Accelerated Computing With HIP: Second Edition

This book is designed for programmers who wish to use GPUs to improve application performance, and it is suitable for both AMD and NVIDIA GPU programmers, as HIP is a portable language that runs on both platforms. ROCm is open sourced, allowing other GPU developers to support this platform. This book does not require knowledge of CUDA programming, however, we highlight how HIP differs from CUDA while explaining how to port those programs to HIP, promoting interoperability such that a single application can be executed on different underlying hardware. For non-CUDA programmers, our book starts with the basics by presenting how HIP is a full-featured parallel programming language. Then, we provide coding examples that cover a wide range of relevant programming paradigms.