Hackathons to Train New Generation on GPU Accelerated Computing
If you’re willing to learn fast, here’s your chance to see your code run fast. The Oak Ridge National Laboratory has organized a series of five hackathons to help people learn how to accelerate their projects with GPUs.
The event’s organizers are looking for current or prospective user groups of big hybrid CPU-GPU systems. Teams of three or more developers with an application that could benefit from GPU accelerators, or one running on accelerators that needs optimization, are encouraged to participate.
If that describes you, the five-day events will help you take advantage of the latest generation of GPU-accelerated supercomputers, which speed up a wide range of applications.
You won’t need any advanced GPU knowledge to join. Just come with science codes you want to have accelerated on GPUs. Prior to the hackathons, participants can get a head start with online training in OpenACC, a standard that simplifies parallel programming on heterogeneous CPU-GPU systems.
Mentors will help guide teams on how to start with GPUs — it can be as easy as using libraries. Or start with OpenACC directives designed to simplify GPU work for scientists and researchers, or CUDA for more flexibility and control.
The hackathons will take place between March and October at the Juelich Supercomputing Center in Germany; the Brookhaven National Laboratory in Upton, New York; the Swiss National Supercomputing Center; the NASA Langley Research Center in Hampton, Virginia; and the Oak Ridge Leadership Computing Facility in Tennessee.
First Hackathon Kicks Off in March
The first hackathon is scheduled for the beginning of March at the Juelich Supercomputing Center (Forschungszentrum Jülich). The event’s organizers originally planned for eight teams, but were able to accommodate 10 of the 14 teams that applied.
In total about 60 people are expected to join for one week of hacking. “We want to accept all teams that apply,” said Fernanda Foertter, creator of the hackathon series. “But we are limited by mentors and space, so we do the best we can. This high demand is what led us to increase the number of offerings.”
“The GPU hackathon is a unique opportunity for the domain scientists at Forschungszentrum Jülich to start fully exploiting the performance of GPUs for their science,” said Dirk Pleiter, a professor of theoretical physics at Regensburg University and one of the organizers of the event.
For more information, and to submit your proposal, visit the Oak Ridge Leadership Computing Facility’s website.
And to get a head start on programming for GPUs, visit our “3 Steps to More Science” video tutorial to start on GPUs with OpenACC and libraries.