Software Week2016: Intro to GPU Programming

The use of graphics cards as general purpose computation accelerators have become a common way to greatly increase the speed of heavy computations, both on clusters and on personal workstations. For the right workloads you can get ten times speed improvements or much more.

We will give a brief introduction to CUDA, a C language extension for programming NVIDIA GPUs, and show how to build and run GPU code on the OIST clusters. You need no experience with parallel programming, but brief previous experience with C or C++ is recommended.

※A laptop is required, and please make sure that you can access the Tombo cluster.