适合于分布式并行计算的一种并行广义乘积型双共轭残差方法(英文)

针对求解大型稀疏非对称线性方程组,提出适合于分布式并行环境的一种并行广义乘积型双共轭残差(GPBiCR)方法(简记为PGPBiCR方法).通过重构GPBiCR方法,新方法将原方法中的三个全局同步点降低到了一个,且内积所需的通讯时间可与向量校正的计算时间有效地重叠.代价仅是稍微增加了一些计算量,而相比于全局通讯时间的降低,这是可以忽略不计的.性能和等效率分析表明,PGPBiCR方法比GPBiCR方法具有更好的并行性和可扩展性,其中可扩展性可改进3倍,而并行通讯性能可改进66.7%.数值试验得到了与理论分析相吻合的结果.     

An efficient block variant of robust structured multifrontal factorization method

Based on the two-dimensional three-temperature （2D3T） radiation diffusion equations and its discrete system, using the block diagonal structure of the three-temperature matrix, the reordering and symbolic decomposition parts of the RSMF method are replaced with corresponding block operation in order to improve the solution efficiency. We call this block form method block RSMF （in brief, BRSMF） method. The new BRSMF method not only makes the reordering and symbolic decomposition become more effective, but also keeps the cost of numerical factorization from increasing and ensures the precision of solution very well. The theoretical analysis of the computation complexity about the new BRSMF method shows that the solution efficiency about the BRSMF method is higher than the original RSMF method. The numerical experiments also show that the new BRSMF method is more effective than the original RSMF method.