A Data-Parallel Implementation of the Gauss-Seidel Iteration Method Applied to the Sliding Box Problem
Abstract
The data-parallel programming is convenient for numerical computation dealing with data collections like arrays. Physicists use very efficient sequential programs to simulate physical processes, but such programs can not be easily adapted to parallel architectures. The Gauss-Seidel fix point computation is an example. This paper aims at showing how data-parallel modeling and programming helps in redesigning sequential programs for an efficient parallel implementation whatever the parallel architecture is.