Introduction to GPUs and CUDA programming

Preface

This short course aims to equip readers and participants with an understanding of GPU architecture and considerations for programming accelerated workloads effectively.

In an era where computational demands are soaring for training large machine learning models to rendering complex graphics, understanding GPU accelerators and how to program them effectively is an essential skill.

In this short course we will cover just enough to build a strong awareness of how accelerators work and how to approach their parallel programming paradigm in C and Python.

Topics we will cover include:

• Parallel vs sequential execution

• Thread hierarchy and organization

• CUDA kernel functions: threads, blocks and grids

• Machine learning performance optimization

If you have requests or suggestions, please submit a pull request here.