skepu::Device_CU Class Reference

A class representing a CUDA device. More...

#include <device_cu.h>

Public Member Functions
	Device_CU (unsigned int id)
	~Device_CU ()
	The destructor.
bool	isOverlapSupported ()
size_t	getMaxBlockSize () const
int	getMajorVersion () const
int	getMinorVersion () const
unsigned int	getSmPerMultiProc () const
std::string	getDeviceName () const
int	getClockRate () const
int	getAsyncEngineCount () const
bool	IsConcurrentKernels () const
unsigned int	getNoConcurrentKernels () const
int	getNumComputeUnits () const
size_t	getGlobalMemSize () const
size_t	getSharedMemPerBlock () const
size_t	getMaxThreads () const
size_t	getMaxBlocks () const
unsigned int	getDeviceID () const

Detailed Description

A class representing a CUDA device.

This class represents one CUDA device which can be used to execute the skeletons on if CUDA is used as backend. Stores various properties about the device and provides functions that return them.

Constructor & Destructor Documentation

skepu::Device_CU::Device_CU ( unsigned int  id )

inline

The constructor creates a device from an ID and gets all its properties.

Parameters

id	Device ID for the device that is to be created.

Member Function Documentation

int skepu::Device_CU::getAsyncEngineCount ( ) const

inline

Returns

The integer specifying whether the overlap is support between pinned memory transfer and kernel launches (value>0 if supported) and/or between pinned memory HTD and DTH (value=2 if supported).

int skepu::Device_CU::getClockRate ( ) const

inline

test

Returns

The clock rate of current GPU.

unsigned int skepu::Device_CU::getDeviceID ( ) const

inline

Returns

Device ID.

Referenced by skepu::DeviceMemPointer_CU< T >::DeviceMemPointer_CU(), skepu::DeviceMemPointer_Matrix_CU< T >::DeviceMemPointer_Matrix_CU(), skepu::scanLargeVectorRecursivelyM_CU(), and skepu::Matrix< T >::transpose_CU().

std::string skepu::Device_CU::getDeviceName ( ) const

inline

Returns

The name of current GPU.

size_t skepu::Device_CU::getGlobalMemSize ( ) const

inline

Returns

The global memory size.

int skepu::Device_CU::getMajorVersion ( ) const

inline

Returns

The major version.

size_t skepu::Device_CU::getMaxBlocks ( ) const

inline

Returns

The maximum number of blocks or groups for a kernel launch.

Referenced by skepu::scanLargeVectorRecursivelyM_CU().

size_t skepu::Device_CU::getMaxBlockSize ( ) const

inline

Returns

The maximum block size.

size_t skepu::Device_CU::getMaxThreads ( ) const

inline

Returns

The maximum number of threads per block or group.

Referenced by skepu::scanLargeVectorRecursivelyM_CU().

int skepu::Device_CU::getMinorVersion ( ) const

inline

Returns

The minor version.

unsigned int skepu::Device_CU::getNoConcurrentKernels ( ) const

inline

Returns

The integer specifying the max number of concurrent kernels supported on this GPU (if not supprted then return 1).

Referenced by skepu::Reduce< ReduceFunc, ReduceFunc >::CU().

int skepu::Device_CU::getNumComputeUnits ( ) const

inline

Returns

The maximum number of compute units available.

size_t skepu::Device_CU::getSharedMemPerBlock ( ) const

inline

Returns

The shared memory size.

unsigned int skepu::Device_CU::getSmPerMultiProc ( ) const

inline

Returns

The number of streaming processors per multiprocessor for current GPU.

bool skepu::Device_CU::IsConcurrentKernels ( ) const

inline

Returns

The boolean specifying whether the concurrent kernels are supported on this GPU or not. If supported maximum kernels count is 16.

bool skepu::Device_CU::isOverlapSupported ( )

inline

Returns whether the device supports overlap (memory,kernel) operation or not

---

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

• include/skepu/src/device_cu.h