Abstract |
---|
A parallel computation model to invert a lower triangular matrix is presented. Method uses Gauss elimination with sweeping technique. Performance charactersistics that we obtain are O(n) time and O(n square) processors leading to an efficiency of O(1/n). A comparative performance study with the available fastest parallel matrix inversion algorithms which run in O(log square n) time on O(n cube) processors, is given. We believe that the method presented here is superior over the existing methods in efficiency measure and in processor complexity. |
Contact |
Ayse Kiper Department of Computer Engineering,Middle East Technical University,06531 Ankara , Turkey ayse@rorqual.cc.metu.edu.tr |