Iterative-Linear-Solver-for-Sparse-System

Introduction

There are 3 most common linear iterative solvers: Jacobi、Gauss-Seidel and Preconditioned-Conjugate-Gradient.

Features

• serial and parallel implementations of Jacobi solver with Chebyshev acceleration

• serial and parallel implemantations of Gauss-Seidel solver

• the convergence of both Jacobi and GS solver can be controled by adjusting the relaxation

• the parallel version of Gauss-Seidel is based on random graph coloring algorithm

• the implementation of the preconditioner contains Jacobi、incomplete-Cholesky factorization and SSOR

the parallel implementation is based on oneTBB

Jacobi

• simulation effects

• comparison of serial and parallel speed

• effects of Chebyshev acceleration

Gauss-Seidel

• simulation effects

• comparison of serial and parallel speed

PCG

• simulation effects

• comparison of different preconditioners

Future Work

• continue to improve the overall performance of the code