Improving the scalabiliy of neutron cross-section lookup codes on multicore NUMA system

TL;DR

This paper investigates and improves the scalability of neutron cross-section lookup codes on multicore NUMA systems using XSBench, achieving significant efficiency and energy improvements through memory and page-size optimizations.

Contribution

It identifies NUMA-related memory allocation issues affecting scalability and demonstrates effective optimization strategies for XSBench, a proxy for neutron cross-section calculations.

Findings

Scalability improved from 70% to 95% efficiency.

Optimizations reduced energy consumption by 25%.

Page-size optimization increased performance by 1.5x.

Abstract

We use the XSBench proxy application, a memory-intensive OpenMP program, to explore the source of on-node scalability degradation of a popular Monte Carlo (MC) reactor physics benchmark on non-uniform memory access (NUMA) systems. As background, we present the details of XSBench, a performance abstraction "proxy app" for the full MC simulation, as well as the internal design of the Linux kernel. We explain how the physical memory allocation inside the kernel affects the multicore scalability of XSBench. On a sixteen-core, two-socket NUMA testbed, the scaling efficiency is improved from a nonoptimized 70% to an optimized 95%, and the optimized version consumes 25% less energy than does the nonoptimized version. In addition to the NUMA optimization we evaluate a page-size optimization to XSBench and observe a 1.5x performance improvement, compared with a nonoptimized one.