OpenCL Programming Book

Overview

The book starts with the basics of parallelization, covers the main concepts, grammar, and setting up a development environment for OpenCL, concluding with source-code walkthroughs of the FFT and Mersenne Twister algorithms written in OpenCL.

It is highly recommended for those wishing to get started on programming in OpenCL.

Download

PDF version


The OpenCL Programming Book
» View Sample
File Size: 3.49MB
Digital: 246 pages
Publisher: Fixstars Corporation
Author: Fixstars Corporation (Ryoji Tsuchiyama, Takashi Nakamura, Takuro Iizuka, Akihiro Asahara, Satoshi Miki)
Published Date: 3/31/2010
Price: USD 19.50
Buy now

New!! HTML version

HTML version of The OpenCL Programming Book is available for FREE.
» Read HTML version

Foreword

OpenCL is an exciting new open standard that is bringing the power of programming complex multi-processor systems to a much wider audience than ever before. Khronos is fully committed not only to creating the specification but also fostering a rich and diverse OpenCL ecosystem. A vital piece of any living standard is enabling the industry to truly understand and tap into the power full potential of the technology. Fixstars is a skilled OpenCL practitioner and is ideally qualified to create state-of-the-art OpenCL educational materials. I wholeheartedly recommend to this book to anyone looking to understand and start using the amazing power of OpenCL.

Neil Trevett
President, The Khronos Group

Sample Code

The sample code contained in the book can be downloaded here:
» Download (zip:825KB)

Limitation for executing the sample code

The sample code will not run on the OpenCL that comes with CUDA SDK3.0 β. Please download the OpenCL from the link below.
» http://developer.nvidia.com/object/get-opencl.html

Also, when using a NVIDIA GPU on Windows, the TDR (a graphics time-out function) must be turned off. For details, please visit the link below.
» http://www.microsoft.com/whdc/device/display/wddm_timeout.mspx

Limitation specific to sample code 6-2/4

Due to the hardware resources used within the kernel, the NVIDIA GPU must contain more than 16 CUDA Cores.