hipsparsesprunedense2csrnnz Interface Reference

hipsparsesprunedense2csrnnz Interface Reference#

HIPFORT API Reference: hipfort_hipsparse::hipsparsesprunedense2csrnnz Interface Reference

hipsparseXpruneDense2csrNnz computes the number of non-zero elements per row and the total number of non-zero elements in a dense matrix after the elements less than the (non-negative) threshold are pruned from the matrix. More...

Public Member Functions
integer(kind(hipsparse_status_success)) function	hipsparsesprunedense2csrnnz_ (handle, m, n, a, lda, threshold, descr, csrrowptr, nnztotaldevhostptr, buffer)

integer(kind(hipsparse_status_success)) function	hipsparsesprunedense2csrnnz_rank_0 (handle, m, n, a, lda, threshold, descr, csrrowptr, nnztotaldevhostptr, buffer)

integer(kind(hipsparse_status_success)) function	hipsparsesprunedense2csrnnz_rank_1 (handle, m, n, a, lda, threshold, descr, csrrowptr, nnztotaldevhostptr, buffer)

integer(kind(hipsparse_status_success)) function	hipsparsesprunedense2csrnnz_full_rank (handle, m, n, a, lda, threshold, descr, csrrowptr, nnztotaldevhostptr, buffer)

Detailed Description

hipsparseXpruneDense2csrNnz computes the number of non-zero elements per row and the total number of non-zero elements in a dense matrix after the elements less than the (non-negative) threshold are pruned from the matrix.

Given an input dense column ordered matrix A, with leading dimension lda where lda>=m, the resulting pruned sparse CSR matrix \(C\) is computed using:

\[ |C(i,j)| = A(i, j) \text{ if |A(i, j)| > threshold} \]

First, the user must determine the size of the required temporary buffer using the routine hipsparseSpruneDense2csr_bufferSize "hipsparseXpruneDense2csr_bufferSize()" and then allocate it. Next, the user allocates csrRowPtr with size m+1. Then the function passes both the temporary storage buffer and csrRowPtr to hipsparseXpruneDense2csrNnz to determine the total number of non-zeros that will exist in the sparse CSR matrix C (after pruning has been performed on A ) as well as fill the output CSR row pointer array csrRowPtr.

For example, given the dense matrix:

\[ \begin{bmatrix} 6 & 2 & 3 & 7 \\% 5 & 6 & 7 & 8 \\% 5 & 4 & 8 & 1 \end{bmatrix} \]

and the threshold value 5, the resulting matrix after pruning is:

\[ \begin{bmatrix} 6 & 0 & 0 & 7 \\% 0 & 6 & 7 & 8 \\% 0 & 0 & 8 & 0 \end{bmatrix} \]

and the corresponding row pointer array and non-zero count:

\[ \begin{align} \text{csrRowPtr} &= \begin{bmatrix} 0 & 2 & 5 & 6 \end{bmatrix} \\% \text{nnzTotalDevHostPtr} &= 6 \end{align} \]

The above example assumes a zero index base for the output CSR matrix. Users can set the desired index base in the output CSR matrix by setting it in the hipsparseMatDescr_t. See hipsparseSetMatIndexBase ().

For a full code example on how to use this routine, see hipsparseSpruneDense2csr().

Note: The routine supports asynchronous execution if the pointer mode is set to device.

Parameters

[in]	handle	- handle to the hipSPARSE library context queue.
[in]	m	- number of rows of the dense matrix `A`.
[in]	n	- number of columns of the dense matrix `A`.
[in]	A	- array of dimensions (`lda`, `n`).
[in]	lda	- leading dimension of the dense array `A`.
[in]	threshold	- pointer to the pruning non-negative threshold, which can exist in either host or device memory.
[in]	descr	- the descriptor of the dense matrix `A`.
[out]	csrRowPtr	- integer array of `m+1` elements that contains the start of every row and the end of the last row plus one.
[out]	nnzTotalDevHostPtr	- total number of non-zero elements in device or host memory.
[out]	buffer	- buffer allocated by the user whose size is determined by calling `hipsparseSpruneDense2csr_bufferSize` "hipsparseXpruneDense2csr_bufferSize()" or `hipsparseSpruneDense2csr_bufferSizeExt` "hipsparseXpruneDense2csr_bufferSizeExt()".

Return values

HIPSPARSE_STATUS_SUCCESS	the operation completed successfully.
HIPSPARSE_STATUS_INVALID_VALUE	`handle`, `m`, `n`, `lda`, `A`, `threshold`, `descr`, `csrRowPtr`, `nnzTotalDevHostPtr`, or `buffer` pointer is invalid.

Member Function/Subroutine Documentation