Document Type

Conference Proceeding

DOI

10.1007/978-3-030-58144-2_5

Publication Date

9-2020

Keywords

OpenMP, Task parallelism, Stencil computation, Loop scheduling

Abstract

Simulations based on stencil computations (widely used in geosciences) have been dominated by the MPI+OpenMP programming model paradigm. Little effort has been devoted to experimenting with task-based parallelism in this context. We address this by introducing OpenMP task parallelism into the kernel of an industrial seismic modeling code, Minimod. We observe that even for these highly regular stencil computations, taskified kernels are competitive with traditional OpenMP-augmented loops, and in some experiments tasks even outperform loop parallelism.

This promising result sets the stage for more complex computational patterns. Simulations involve more than just the stencil calculation: a collection of kernels is often needed to accomplish the scientific objective (e.g., I/O, boundary conditions). These kernels can often be computed simultaneously; however, implementing this simultaneous computation with traditional programming models is not trivial. The presented approach will be extended to cover simultaneous execution of several kernels, where we expect to fully exploit the benefits of task-based programming.

Share

COinS
 
 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.