Pull engine from device. If engine is not on a device, then the call is ignored.

Returns

HRESULT

getDevice()

int TissueForge::cuda::EngineConfig::getDevice

(

)

Get the id of the device running the engine.

Returns -1 if engine is not on a device.

Returns

int

onDevice()

bool TissueForge::cuda::EngineConfig::onDevice

(

)

Check whether the engine is currently on a device.

Returns

true

false

refresh()

HRESULT TissueForge::cuda::EngineConfig::refresh

(

)

Update the image of the engine on a CUDA device.

Necessary to notify the device of changes to engine data that are not automatically handled by Tissue Forge. Refer to documentation of specific functions and members for which Tissue Forge automatically handles.

If engine is not on a device, then the call is ignored.

Returns

HRESULT

refreshBoundaryConditions()

HRESULT TissueForge::cuda::EngineConfig::refreshBoundaryConditions

(

)

Update boundary conditions on a CUDA device.

Useful for notifying the device that a boundary condition has changed.

If engine is not on a device, then the call is ignored.

Returns

HRESULT

refreshFluxes()

HRESULT TissueForge::cuda::EngineConfig::refreshFluxes

(

)

Update fluxes on a CUDA device.

Useful for notifying the device that a flux has changed.

If engine is not on a device, then the call is ignored.

Returns

HRESULT

refreshPotentials()

HRESULT TissueForge::cuda::EngineConfig::refreshPotentials

(

)

Update potentials on a CUDA device.

Useful for notifying the device that a potential has changed.

If engine is not on a device, then the call is ignored.

Returns

HRESULT

setBlocks()

HRESULT TissueForge::cuda::EngineConfig::setBlocks	(	unsigned int	numBlocks,
		int	deviceId = -1 )

Set the number of blocks of the CUDA configuration for a CUDA device.

Throws an error if called when the engine is already deployed to a CUDA device.

Parameters

numBlocks	number of blocks
deviceId	device ID (optional)

Returns

HRESULT

setDevice()

HRESULT TissueForge::cuda::EngineConfig::setDevice

(

int

deviceId = 0

)

Set the id of the device for running the engine.

Fails if engine is currently on a device.

Parameters

deviceId

Returns

HRESULT

setThreads()

HRESULT TissueForge::cuda::EngineConfig::setThreads	(	unsigned int	numThreads,
		int	deviceId = -1 )

Set the number of threads of the CUDA configuration for a CUDA device.

Throws an error if called when the engine is already deployed to a CUDA device.

Parameters

numThreads	number of threads
deviceId	device ID (optional)

Returns

HRESULT

toDevice()

HRESULT TissueForge::cuda::EngineConfig::toDevice

(

)

Send engine to device. If engine is already on device, then the call is ignored.

Returns

HRESULT

---

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

• source/cuda/tfEngineConfig.h