GPU Offloading in ExaHyPE Through C++ Standard Algorithms

TL;DR

This paper explores porting ExaHyPE's GPU offloading to ISO C++17 parallel algorithms, demonstrating feasibility but with performance trade-offs compared to OpenMP.

Contribution

It presents a novel approach to GPU offloading in ExaHyPE using standard C++17 parallel algorithms, avoiding vendor-specific extensions.

Findings

ISO C++17 parallel algorithms are feasible for complex scientific applications.

Performance is slower than OpenMP target offloading.

No vendor-specific extensions are needed for this implementation.

Abstract

The ISO C++17 standard introduces \emph{parallel algorithms}, a parallel programming model promising portability across a wide variety of parallel hardware including multi-core CPUs, GPUs, and FPGAs. Since 2019, the NVIDIA HPC SDK compiler suite supports this programming model for multi-core CPUs and GPUs. ExaHyPE is a solver engine for hyperbolic partial differential equations for complex wave phenomena. It supports multiple numerical methods including Finite Volumes and ADER-DG, and employs adaptive mesh refinement with dynamic load balancing via space-filling curves as well as task-based parallelism and offloading to GPUs. This study ports ExaHyPE's tasks over blocks of Finite Volumes to the ISO C++ parallel algorithms programming model, and compares its performance and usability against an OpenMP implementation with offloading via OpenMP target directives. It shows that ISO C++ is a feasible programming model for non-trivial applications like our task-based AMR code. The realisation is bare of vendor-specific or non-C++ extensions. It however is slower than its OpenMP counterpart. \vspace{-1cm}