Efficient data parallel computing on GPUs

Abstract

In this part of the course, we'll look at some tricks and traps for programming the GPU for general purpose computation more effectively. This first part of this section will attempt to get you, the CPU programmer, to start to think in "GPU terms." There are some mistakes that every beginning GPU programmer (even those who are experienced CPU programmers) seem to make; those are the things I'm aiming to address in this part of the talk. After this introduction to "GPU thinking," Aaron and Ian will delve into more details about exactly how you can sculpt your algorithms so that they make the best use of what the GPU has to offer.