Adaptive sparse matrix vector multiplication strategy selection and optimization method

Abstract

The invention discloses a self-adaptive sparse matrix vector multiplication strategy selection and optimization method, which is suitable for a GPU (Graphics Processing Unit) architecture, and comprises the following steps: partitioning a to-be-processed matrix according to rows, counting the number of non-zero elements of each matrix sub-block, and if the difference multiple of the number of non-zero elements of each matrix sub-block is higher than a first preset threshold value, determining that the matrix is not processed; if yes, processing is carried out by adopting a self-adaptive CSR-Vector algorithm; counting the row average non-zero element number of the to-be-processed matrix, and if the row average non-zero element number of the matrix is lower than a second preset threshold value, adopting an improved CSR-Stream algorithm for solving; counting the number of non-zero elements of the to-be-processed matrix, and if the number of the non-zero elements of the to-be-processed matrix is larger than a third preset threshold value, adopting a hola algorithm for solving; and if all the conditions are not met, a CSR-Vector algorithm is adopted for solving. According to the method, adaptive efficient SpMV solving for different application problems is realized.

Description

technical field [0001] The invention relates to the technical field of high-performance computing, in particular to an adaptive sparse matrix-vector multiplication strategy selection and optimization method suitable for GPU architecture. Background technique [0002] SpMV (Sparse matrix–vector multiplication, sparse matrix-vector multiplication) is a matrix-vector multiplication operation of the form y=αAx+βy (where A is a sparse matrix, x and y are dense vectors, and α and β are scalars). It is widely used in scientific computing, economic construction, signal processing, document retrieval and other fields. Usually, the matrix involved in the calculation is sparse, such as the matrix produced by the discretization of physical processes. In recent years, SpMV has been classified among seven numerical methods considered to be of great importance to science and engineering for at least the next decade. [0003] For different storage formats of sparse matrix A, there will be...

AI Technical Summary

Problems solved by technology

[0008] The invention provides an adaptive sparse matrix-vector multiplication strategy selection and optimization method to solve the problem that although the existing SpMV algorithm improves the calculation performance under specific matrix conditions, it has poor versatility in solving and optimizing problems in different fields question

Images