mxnet::common::cuda Namespace Reference

common utils for cuda More...

Classes
struct	CublasType
	Converts between C++ datatypes and enums/constants needed by cuBLAS. More...
struct	CublasType< double >
struct	CublasType< float >
struct	CublasType< int32_t >
struct	CublasType< mshadow::half::half_t >
struct	CublasType< uint8_t >
class	DeviceStore

Functions
const char *	CublasGetErrorString (cublasStatus_t error)
	Get string representation of cuBLAS errors. More...
const char *	CusolverGetErrorString (cusolverStatus_t error)
	Get string representation of cuSOLVER errors. More...
const char *	CurandGetErrorString (curandStatus_t status)
	Get string representation of cuRAND errors. More...
template<typename DType >
DType __device__	CudaMax (DType a, DType b)
template<typename DType >
DType __device__	CudaMin (DType a, DType b)
int	get_load_type (size_t N)
	Get the largest datatype suitable to read requested number of bytes. More...
int	get_rows_per_block (size_t row_size, int num_threads_per_block)
	Determine how many rows in a 2D matrix should a block of threads handle based on the row size and the number of threads in a block. More...

Detailed Description

common utils for cuda

Function Documentation

CublasGetErrorString()

const char* mxnet::common::cuda::CublasGetErrorString ( cublasStatus_t  error )

inline

Get string representation of cuBLAS errors.

Parameters

error

The error.

Returns

String representation.

CudaMax()

template<typename DType >

DType __device__ mxnet::common::cuda::CudaMax ( DType  a, DType  b  )

inline

CudaMin()

template<typename DType >

DType __device__ mxnet::common::cuda::CudaMin ( DType  a, DType  b  )

inline

CurandGetErrorString()

const char* mxnet::common::cuda::CurandGetErrorString ( curandStatus_t  status )

inline

Get string representation of cuRAND errors.

Parameters

status

The status.

Returns

String representation.

CusolverGetErrorString()

const char* mxnet::common::cuda::CusolverGetErrorString ( cusolverStatus_t  error )

inline

Get string representation of cuSOLVER errors.

Parameters

error

The error.

Returns

String representation.

get_load_type()

int mxnet::common::cuda::get_load_type

(

size_t

N

)

Get the largest datatype suitable to read requested number of bytes.

Number of bytes to be read

Returns

mshadow representation of type that could be used for reading

get_rows_per_block()

int mxnet::common::cuda::get_rows_per_block	(	size_t	row_size,
		int	num_threads_per_block
	)

Determine how many rows in a 2D matrix should a block of threads handle based on the row size and the number of threads in a block.

Parameters

row_size	Size of the row expressed in the number of reads required to fully load it. For example, if the row has N elements, but each thread reads 2 elements with a single read, row_size should be N / 2.
num_threads_per_block	Number of threads in a block.

Returns

the number of rows that should be handled by a single block.