Getting Ready to Use OpenMP in Exascale Computing

An article on phys.org by Ariana Tantillo of Brookhaven National Laboratory (BNL) describes activities during the recent four-day OpenMP hackathon (aka “OpenMP Brookathon 2019”) hosted at BNL by the lab’s Computational Science Initiative (CSI).

Eight teams who were chosen based on their extensive experience in programming worked with national lab, university, and industry mentors to learn about the changes they should expect in OpenMP and how the differences can be advantageous to their scientific applications.

The teams worked on porting their scientific applications from CPUs to GPUs and on optimizing them using the latest OpenMP version (4.5+). They also learned about OpenMP features via tutorials. In addition, some teams had the opportunity to bring to bear the latest OpenMP functionalities to program IBM Power9 CPUs accelerated with NVIDIA GPUs.

The US Department of Energy’s Exascale Computing Project (ECP) sponsored the event, which was driven by ECP’s SOLLVE project and jointly organized by Oak Ridge National Laboratory and IBM. The OpenMP hackathon was the latest in a series of such events offered by CSI.