摘要
多维递归方程组在并行求解时存在串并行不一致问题,提供三种Krylov子空间迭代求解方法——PCG/ATCG和GMRES来解决这一问题,并采用典型算例对这三种Krylov子空间迭代方法进行正确性验证和加速比测试。试验表明这三种Krylov子空间迭代法在并行规模较大的情况下,均能够正确求解多维递归方程组,并且加速特性良好。
The parallel solving of multi-dimension recursive equations has inconformity between the serial computation and the parallel computation. To solve this problem, in the dissertation we present three Krylov subspace iteration methods, PCG, ATCG and GMRES. Some typical numerical examples are used to verify the correetness and conduct the acceleration ratio test for these three methods. Test results show that these three Krylov subspace iteration methods are able to correctly resolve the multi-dimension recursive equations with satisfied acceleration property even in the condition of bigger parallel scale.
出处
《计算机应用与软件》
CSCD
北大核心
2012年第11期83-86,共4页
Computer Applications and Software
基金
国家高技术研究发展计划(10072077)
