NERSCPowering Scientific Discovery for 50 Years

Introduction to OpenMP Device Offload, Sept 22-23, 2021

September 22, 2021

Introduction

Oak Ridge Leadership Computing Facility (OLCF) and NERSC will offer a (virtual) Introduction to OpenMP GPU Offloading. This training is meant for users who are already familiar with the basic ideas of GPU programming but who want to learn about the core GPU offloading capabilities of OpenMP.

Date and Time: 10 am - 2 pm (Pacific time), Sept 22 - Sept 23 (Wednesday - Thursday)

During each day of the training, OLCF/NERSC staff will give a lecture followed by hands-on exercises for the participants to work through. On day 1, we will give a general overview of the OpenMP programming model and cover the basics of using OpenMP directives to offload work to GPUs. On day 2, we will discuss optimization strategies and show how efficient data movement and a better understanding of the hierarchy of parallelism available can lead to improved performance.

The hands-on exercises will be available in both C and Fortran and will be designed to work with a variety of compilers on both the OLCF and NERSC Cori GPU systems. Time will be reserved for users to work through the examples during the training and a slack channel will be available for questions during and after each training session.

Agenda

All times are given in PDT

Day 1, September 22, Wednesday

  • 10:00 am -11:00 am    Introductions to OpenMP Offloading
  • 11:00 am -12:30 pm    Hands-on examples
  • 12:30 pm - 2:00 pm     Help available on slack

Day 2, September 23, Thursday

  • 10:00 am - 11:00 am    The Memory model and Efficient Data Transfer.
  • 11:00 am - 12:30 pm    Hands-on examples
  • 12:30 pm - 2:00 pm      Help available on Slack

Registration

Please click the "Registration" drop-down on this page to register.  

Remote Connection Information 

Please click the "Remote Connection Details" and "Slack Channel Link" drop-down menus on this page for more information.

Presentation Materials