CUDA Programming from Basics to Advanced - Paperback

by Finbarrs Oketunji (Author)

The book explores CUDA programming from fundamental concepts to advanced techniques. It covers the latest CUDA 12.6 environment, detailing GPU hardware evolution and parallel computing enhancements. It offers practical insights into memory management, utilisation of GPU-compatible libraries, and tackling computational bottlenecks in various scientific applications. Additionally, it includes chapters on OpenCL, performance tuning with Nsight Compute 2024, and debugging at scale with Nsight Systems 2024. With code samples and appendices, it serves as a priceless resource for novice and experienced programmers.

The book is divided into 13 chapters:

1. Introduction to GPU and CUDA Programming
2. Setting Up and Running CUDA 12.6
3. CUDA Program and Memory Hierarchy
4. Utilising GPU-Compatible Libraries
5. Tackling Computational Bottlenecks - Computer-Generated Holography
6. Conditional Branching in Simulations - Monte Carlo Method for Optical Properties
7. Overcoming Memory Access Bottlenecks - Electromagnetic Field Simulation Using FDTD Method
8. Fortran Implementation in CUDA - Numerical Solutions to Heat Conduction
9. GPU Programming with OpenCL
10. Using Nsight Compute 2024 for Performance Tuning
11. Debugging at Scale with Nsight Systems 2024
12. Appendix A: Extended Sample Programs for Numerical Calculations
13. Appendix B: Further Reading

After reading this book, you will know the following and much more:

- How to write CUDA programs for various real-world applications
- Techniques for managing memory, control flow, and parallelism in CUDA programming
- Methods for working with GPU-compatible libraries and optimizing performance
- Strategies for debugging, profiling, and maintaining CUDA applications
- Ways to integrate CUDA with other programming languages and tools like Fortran and OpenCL
- Best practices for GPU programming and optimization in modern high-performance computing environments