Improving the Performance of a Verified Linear System Solver Using Optimized Libraries and Parallel Computation

Numerical Validation in Current Hardware Architectures

计算机科学;物理学

Mariana Kolberg ; Gerd Bohlender ; Dalcidio Claudio

PID  :  81054

学科分类：计算机科学（综合）

来源: CEUR

【 摘 要 】

A parallel version of the selfverified method for solving linear systems was presented in [16, 15]. In this research we propose improvements aiming at a better performance. The idea is to implement an algorithm that uses tech nologies as MPI communication primitives associated to libraries as LAPACK, BLAS and CXSC, aiming to provide both selfverification and speedup at the same time. The algorithms should find an enclosure even for very illconditioned problems. In this scenario, a parallel version of a selfverified solver for dense linear systems appears to be essential in order to solve bigger problems. More over, the major goal of this research is to provide a free, fast, reliable and accurate