Here are the slides from my presentation from late 2019 giving an overview of how GPUs can accelerate scientific code, with CUDA and OpenACC. Examples mostly in hydrodynamics simulations in the context of astrophysics.

Abstract: Graphics processing units (GPUs) aren’t just of interest to gamers and cryptocurrency miners. Increasingly, they’re being used to turbocharge academic research, too. GPUs provide massive parallelism enabling one to perform computations on a basic computer that until very recently required a supercomputer. I will describe how GPUs can be leveraged to dramatically accelerate scientific calculations. I will also give an overview of recent applications of GPUs to astrophysical simulations, with a particular focus on radiative transfer in curved spacetimes and (magneto)hydrodynamics simulations of black hole accretion.