Multicore-based performance optimization for dense matrix computation

Abstract

To make the traditional applications benefit from multicore processors, the traditional Gaussian Elimination algorithm is improved to enhance its parallel performance under multicore architecture by matrix partition. The stability of the original algorithm is guaranteed. The hit rate of cache is improved by adjusting the computation sequence, the experiment shows that the speedup can reach 1.8 under duo core CPU environment when evaluating the inverse of dense matrix. © 2012 Springer-Verlag.