Accelerating Communication for Parallel Programming Models on GPU Systems

TL;DR

This paper demonstrates how the UCX framework can be used to accelerate GPU-aware communication in parallel programming models, significantly improving latency and bandwidth for microbenchmarks and real-world applications on GPU systems.

Contribution

The work presents a unified GPU-aware communication layer for Charm++ ecosystem models using UCX, achieving substantial performance improvements.

Findings

Latency improvements up to 17.4x across models

Bandwidth increases up to 10.5x

Enhanced application performance by up to 19.7x

Abstract

As an increasing number of leadership-class systems embrace GPU accelerators in the race towards exascale, efficient communication of GPU data is becoming one of the most critical components of high-performance computing. For developers of parallel programming models, implementing support for GPU-aware communication using native APIs for GPUs such as CUDA can be a daunting task as it requires considerable effort with little guarantee of performance. In this work, we demonstrate the capability of the Unified Communication X (UCX) framework to compose a GPU-aware communication layer that serves multiple parallel programming models of the Charm++ ecosystem: Charm++, Adaptive MPI (AMPI), and Charm4py. We demonstrate the performance impact of our designs with microbenchmarks adapted from the OSU benchmark suite, obtaining improvements in latency of up to 10.1x in Charm++, 11.7x in AMPI, and 17.4x in Charm4py. We also observe increases in bandwidth of up to 10.1x in Charm++, 10x in AMPI, and 10.5x in Charm4py. We show the potential impact of our designs on real-world applications by evaluating a proxy application for the Jacobi iterative method, improving the communication performance by up to 12.4x in Charm++, 12.8x in AMPI, and 19.7x in Charm4py.