NERSCPowering Scientific Discovery for 50 Years

NVIDIA HPC SDK - OpenMP Target Offload Training, December 2020

December 8, 2020

NVIDIA will present a 2-part training series for NERSC and OLCF users about using OpenMP target offload with NVIDIA’s HPC SDK compilers. The training will introduce OpenMP target offload, the NVIDIA compilers, and best practices for achieving high performance with OpenMP target offload on NVIDIA GPUs.  Access to Cori GPU nodes will be provided.


Date & Time Topics and Speakers

Day 1, Tuesday, December 8

9:00 am - 10:30 am PST

  • Overview of OpenMP Target Offload project (Michael Wolfe)
  • Best practices for NVIDIA GPU’s using OpenMP Target Offload directives (Guray Ozen)
  • Compiler Usage, system requirements, runtime environment variables, etc… (Simone Atzeni)
  • Live Demo, homework assignment (Scott Biersdorff)

Day 2, Thursday, December 10

9:00 am - 11:30 am PST
  • Review homework assignment (Scott Biersdorff)
  • Code Example walk-throughs (Scott Biersdorff, Mat Colgrove)
  • 10-min break
  • Interoperability with CUDA, with CUDA Fortran, with OpenACC (Brent Leback)
  • NVLAmath Introduction (Yu You)


Presentation Materials