Porting a sparse linear algebra math library to Intel GPUs

TL;DR

This paper introduces a platform-portable open source sparse linear algebra library supporting Intel GPUs via DPC++, benchmarks its performance, and compares it with Intel's oneMKL to evaluate efficiency and potential for scientific computing.

Contribution

It presents the first open source, platform-portable sparse linear algebra library for Intel GPUs using DPC++, and benchmarks its performance against existing Intel libraries.

Findings

The library achieves competitive performance on Intel GPUs.

DPC++ effectively translates raw hardware performance into application-level performance.

The library's efficiency is comparable to Intel's oneMKL routines.

Abstract

With the announcement that the Aurora Supercomputer will be composed of general purpose Intel CPUs complemented by discrete high performance Intel GPUs, and the deployment of the oneAPI ecosystem, Intel has committed to enter the arena of discrete high performance GPUs. A central requirement for the scientific computing community is the availability of production-ready software stacks and a glimpse of the performance they can expect to see on Intel high performance GPUs. In this paper, we present the first platform-portable open source math library supporting Intel GPUs via the DPC++ programming environment. We also benchmark some of the developed sparse linear algebra functionality on different Intel GPUs to assess the efficiency of the DPC++ programming ecosystem to translate raw performance into application performance. Aside from quantifying the efficiency within the hardware-specific roofline model, we also compare against routines providing the same functionality that ship with Intel's oneMKL vendor library.