Introduction to GPU Programming

By Rekha R. Nair, Kuldeep Singh Kaswan, Balamurugan Balusamy

This book provides a detailed guide to programming Graphics Processing Units (GPUs) for high-performance computing, covering both foundational concepts and advanced techniques. It introduces GPU architectures, programming frameworks like CUDA and OpenCL, and optimization strategies for memory and thread management. Readers will learn to implement parallel algorithms, integrate GPU programming with libraries like cuBLAS and TensorRT, and explore advanced topics such as multi-GPU programming and unified memory.

Explains GPU execution models (SIMD, SIMT), memory hierarchies (global, shared, constant), and thread-level parallelism with practical illustrations. Guides through writing kernels, managing grids and threads, synchronization techniques, and error handling in CUDA and OpenCL. Covers GPU memory allocation, coalesced memory access, latency hiding, and bandwidth maximization for performance efficiency. Provides coding examples for reduction, scan, sorting, and matrix operations, analyzing synchronization, scalability, and performance bottlenecks. Demonstrates GPU acceleration in AI and data science workflows using cuBLAS, TensorRT, RAPIDS, PyCUDA, and Numba. Explores multi-GPU programming, unified memory, dynamic parallelism, and GPU virtualization, addressing challenges in energy efficiency and portability.

Featuring comprehensive real-world case studies in artificial intelligence, data science, and scientific computing, this resource provides a rigorous and practical foundation for understanding and leveraging GPU acceleration in computationally intensive applications. This book is ideal for students, researchers, and professionals seeking to harness GPU acceleration for computational tasks.

• Format: Paperback
• Language: English
• ISBN: 1041306792
• ISBN13: 9781041306795
• Release Date: March 2027
• Publisher: Chapman and Hall/CRC
• Length: 360 Pages
• Dimensions: 6.1" x 9.2"