Today, the SYCL standard represents the most advanced programming model for heterogeneous computing, delivering both productivity, portability, and performance in pure C++17. SYCL 2020, in particular, represents a major enhancement that pushes the boundaries of heterogeneous programming by introducing a number of new features. As the new features are implemented by existing compilers, it becomes critical to assess the maturity of the implementation through accurate and specific benchmarking. This paper presents SYCL-Bench 2020, an extended benchmark suite specifically designed to evaluate six key features of SYCL 2020: unified shared memory, reduction kernel, specialization constants, group algorithms, in-order queue, and atomics. We experimentally evaluate SYCL-Bench 2020 on GPUs from the three major vendors, i.e., AMD, Intel, and NVIDIA, and on two different SYCL implementations AdaptiveCPP and oneAPI DPC++.

SYCL-Bench 2020: Benchmarking SYCL 2020 on AMD, Intel, and NVIDIA GPUs

Crisci L.;Carpentieri L.;Cosenza B.
2024

Abstract

Today, the SYCL standard represents the most advanced programming model for heterogeneous computing, delivering both productivity, portability, and performance in pure C++17. SYCL 2020, in particular, represents a major enhancement that pushes the boundaries of heterogeneous programming by introducing a number of new features. As the new features are implemented by existing compilers, it becomes critical to assess the maturity of the implementation through accurate and specific benchmarking. This paper presents SYCL-Bench 2020, an extended benchmark suite specifically designed to evaluate six key features of SYCL 2020: unified shared memory, reduction kernel, specialization constants, group algorithms, in-order queue, and atomics. We experimentally evaluate SYCL-Bench 2020 on GPUs from the three major vendors, i.e., AMD, Intel, and NVIDIA, and on two different SYCL implementations AdaptiveCPP and oneAPI DPC++.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11386/4886391
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 3
  • ???jsp.display-item.citation.isi??? ND
social impact