Research on parallel Gauss elimination method in cloud computing platform

Research on parallel Gauss elimination method in cloud computing platform

Pan Xiaohui

In order to solve the problem that the Gaussian elimination algorithm of the serial partial selection principal element cannot make full use of the multi-core processor, this paper proposes and realizes the Gaussian elimination algorithm of the partial selection principal element of parallel multithreading, and analyzes and optimizes the whole algorithm, which makes the data storage layout and algorithm match the memory pattern, thus greatly improving the program performance. By comparing and analyzing the experimental results on the local Linux server and the various platforms of the Amazon EC2 Cloud in the United States, the Gaussian elimination algorithm of the partially selected principal element is affected by the cache, so the performance is best on the platform with more balanced CPU and memory/cache configuration. This paper presents an efficient and scalable Gaussian elimination algorithm for multithreaded parallel partial-selection principal elements and a method for parallelization and optimization of general serial algorithms.

Research on parallel Gauss elimination method in cloud computing platform