Improving Cache Performance for Uniprocessors

9525767麦金利在这项研究计划拨款(RPG)中，PI将开发新的方法，将编译时分析、程序转换和数据转换结合在一起，以生成实现良好高速缓存性能的程序。在规划期间，PI将探索和比较改善循环嵌套内局部性的技术和改善循环嵌套之间局部性的技术(尽管数据量很大)的好处。成功的完成将导致一个更雄心勃勃的研究计划，研究编译时技术，以提高单处理器和共享内存多处理器上的科学应用程序的内存性能。***

9525767 McKinley In this Research Planning Grant (RPG) the PI will develop new approaches for combining compile time analysis, program transformation, and data transformation to generate programs that achieve good cache memory performance. During the planning period, the PI will explore and compare benefits of techniques that improve locality within loop nests and techniques that improve locality between loop nests in spite of large data sizes. A successful completion will lead to a more ambitious research plan to investigate compile time techniques to improve memory performance of scientific applications on uniprocessors and shared memory multiprocessors. ***