July 2019 GPU Hackathon

The July 2019 OLCF GPU Hackathon is coming to Oakland!

If you have an application that you are thinking about porting to GPUs, or an application already using GPUs that could use a helping hand getting that next level of performance, you should consider submitting to the Oakland ORNL GPU Hackathon.

What is a GPU hackathon?

OLCF GPU hackathons are 5-day intensive coding events that put application teams together with experts in programming and performance. The goal of the event is to port and optimize codes on GPU technologies in a focused, highly collaborative environment.

What is the format?

Each code team should consist of 3-5 code developers that together are intimately knowledgeable of the code. If a team proposal consists of a suite of applications, no more than two applications should be sent to the hackathon. For each individual code, at least 2 people must attend. Teams will be complimented by mentors that will be assigned based on expert knowledge that matches the needs of the code.

Selected teams will have access to ORNL's Summit training system, Ascent, and NERSC's Cori-GPU cluster for the duration of the event, including preparation and follow-up testing.

Who should attend?

Any HPC scientific codes seeking performance improvements, or porting to GPU systems in a cooperative, hands-on environment should apply. Codes should primarily be written in a GPU compatible language, such as C++, Fortran and/or Python. Codes can use/want any standard method of offloading work to the GPUs, including CUDA, CUDA Fortran, OpenACC, OpenCL or OpenMP.

The NERSC GPU hackathon is specifically interested in ECP application teams and application teams planning on using NERSC's upcoming CPU-GPU system, Perlmutter. However, all coding teams looking to use GPU technologies are encouraged to apply. A broad range of code types and GPU experience levels are sought for this event.

Participant Costs

There is no registration fee to attend the event. Breakfast, lunch, coffee and snacks will be provided during the hackathon. Participants are responsible for all other expenses, including their transportation to and from the event, lodgings and other meals. Participants will also be expected to arrive with a laptop that is capable of wireless internet to work on for the duration of the event.

Accepted Teams Have Been Notified

Accepted teams should:

• Register all team members for the event.
• Review the attendee guide.

Mentor introductions and computational resource access instructions will be coming soon.

Directions to the Venue:

For local travelers and travelers flying into Oakland International Airport: we recommend using the local Bay Area Rapid Transport (BART) public transportation system. The venue and hotel block is only one block away from the 12th St./Oakland City Center station, which is just 4 stops (~20 minutes) from the airport. Using BART also makes any lodgings near the rail line convenient for the event.

For travelers coming into San Francisco International Airport: the BART system is also a convenient option, but lengthier (~1 hour). If you are looking for a quicker route, we recommend making plans with a van-sharing program. You can search for options at https://www.flysfo.com/to-from/ground-transportation. App-based rides and taxis from SFO are also possible, although they will be more expensive when traveling from SFO.

For those wishing to rent a car or drive to the event: Parking can be difficult and expensive in the Bay Area, so if you plan on driving, plan your parking too. Confirm parking availability with your lodgings and determine where you will park during the event. One good parking option is the Trans Pacific Garage next to the venue. It does not have over-night parking, but is only $20 dollars for the day, $16 if you arrive before 8 AM.

Questions?

You may submit any questions about the GPU Community Hackathon to Kevin Gott: kngott@lbl.gov