Triangular (LU) factorization has long been a widely used method for solving sparse linear matrix equations arising from power system analysis problems in the industry. Recent interest has been shown in parallel processing in order to further improve the speed of solution of the linear matrix equation. Parallel triangular factorization scheduling using two fast coarse grain scheduling schemes is presented in this paper along with performance results obtained using the iPSC hypercube parallel processing computer. The study shows that the solution of the parallel LU factorization problem using different coarse grain scheduling schemes on a coarse grain parallel processing computer can have widely varying performance results.
ASJC Scopus subject areas
- Energy Engineering and Power Technology
- Electrical and Electronic Engineering