the traditional methods are to solve the linear algebra equations directly, based on matrix factorization such as lu decomposition . with this kind of methods, the " true " solution can be derived if there is no consideration of the round error 解线性代数方程组的传统方法是利用lu分解等直接求解,虽然传统方法具有理论上直接得到真解的优点,但当系数矩阵条件数很大时,存在严重的稳定性问题。
the high-performance solution of sparse linear algebra equations is very important in solving many problems from science and engineering applications, including computational fluid, simulation and design of materials, data processing in oil exploitation and earthquake prediction, numerical forecast of weather, and numerical simulation of nuclear blast 稀疏线性代数方程组的高效求解是许多科学与工程计算的核心,如计算流体力学、材料模拟与设计、石油地震数据处理、数值天气预报从核爆数值模拟等都离不开稀疏线性代数方程组的求解。