Basics of OpenCL

Abstract

Open Computing Language (OpenCL), which is generally called a heterogeneous computing system, is an open programming framework of parallel computing for a calculation system comprising different computers (e.g., CPU, GPU, DSP, FPGA). Although CUDA only applies to NVIDIA’s GPU, OpenCL can drive the GPUs of different vendors (AMD, NVIDIA, Intel, Qualcomm), as well as the CPU or another computer, via the same OpenCL-written source code. Thus, OpenCL is more portable than CUDA. In this chapter, OpenCL, as well as the strategy for constructing a calculation environment, is briefly introduced employing a source code example for calculating a computer-generated hologram (CGH). Based on the contents of this chapter, holography calculations employing OpenCL can be attempted. Readers who wish to improve their OpenCL coding skills, programming guides that are published by chip vendors, etc., may be consulted.