Towards heterogeneous parallelism for SPHinXsys

TL;DR

This paper explores using the SYCL programming standard to enable heterogeneous parallelism in SPHinXsys, allowing the same code to run efficiently on CPUs and GPUs, with significant speed-ups demonstrated.

Contribution

It introduces a SYCL-based execution model for SPHinXsys, facilitating hardware-agnostic parallelism and improving performance across diverse computing architectures.

Findings

SYCL implementation achieves substantial speed-up over CPU-only version

Data-structure and communication strategies are optimized for heterogeneous hardware

Performance benchmarks demonstrate effective multi-hardware execution

Abstract

Simulations based on particle methods, such as Smoothed Particle Hydrodynamics (SPH), are known to be computationally demanding. While such methods have for long been executed in parallel on multi-core CPUs, in recent years the increasing adoption of many-core accelerators, such as GPUs. However, hardware fragmentation and vendor-specific programming interfaces are still characterizing their market. Hence, support for various hardware configurations may easily lead to non-trivial and less maintainable implementations. To leverage over some higher-level specifications have become available recently, such as the SYCL programming standard, this work highlights the initial effort in adopting the SYCL standard for the execution of SPHinXsys, an open-source multi-physics library. The result is an execution model able to run the same implementation on variable (heterogeneous) hardware, with considerable speed-up compared to the current multi-core CPU parallelization. Among others, representation of data-structures for parallel access, communication strategies, and parallel methods for data sorting will be topics discussed in depth. Benchmarks has also been presented, showcasing performance comparisons between the current multi-core CPU implementation and the newly introduced SYCL parallelization with a GPU back-end.