GPU Programming Bundle

This book takes you from the fundamentals of GPU architecture and the Linux graphics stack all the way to designing, implementing, and debugging a complete Direct Rendering Manager driver for the Linux kernel. If you can write C and understand the basics of kernel modules, this book will teach you how to bridge the gap between graphics hardware and userspace applications through production-quality Linux kernel code. Every chapter builds on the previous one, progressing from a minimal loadable module to a fully functional DRM driver with memory management, command submission, synchronization, display support, and error recovery. Real code, real build instructions, no shortcuts.

GPU parallel computing has fundamentally changed how software achieves performance, but writing kernels has historically demanded deep expertise in low-level hardware details and C++ era programming patterns. This book shows Rust developers how to harness NVIDIA GPUs using CUDA Rust, the emerging ecosystem that brings memory safety, expressive type systems, and modern tooling to GPU programming.

You will learn both of NVIDIA's official CUDA Rust tracks: the SIMT-style cuda-oxide compiler backend for fine-grained thread-level control, and the tile-based cutile-rs framework for safe, high-level kernel authoring. Along the way, you will build a solid understanding of GPU architecture, memory hierarchies, synchronization, and performance optimization. Every chapter includes complete, runnable code examples and explains the why behind each concept, not just the how. This is a practical guide for Rust programmers ready to write production-quality GPU kernels without sacrificing the safety and clarity that make Rust compelling.

This book is a technically rigorous treatment of GPU virtualization on Linux, focused on the KVM/QEMU stack and the emerging virtio-nvgpu architecture for shared NVIDIA GPU access in virtual machines. It covers the complete technology stack from hardware fundamentals through production deployment, with working configurations, architecture diagrams and source code analysis drawn from upstream Linux kernel, QEMU and NVIDIA implementations. The audience is experienced systems engineers who need to understand not just how to deploy GPU virtualization, but how the entire stack works internally.