Foundations of AI: A multi-GPU softmax function in Chapel
AI relies on optimized math done on GPUs. In this video, we demonstrate how the crucial softmax function can be implemented in Chapel and run on many GPUs. We also show off more of Chapel's parallel-first language features, which are essential to writing portable, readable, and maintainable GPU code.
Check out the code at: https://github.com/jabraham17/hpe-dev...
This demo was part of larger talk about GPU programming in Chapel, available here: • Video
A playlist containing all five demos is available here: • Demos of GPU Programming in Chapel