N体问题直接法的众核实现研究

A Study on Mass Nuclear Realization of Direct Method of N Body Problem

展示了一个Intel众核处理架构上优化的N体内核代码,该优化版本在传统处理器上也获得了性能提升;探讨如何快速为优化过程确定一个目标,最终的优化版本达到了峰值上界的89%,因此这是一个较为合理的终点;还展示了在Intel Xeon Phi协处理器上的所有这些优化也能够在Intel Xeon处理器上获得性能提升。

This paper presents a N kernel code optimized Intel the kernel architecture,which also achieves performance improvements on traditional processors.It explores how to quickly set a goal for the optimization process,and the final optimization version reaches 89%of the peak upper bound,so this is a more reasonable end point.It also shows that all these optimizations on Intel Xeon Phi coprocessors can also achieve performance improvements on Intel Xeon processors.