Optimizing the Performance of Streaming Numerical Kernels on the IBM Blue Gene/P PowerPC 450 Processor

TL;DR

This paper presents a high-level assembly synthesis approach to optimize streaming numerical kernels on the IBM Blue Gene/P, achieving significant speedups for 3D stencil computations on energy-efficient in-order processors.

Contribution

It introduces a novel high-level assembly synthesis method tailored for the Blue Gene/P's architecture, improving the performance of streaming kernels through optimized scheduling.

Findings

Achieved a 1.7x speedup over previous results for 3D stencil kernels.

Demonstrated effectiveness of mechanically scheduled variants in various memory scenarios.

Validated the approach through simulation, verification, and analysis.

Abstract

Several emerging petascale architectures use energy-efficient processors with vectorized computational units and in-order thread processing. On these architectures the sustained performance of streaming numerical kernels, ubiquitous in the solution of partial differential equations, represents a challenge despite the regularity of memory access. Sophisticated optimization techniques are required to fully utilize the Central Processing Unit (CPU). We propose a new method for constructing streaming numerical kernels using a high-level assembly synthesis and optimization framework. We describe an implementation of this method in Python targeting the IBM Blue Gene/P supercomputer's PowerPC 450 core. This paper details the high-level design, construction, simulation, verification, and analysis of these kernels utilizing a subset of the CPU's instruction set. We demonstrate the effectiveness of our approach by implementing several three-dimensional stencil kernels over a variety of cached memory scenarios and analyzing the mechanically scheduled variants, including a 27-point stencil achieving a 1.7x speedup over the best previously published results.