The use of GPGPUs for scientific programming is getting increasingly popular, and the number of applications that support GPUs increases everyday. However, taking advantage of this new paradigm often requires rewriting your code.
A promising alternative has emerged by PGI, Cray, and Nvidia called OpenACC. OpenACC allows you to take a directive based approach to minimally modify your C or Fortran code, allowing existing applications to run on GPUs and other accelerators (e.g. Xeon-Phi). The broader objective of OpenACC is to create a generic, high-level device-agnostic programming languge similar to OpenMP, but able to run on GPUs, CPUs, Xeon-Phi cards, etc.
In this workshop, we will introduce the basics of OpenACC and discuss how to convert your C or Fortran code by working through examples. We will assume that you have some background in C or Fortran programming.
This is an interactive, hands-on workshop. It will be held in a fully-equipped classroom laboratory with computers. Attendees may bring their own notebook computers, but this will not be necessary.
- Introduction to OpenACC
- Modifying existing code for use with OpenACC
This workshop will be led by Prasad Maddumage, RCC Application Specialist. Dr. Maddumage has over nine years of parallel scientific code development experience using a variety of platforms and technologies.
The location of these workshops is Dirac Science Library Room 152 (map). NOTE: The entrance is located on the southeast corner of the building, across from the Oceanography building next to the parking garage. Use that entrance, not the main library entrance.