The Feasibility of Using OpenCL Instead of OpenMP for Parallel CPU Programming

TL;DR

This paper compares OpenCL and OpenMP for CPU programming, analyzing their performance and ease of use in compute-heavy tasks, and finds that OpenCL can be a viable alternative to OpenMP under certain conditions.

Contribution

It provides a detailed comparison of OpenCL and OpenMP for CPU parallel programming, highlighting scenarios where OpenCL performs competitively.

Findings

OpenCL and OpenMP each excel in different execution scenarios.

OpenCL can match OpenMP performance with high thread invocation counts.

Using OpenCL requires careful consideration of potential pitfalls.

Abstract

OpenCL, along with CUDA, is one of the main tools used to program GPGPUs. However, it allows running the same code on multi-core CPUs too, making it a rival for the long-established OpenMP. In this paper we compare OpenCL and OpenMP when developing and running compute-heavy code on a CPU. Both ease of programming and performance aspects are considered. Since, unlike a GPU, no memory copy operation is involved, our comparisons measure the code generation quality, as well as thread management efficiency of OpenCL and OpenMP. We evaluate the performance of these development tools under two conditions: a large number of short-running compute-heavy parallel code executions, when more thread management is performed, and a small number of long-running parallel code executions, when less thread management is required. The results show that OpenCL and OpenMP each win in one of the two conditions. We argue that while using OpenMP requires less setup, OpenCL can be a viable substitute for OpenMP from a performance point of view, especially when a high number of thread invocations is required. We also provide a number of potential pitfalls to watch for when moving from OpenMP to OpenCL.