#include <SpImplNoSR.h>

Static Public Member Functions
static void	SpMXSpV (const Dcsc< IT, bool > &Adcsc, int32_t mA, const int32_t indx, const VT numx, int32_t veclen, int32_t indy, VT numy, int cnts, int dspls, int p_c, BitMap *isthere)
static void	SpMXSpV_ForThreading (const Dcsc< IT, bool > &Adcsc, int32_t mA, const int32_t indx, const VT numx, int32_t veclen, vector< int32_t > &indy, vector< VT > &numy, int32_t offset)

Detailed Description

Dcsc and vector index types do not need to match However, input and output vector numerical types should be identical

Definition at line 80 of file SpImplNoSR.h.

Member Function Documentation

template<typename IT , typename VT >

void SpImplNoSR< IT, bool, VT >::SpMXSpV	(	const Dcsc< IT, bool > &	Adcsc,
		int32_t	mA,
		const int32_t *	indx,
		const VT *	numx,
		int32_t	veclen,
		int32_t *	indy,
		VT *	numy,
		int *	cnts,
		int *	dspls,
		int	p_c,
		BitMap *	isthere
	)

static

Parameters:

[in,out]	indy,numy,cnts	{preallocated arrays to be filled}
[in]	dspls	{displacements to preallocated indy,numy buffers} This version determines the receiving column neighbor and adjust the indices to the receiver's local index It also by passes-SPA by relying on the fact that x (rhs vector) is sorted and values are indices (If they are not sorted, it'd still work but be non-deterministic) Hence, Semiring operations are not needed (no add or multiply) Also allows the vector's indices to be different than matrix's (for transition only) : Disable?

Definition at line 100 of file SpImplNoSR.h.

template<typename IT , typename VT >

void SpImplNoSR< IT, bool, VT >::SpMXSpV_ForThreading	(	const Dcsc< IT, bool > &	Adcsc,
		int32_t	mA,
		const int32_t *	indx,
		const VT *	numx,
		int32_t	veclen,
		vector< int32_t > &	indy,
		vector< VT > &	numy,
		int32_t	offset
	)

static

Definition at line 206 of file SpImplNoSR.h.

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