The Comparisons of OpenCL and OpenMP Computing Paradigm

Abstract

Graphics processing units (GPUs) in a computer system are increasingly adopted to create a realistic environment in 3D applications. Despite the extremely high parallelism of these devices with a trem endous amount of processing elements, GPUs are seldom used in scientific applications owing to their difficulty in programming. A dditionally, vendor-provided APIs are too specific to cross platform. An open standard, OpenCL, has subsequently been developed to provide universal APIs and programming paradigms for various GPUs. This work demonstrates the capabilities of OpenCL with several platforms, based on a preliminary example. The proposed benchmarks with different attributes are implemented by OpenMP and OpenCL sequentially to compare their differences. The benchmarks are then targeted on several GPUs and conventional servers. Experimental results indicate that inexpensive GPUs perform better than servers if OpenCL paradigms are adopted.