Multi-level Memory-Centric Profiling on ARM Processors with ARM SPE

TL;DR

This paper introduces a multi-level memory profiling tool for ARM processors using the ARM SPE extension, providing insights into profiling overhead and accuracy for HPC and cloud workloads.

Contribution

It presents the first quantitative evaluation of ARM SPE-based memory profiling, analyzing overhead and accuracy across different sampling configurations.

Findings

ARM SPE enables effective memory profiling on ARM processors.

Profiling overhead varies with sampling period and buffer size.

The tool provides accurate insights into memory bottlenecks in HPC and cloud workloads.

Abstract

High-end ARM processors are emerging in data centers and HPC systems, posing as a strong contender to x86 machines. Memory-centric profiling is an important approach for dissecting an application's bottlenecks on memory access and guiding optimizations. Many existing memory profiling tools leverage hardware performance counters and precise event sampling, such as Intel PEBS and AMD IBS, to achieve high accuracy and low overhead. In this work, we present a multi-level memory profiling tool for ARM processors, leveraging Statistical Profiling Extension (SPE). We evaluate the tool using both HPC and Cloud workloads on the ARM Ampere processor. Our results provide the first quantitative assessment of time overhead and sampling accuracy of ARM SPE for memory-centric profiling at different sampling periods and aux buffer sizes.