Intel Knights Corner的结点级内存访问优化

Node-level Memory Access Optimization on Intel Knights Corner

传统编程优化(Traditional Programming Optimization,TPO)在Intel Knights Corner(KNC)上收效甚微,因此提出内存访问优化(Memory Access Optimization,MAO)。将MAO应用到已经过TPO的程序Diffusion 3D上,发现其性能仍然提高了39.1%。主要有2个贡献:1)提出MAO,认为TPO+MAO有助于在KNC上获取最优化性能;2)发现对于stencil代码,基于intrinsic的MAO比基于编译器的MAO更高效。这些发现对于在KNC上优化大规模应用有启发意义。

Traditional programming optimization （TPO） has limited effects on Intel Knights Corner （KNC）. Therefore, we proposed memory access optimization （MAO） for KNC. We applied MAO to TPO version of Diffusion 3D, and its performance is improved by 39. 1%. We made two contributions in this paper： 1） MAO is indispensable to KNC and TPOq-MAO is the path to Ninja Performance—the best optimized performance. 2） Intrinsic-based MAO is more effi- cient to stencil code than compiler-based MAO. Our findings on MAO will inspire optimizations of large-scale applica-tions on KNC.