A Fast Solution Method for Large Scale Linear Sparse Equations Based on Parallelism

Abstract

The solution of large sparse linear matrix is an important research problem to improve the efficiency of numerical simulation technology. As the complexity of the model increases, the memory and computing time required to solve the sparse linear matrix increase sharply. To solve these problems, this paper studies the fast solution method of large-scale linear sparse equations based on parallel. In this paper, through semi storage and CSR compression format, the memory occupied by sparse matrix storage is greatly reduced, and the CPU parallel and GPU parallel of preprocessing conjugate gradient method are used to speed up the solution of sparse matrix. By solving the sparse matrix obtained by finite element potential calculation, compared with the general sparse matrix storage method and solution method, the proposed method can reduce the storage space by about 40%, and obtain a GPU parallel acceleration ratio of about 32, which verifies the effectiveness of the proposed method.