namespace to support sse2 vectorization More...

Classes
struct	FVec
	float vector real type, used for vectorization More...

struct	FVec< float >
	vector real type for float More...

struct	FVec< double >
	vector real type for float More...

struct	SSEOp
	sse2 operator type of certain operator More...

struct	SSEOp< op::plus >

struct	SSEOp< op::minus >

struct	SSEOp< op::mul >

struct	SSEOp< op::div >

struct	SSEOp< op::identity >

struct	Saver

struct	Saver< sv::saveto, TFloat >

Functions
void *	AlignedMallocPitch (size_t &pitch, size_t lspace, size_t num_line)
	analog to cudaMallocPitch, allocate a aligned space with num_line * lspace cells More...

void	AlignedFree (void *ptr)
	free aligned space More...

bool	CheckAlign (size_t pitch)
	check if a pointer is aligned

bool	CheckAlign (void *ptr)
	check if a pointer is aligned

index_t	UpperAlign (index_t size, size_t fsize)
	get upper bound of aligned index of size More...

index_t	LowerAlign (index_t size, size_t fsize)
	get lower bound of aligned index of size More...

Detailed Description

namespace to support sse2 vectorization

Function Documentation

void mshadow::sse2::AlignedFree ( void * ptr )

inline

free aligned space

Parameters

ptr	pointer to space to be freed

void* mshadow::sse2::AlignedMallocPitch	(	size_t &	pitch,
		size_t	lspace,
		size_t	num_line
	)

inline

analog to cudaMallocPitch, allocate a aligned space with num_line * lspace cells

Parameters

pitch	output parameter, the actuall space allocated for each line
lspace	number of cells required for each line
num_line	number of lines to be allocated

index_t mshadow::sse2::LowerAlign	(	index_t	size,
		size_t	fsize
	)

inline

get lower bound of aligned index of size

Parameters

size	size of the array
fsize	size of float

index_t mshadow::sse2::UpperAlign	(	index_t	size,
		size_t	fsize
	)

inline

get upper bound of aligned index of size

Parameters

size	size of the array
fsize	size of float