Book Image

Hands-On GPU Programming with Python and CUDA

By : Dr. Brian Tuomanen
Book Image

Hands-On GPU Programming with Python and CUDA

By: Dr. Brian Tuomanen

Overview of this book

Hands-On GPU Programming with Python and CUDA hits the ground running: you’ll start by learning how to apply Amdahl’s Law, use a code profiler to identify bottlenecks in your Python code, and set up an appropriate GPU programming environment. You’ll then see how to “query” the GPU’s features and copy arrays of data to and from the GPU’s own memory. As you make your way through the book, you’ll launch code directly onto the GPU and write full blown GPU kernels and device functions in CUDA C. You’ll get to grips with profiling GPU code effectively and fully test and debug your code using Nsight IDE. Next, you’ll explore some of the more well-known NVIDIA libraries, such as cuFFT and cuBLAS. With a solid background in place, you will now apply your new-found knowledge to develop your very own GPU-based deep neural network from scratch. You’ll then explore advanced topics, such as warp shuffling, dynamic parallelism, and PTX assembly. In the final chapter, you’ll see some topics and applications related to GPU programming that you may wish to pursue, including AI, graphics, and blockchain. By the end of this book, you will be able to apply GPU programming to problems related to data science and high-performance computing.
Table of Contents (15 chapters)

Summary

We started with an implementation of Conway's Game of Life, which gave us an idea of how the many threads of a CUDA kernel are organized in a block-grid tensor-type structure. We then delved into block-level synchronization by way of the CUDA function, __syncthreads(), as well as block-level thread intercommunication by using shared memory; we also saw that single blocks have a limited number of threads that we can operate over, so we'll have to be careful in using these features when we create kernels that will use more than one block across a larger grid.

We gave an overview of the theory of parallel prefix algorithms, and we ended by implementing a naive parallel prefix algorithm as a single kernel that could operate on arrays limited by a size of 1,024 (which was synchronized with ___syncthreads and performed both the for and parfor loops internally), and...