GPUs have immensely contributed to various applications: in graphics, AI, scientific computing, you name it. But their programming ecosystem is a mess - in my opinion. You can program GPUs with several frameworks (CUDA, OpenGL, OpenCL, Triton, ...), but they often struggle with portability (e.g. locked to a specific OS or GPU manufacturer) and are generally hard to use (e.g. GPU code has to be written with special programming languages, optimization is complicated, ...). This video presents a brief history of GPU programming and how to develop GPU-accelerated software. To make it more approachable, I decided to focus on the main features of the frameworks instead of explaining the code in detail. Of course, you can view the full source code that I used to make the video in the repository linked below, and even run it on your system! Full source code: Script: References: #references Chapters: - Introduction: 00:00 - How GPUs Work: 01:10 - Graphics APIs: 03:45 - General-Purpose APIs: 09:50 - The Future: 14:10