CUDA Programming from Basics to Advanced – Paperback

Description

Author(s): Finbarrs Oketunji
Publisher: Independently Published
Publisher Imprint: Independently Published
BISAC: Microprocessors

CUDA Programming from Basics to Advanced teaches you how to write fast, correct GPU code using CUDA 13.4. It starts with how modern NVIDIA GPUs run grids, blocks, threads, and warps, then builds practical kernels step by step.

You will learn the CUDA memory hierarchy, including global, shared, constant, texture, and unified memory, with patterns you can apply to reduce memory traffic and improve occupancy. You will use streams, events, and CUDA Graphs to overlap transfers and compute, then scale from one GPU to multi-GPU and multi-node systems with NCCL.

The book also shows you how to profile and tune real workloads using Nsight Compute 2026, Nsight Systems 2026, and NVML. Worked examples include computer-generated holography, Monte Carlo simulations for optical properties, finite-difference time-domain electromagnetic field simulation, and heat conduction solvers, including CUDA Fortran, plus a clear comparison with OpenCL.

This book suits software engineers, researchers, and students who want a hands-on guide to GPU programming, with focused explanations and ready-to-run code.