CUDPPSparseMatrixVectorMultiplyPlan Class Reference

Plan class for sparse-matrix dense-vector multiply. More...

#include <cudpp_plan.h>

Inherits CUDPPPlan.

List of all members.

Public Member Functions
	CUDPPSparseMatrixVectorMultiplyPlan (CUDPPConfiguration config, size_t numNZElts, const void *A, const unsigned int *rowindx, const unsigned int *indx, size_t numRows)
	SparseMatrixVectorMultiply Plan constructor.
virtual	~CUDPPSparseMatrixVectorMultiplyPlan ()
	Sparse matrix-vector plan destructor.
Public Attributes
CUDPPSegmentedScanPlan *	m_segmentedScanPlan
void *	m_d_prod
unsigned int *	m_d_flags
unsigned int *	m_d_rowFinalIndex
unsigned int *	m_d_rowIndex
unsigned int *	m_d_index
void *	m_d_A
unsigned int *	m_rowFinalIndex
size_t	m_numRows
	Number of rows.
size_t	m_numNonZeroElements
	Number of non-zero elements.

---

Detailed Description

Plan class for sparse-matrix dense-vector multiply.

---

Constructor & Destructor Documentation

CUDPPSparseMatrixVectorMultiplyPlan::CUDPPSparseMatrixVectorMultiplyPlan	(	CUDPPConfiguration	config,
		size_t	numNonZeroElements,
		const void *	A,
		const unsigned int *	rowIndex,
		const unsigned int *	index,
		size_t	numRows
	)

SparseMatrixVectorMultiply Plan constructor.

Parameters:

[in]	config	The configuration struct specifying options
[in]	numNonZeroElements	The number of non-zero elements in sparse matrix
[in]	A	Array of non-zero matrix elements
[in]	rowIndex	Array of indices of the first element of each row in the "flattened" version of the sparse matrix
[in]	index	Array of indices of non-zero elements in the matrix
[in]	numRows	The number of rows in the sparse matrix

---

Member Data Documentation

CUDPPSegmentedScanPlan* CUDPPSparseMatrixVectorMultiplyPlan::m_segmentedScanPlan

Performs a segmented scan of type T using this plan

void* CUDPPSparseMatrixVectorMultiplyPlan::m_d_prod

Vector of products (of an element in A and its corresponding (thats is belongs to the same row) element in x; this is the input and output of segmented scan

unsigned int* CUDPPSparseMatrixVectorMultiplyPlan::m_d_flags

Vector of flags where a flag is set if an element of A is the first element of its row; this is the flags vector for segmented scan

unsigned int* CUDPPSparseMatrixVectorMultiplyPlan::m_d_rowFinalIndex

Vector of row end indices, which for each row specifies an index in A which is the last element of that row. Resides in GPU memory.

unsigned int* CUDPPSparseMatrixVectorMultiplyPlan::m_d_rowIndex

Vector of row end indices, which for each row specifies an index in A which is the first element of that row. Resides in GPU memory.

unsigned int* CUDPPSparseMatrixVectorMultiplyPlan::m_d_index

Vector of column numbers one for each element in A

void* CUDPPSparseMatrixVectorMultiplyPlan::m_d_A

The A matrix

unsigned int* CUDPPSparseMatrixVectorMultiplyPlan::m_rowFinalIndex

Vector of row end indices, which for each row specifies an index in A which is the last element of that row. Resides in CPU memory.

---

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

• cudpp_plan.h
• cudpp_plan.cpp