关键词: Imagine Jaeobi迭代 Seidel迭代 流编程模型

Abstract:

The Imagine stream processor of Stanford University has powerful computing ability, and the key point is how to apply this architecture into the domai  n of scientific computing. The iteration method in resolving linear algebraic equations has broad applications in engineering and scientific computing. The algorithm has great computing intensity and parallelism, which are suitable for the computing model, in stream processors. In this paper, we introdu ce the mapping of the Jacobi iteration and the Seidel iteration onto the Imagine stream processor according to the size and denseness of the coefficient  matrix respectively. The result shows that the iteration algorithm can exploit the computing ability of Imagine effectively and get high performance sp eedup.

Key words: （Imagine, Jaeobi iteration, Seidel iteration, stream programming model）