Performance of OpenCL

Tadeusz Puzniakowski

The University of Gdansk, Wit Stwosz St. 57, 80-952 Gdansk, Poland

The University of Gdansk, 2013

BibTeX @article{puzniakowski2013performance, title={Performance of OpenCL}, author={PU{‘Z}NIAKOWSKI, Tadeusz}, year={2013} }

OpenCL is a relatively new standard that supports computation on a variety of parallel architectures. The author was unable to find reliable information about performance of OpenCL programs on CPU’s in comparison to traditional parallel processing standards like OpenMP. This paper describes the results of an experiment that tries to answer the following question: "Which standard performs better on multi-core CPU: OpenMP or OpenCL?". The experiment involves analysis of the efficiency of a linear algebra algorithm implemented in OpenCL with different optimization methods and run on different hardware, and compared with the efficiency of the same algorithm implemented in OpenMP.

Rate this item: 1.00 2.00 3.00 4.00 5.00 Submit Rating No votes yet. Please wait...