You're reading from GPU-Accelerated Computing with Python 3 and CUDA From low-level kernels to real-world applications in scientific computing and machine learning

Product type Paperback

Published in Mar 2026

Publisher Packt

ISBN-13 9781803245423

Length 534 pages

Edition 1st Edition

Languages

Python

Tools

CUDA

Concepts

Programming Language

Authors (2):

Niels Cautaerts

Hossein Ghorbanfekr

View More author details

Table of Contents (24) Chapters

Preface

Free benefits with your book

1. Part 1: Fundamentals of GPU programming with CUDA in Python 3

2. Chapter 1: Why GPU Programming with CUDA in Python 3? FREE CHAPTER

3. Chapter 2: Setting Up a GPU Programming Environment Locally and in the Cloud

4. Chapter 3: Writing and Executing CUDA Kernels with Numba-CUDA

5. Chapter 4: Profiling and Debugging CUDA Code

6. Part 2: Performance Optimization and Advanced CUDA Topics

7. Chapter 5: Optimizing the Performance of CUDA Code

8. Chapter 6: Enabling Concurrency Using CUDA Streams

9. Chapter 7: Scaling to Multiple GPUs

10. Part 3: Using High-Level Python Libraries for GPU Computation

11. Chapter 8: Bringing NumPy and SciPy to the GPU with CuPy

12. Chapter 9: Bringing pandas and scikit-learn to the GPU with Rapids

13. Chapter 10: Solving Optimization Problems on the GPU with JAX

14. Part 4: Real-World Example Applications

15. Chapter 11: Solving the Heat Equation on the GPU

16. Chapter 12: Image Processing and Computer Vision on the GPU

17. Chapter 13: Simulating Atomic Interactions on the GPU

18. Chapter 14: Implementing Your Own Transformer-Based Language Model

19. Part 5: Beyond This Book

20. Chapter 15: Expanding and Deepening Your GPU Programming Knowledge

21. Chapter 16: Unlock Your Exclusive Benefits

Unlock this Book's Free Benefits in 3 Easy Steps

22. Other Books You May Enjoy

Subscribe to Deep Engineering

23. Index

NVIDIA Nsight profiling tools

NVIDIA Nsight provides a suite of specialized profiling tools designed to help understand and optimize the performance of CUDA and GPU-accelerated applications. Here, we will focus on Nsight Systems and Nsight Compute for performance analysis. Nsight Systems offers a high-level timeline-based view of application execution as it visualizes how CPU threads, CUDA kernels, and memory transfers interact over time. It helps identify performance bottlenecks such as synchronization delays, data transfer overheads, and suboptimal kernel launch patterns. This makes Nvidia Systems ideal for analyzing application flow, concurrency, and overall GPU utilization. Nsight Compute focuses on low-level kernel performance analysis. It provides detailed metrics on GPU hardware utilization, such as memory bandwidth, occupancy, instruction throughput, and cache efficiency, allowing us to pinpoint and optimize performance issues within individual kernels.

Legacy NVIDIA profiler...