Programming Research Group Technical Report TR-15-97

Parallel Algorithms for Structured Matrix Computations

Yuguang Huang, Date to be advised, 25pp In this report, we give some efficient algorithms for structured matrix computations. The matrices involved are tridiagonal matrix and skew-symmetric matrix. More specifically, we first investigate into the tridiagonal system solvers using linear recurrence and a two-way strategy. Based on that, an optimal tridiagonal matrix inversion algorithm is further derived. For the skew-symmetric system, we give an efficient iterative algorithm using the idea of conjugate gradient method. The parallel complexity issues are discussed using the Bulk Synchronous Parallel model.


This paper is available as a 327,918 gzipped PostScript file.