The central processing unit (CPU) handles all the tasks required for all software on the computer or server to run correctly. A graphics processing unit (GPU), on the other hand, supports the CPU to perform concurrent calculations. A GPU can complete simple and repetitive tasks much faster because it can break the task down into smaller components and finish them in parallel. These cores were initially designed to process images, video game computer graphics, and visual data. General-purpose GPUs (GPGPUs) were adopted to enhance other computational processes, such as transformers and deep learning. More recently, AI is driving GPU tensor cores that achieve significantly higher throughput compared to traditional cores. This book provides a comprehensive overview of GPGPU internals, including its memory structure and optimization, the operation of tensor cores, and the current state-of-the-art technology. It contains programming tips and code in using the NVIDIA CUDA parallel computing platform that allows software developers to use GPGPUs for specific applications. In addition to key examples involving scientific problems, neural networks, image processing, and video coding, the book also covers the graphics pipeline and steps to construct a scene, pipelining of GPU and CPU processes, as well as practical techniques in parallelizing an implementation.
About the AuthorBenny Bing has developed 6 video U.S. patents that were licensed to industry and received the National Association of Broadcasters Technology Innovation Award for demonstrations of advanced media technologies. He has conducted many custom video courses for Comcast/NBC, Qualcomm, Motorola, universities, and the IEEE, and has served as an expert witness on cable TV and video streaming. More recently, two of his video books were adopted by the AI company Anthropic as training material. He is a Senior Member of the IEEE and his current research interest is in optimizing media processing.
