A Performance Evaluation of OpenCL and Intel Cilk Plus on a Graphic Rendering Problem
Y. J. Gambo *
Department of Computer Science, Federal University Wukari, Wukari, Taraba State, Nigeria.
S. B. Junaidu
Department of Mathematics, Ahmadu Bello University, Zaria, Kaduna State, Nigeria.
S. E. Abdullahi
Department of Mathematics, Ahmadu Bello University, Zaria, Kaduna State, Nigeria.
C. I. Saidu
Bingham University, Karu, Nasarawa State, Nigeria.
*Author to whom correspondence should be addressed.
Abstract
Parallel programming is fast evolving into a central core of high performance computing. Several computing hardware vendors now embed several processors into computers to increase speed and performance thus the need to maximize and utilize computing resources for maximum performance. Rendering of graphics has also been a challenge in terms of speed that architects and graphic experts have found it difficult to render graphics in a timely manner at minimal cost. In this paper, a comparative study was carried out on two recent and powerful multiprocessor parallel programming languages OpenCL and Intel Cilk Plus which were used to separately program an embarrassingly parallel graphic rendering algorithm, raytracing. The programs were ran and profiled on three different computers with varying specifications on windows using Microsoft Visual Studio 2012 as IDE. The result showed that OpenCL was consistent after first run without restart than Intel Cilk Plus and as raytracing depth increased the performance gap between OpenCL and Intel Cilk Plus also increased with OpenCL showing the better performance.
Keywords: OpenCL, intel cilk plus, parallel programming, raytracing, high performance computing, rendering