摘要
随着半导体工艺的发展,GPU的浮点计算能力迅速提高。如何把GPU处理技术应用到非图形计算领域成为体系结构以及高性能计算领域的热点研究问题。Jacobi和Laplace是科学计算领域常用的计算核心。本文基于AMD的流处理GPU平台设计并实现了这两个算法,相对于CPU平台取得了很好的加速效果。
With the development of the semiconductor technology, the GPU's floating-point computing capacity improves rapidly. How to apply the GPU technology to the non-graphic computing field becomes a highlight in the research of architecture and high performance computing. Jacobi and Laplace algorithms are typical application kernels in scientific computing. This paper implements these two algorithms on AMD's stream processing GPU platform and gets a good speedup compared to the CPU platform.
出处
《计算机工程与科学》
CSCD
北大核心
2009年第A01期93-96,159,共5页
Computer Engineering & Science
基金
国家自然科学基金资助项目(60621003
60873016)
关键词
GPU
流
科学计算
GPU
stream
scientific computing