../_images/gpu_hackathons.jpg

OLCF GPU Hackathons

Each year, the Oak Ridge Leadership Computing Facility (OLCF) works with our vendor partners to organize a series of GPU hackathons at a number of host locations around the world. The table below lists the OLCF-supported events scheduled for 2020.

Location (virtual) Proposal Deadline Event Dates
San Diego Supercomputing Center - San Diego, CA March 11 April 27, May 11 - 13
Princeton University - Princeton, NJ March 2 June 1, 8 - 10
NERSC - Oakland, CA May 13 July 7, 13 - 15
Brookhaven National Laboratory - Upton, NY June 28 August 10, 17 - 19
OLCF - Knoxville, TN August 6 October 19, 26-28

Note

The OLCF-supported events are a subset of a larger number of GPU hackathons organized within the community. For the full list, please visit gpuhackathons.org

What is a GPU hackathon?

A GPU hackathon is a multi-day coding event in which teams of developers port their own applications to run on GPUs, or optimize their applications that already run on GPUs. Each team consists of 3 or more developers who are intimately familiar with (some part of) their application, and they work alongside 1 or more mentors with GPU programming expertise. Our mentors come from universities, national laboratories, supercomputing centers, and industry partners.

Note

There are a variety of programming models available to program GPUs (e.g. CUDA, OpenACC, OpenMP offloading, etc.) and your are welcome to use any of them at these events.

Why participate?

If you want/need to get your code running (or optimized) on a GPU-accelerated system, these hackathons offer a unique opportunity to set aside the time, surround yourself with experts in the field, and push toward your development goals. By the end of the event, each team should have part of their code running (or more optimized) on GPUs, or at least have a clear roadmap of how to get there.

Target audience

We are looking for teams of 3-6 developers with a scalable** application to port (or optimize on) GPUs. Collectively, the team should know the application intimately.

[** We say scalable here because we’re typically looking for codes intended to run on multiple nodes (e.g. MPI-enabled), although porting/optimizing such codes on a single node during the events is encouraged whenever possible.]

Note

Please keep in mind that we are looking for teams with plans to develop GPU code - not to simply run their code on GPUs.

Virtual hackathon format

Typically, these hackathons are in-person events, where each team (app developers + mentors) sits at their own round table in a single large conference room. This structure allows teams to hack away on their own codes, but also to interact (ask questions, give advice, etc.) with members/mentors from other teams when needed.

To recreate this environment in a virtual setting, we will be using Zoom + Slack. Zoom will be used as the main online tool due to its breakout room capabilities; there will be a single Zoom session, where the main room will be used for presentations, and breakout rooms will be used by individual teams (for screen sharing and verbal communication). We will also set up a Slack workspace for communication between all participants, and individual team channels for communication within teams (chat, sharing code snippets, etc.).

Ok, so how do I attend?

First, you must decide which event you’d like to attend (see table of events above for locations and proposal deadlines), and then submit a short proposal form describing your application and team. The organizing committee will then review all proposals after the call for that event closes and select the teams they believe are best suited for the event.

To submit a proposal, please visit gpuhackathons.org/events, click on the event you’d like to attend, and submit the form.

Want to be a mentor?

If you would like more information about how you can volunteer to mentor a team at an upcoming GPU hackathon, please visit our Become a Mentor page.

Who can I contact with questions?

If you have any questions about the OLCF GPU Hackathon Series, please contact Tom Papatheodore (papatheodore@ornl.gov).