A stand-alone short course in GPU computing for statisticians
Day: Monday, September 9Time: 2:10 PM - 4:00 PM
Place: Snedecor Hall 2113
This is a stand-alone, 2-hour crash course in GPU computing for statisticians. It is meant to be an alternative to the 11 main talks listed at http://will-landau.com/gpu/talks.html. Although intended mainly for faculty, this talk welcomes everyone.
Viewers will be introduced to GPUs in general, the Iowa State Statistics Department's GPU clusters, parallelizing statistical algorithms in CUDA C, and NVIDIA's GPU-accelerated CUDA C libraries. Examples include pairwise summation, K-means, Markov chain Monte Carlo, and ordinary least squares. These examples will showcase some best practices in CUDA C programming, including error checking with Valgrind and CUDA MEMCHECK.
Materials
SlidesVideo
Example Code
MakefileMakefile
Makefile