Task inefficiency patterns for a wave equation solver

TL;DR

This paper analyzes task scheduling inefficiencies in OpenMP 5.0 when used with a wave equation solver across multiple architectures, identifying flaws, their performance impacts, workarounds, and proposing standard extensions.

Contribution

It identifies specific task-scheduling flaws in OpenMP 5.0 runtimes for wave equation solvers, demonstrating their impact and suggesting standard extensions to improve performance.

Findings

Identified key task-scheduling flaws affecting performance

Demonstrated performance impacts across three architectures

Proposed extensions to the OpenMP standard

Abstract

The orchestration of complex algorithms demands high levels of automation to use modern hardware efficiently. Task-based programming with OpenMP 5.0 is a prominent candidate to accomplish this goal. We study OpenMP 5.0's tasking in the context of a wave equation solver (ExaHyPE) using three different architectures and runtimes. We describe several task-scheduling flaws present in currently available runtimes, demonstrate how they impact performance and show how to work around them. Finally, we propose extensions to the OpenMP standard.