CUDA C++ Debugging: Safer GPU Kernel Programming

by David Spuler

This book covers CUDA C++ programming tools and techniques for safely running GPU kernels in the NVIDIA CUDA C++ environment, with coverage from beginner to advanced. Improve reliability without sacrificing performance and reduce development time by finding, fixing, and forgetting GPU coding errors.

Main
- Debugging techniques for CUDA C++ kernels
- Safely run CUDA C++ kernels without losing speed
- Common CUDA C++ bugs from beginner to advanced
- CUDA tools for debugging and memory checking
- Shaking out more bugs with self-testing code
- Tolerating and recovering from errors
- Prevention of bugs with resilient coding

Table of
Part Introduction
1. CUDA Introduction
2. Debugging Hello World
3. CUDA for C++ Programmers
4. CUDA Emulation
5. Debugging Simple Kernels
Part Debugging CUDA C++
6. Debugging Strategies
7. CUDA Debugging Tools
8. Error Checking
9. Sticky Errors
10. GPU Kernel Debugging
11. Basic CUDA C++ Bugs
12. Advanced CUDA Bugs
Part Advanced CUDA Debugging
13. Self-Testing Code
14. Assertions
15. Debug Wrapper Functions
16. Debug Tracing
17. CUDA Portability
CUDA Puzzles

210 pages, Paperback

Published October 13, 2024