Matrix Parallel Algorithm Application Research Base On MPI And Linux Cluster

With the development of high-performance calculates technology, brought forward much higher request about the Arithmetic speed , timeliness handling accuracy , and fleetness of computer. Thereupon, the PC cluster system based on fleet technology and distributed store technology becomes the effective approach satisfying the high-performance data handling request . Hou to make better use of the PC fleet, design out high-effect and stable parallel algorithms , is one of the hot spot in the computer science field at present, also has broadly applying background and practical value.Common use about parallel programming standards is MPI , PVM etc. , information delivery interface MPI has good transplant, powerful function, and high efficiency etc., so it becomes the most important parallel programming implement at present. The thesis combined MPI base on the parallel cluster environment , carried out thorough analysis and study on the parallel algorithm.This paper introduced the builds of parallel cluster under the Windows/Linux environment, set forth parallel machine, fundamental theory of parallel algorithm, and MPI parallel programming methos and coming true. Taken the matrix multiplies and shortest path algorithm as example respectively, analysed the deficiency that existing in now available algorithm , brought forward new parallel methods, and applying them to the reality use of graphics study, got the very good parallel function by verifying.Finally, the algorithms brought forward by this paper were wrote in C+MPI language, taken reality measure analysis under the cluster environment, and summed up experience and deficiency of this thesis, pointed out the problem needing to continue studying.