People - Written by on January 4, 2017

Mini Hackathon Offers Shorter Schedule, Introductory Focus

Tags: , , ,

Brent Leback of NVIDIA, left, advises a GPU Hackathon team on accelerating its application at the OLCF’s weeklong hackathon in October. This year, the OLCF also offered the first mini-hackathon in November to accommodate record-high attendance.

OLCF debuts new way to learn about programming for GPUs

In November, the Oak Ridge Leadership Computing Facility (OLCF) hosted its first 3-day mini GPU Hackathon, an extension of the center’s annual 5-day GPU Hackathon, which began in 2014. The mini hackathon took place November 1–3 at the Crowne Plaza in Knoxville, Tennessee, and was open to the public.

“The 5-day hackathon was oversubscribed, but we wanted everyone who wanted to participate to have the chance,” said Fernanda Foertter, OLCF high-performance computing (HPC) user assistance specialist. The OLCF is a US Department of Energy (DOE) Office of Science User Facility located at DOE’s Oak Ridge National Laboratory (ORNL).

Only in its first year, the mini hackathon drew 41 attendees, many of whom had initially submitted applications to the weeklong event that accommodated 61 attendees.

At such hackathons participants work intensely to code applications for a hybrid CPU/GPU architecture—namely, Titan, the OLCF’s 27-petaflop supercomputer.

GPU accelerators can offer significant performance boosts, particularly for scientific simulations that benefit from running many similar calculations at the same time. However, accelerated architectures have some unique programming requirements compared with CPU-only architectures, and hackathons are a fast way to acquaint users with programming for GPUs. Most OLCF hackathon participants are scientists who use modeling and simulation on Titan to solve large-scale scientific problems, but anyone with an interest in programming for accelerated computing is welcome to register.

Unlike the full hackathon, which typically draws teams with some experience using GPUs, this year’s mini hackathon was geared toward participants with little to no experience accelerating applications for GPUs.

“Some people worked on their own applications. Others didn’t have an application but got a feel for what they might need,” Foertter said. “We had several researchers from around the laboratory [ORNL] who aren’t necessarily experienced HPC users. This is an opportunity to grow the HPC community.”

Foertter and Jeff Larkin of the OLCF and Brent Leback of NVIDIA, the GPU vendor partner for Titan, led the event’s lectures and hands-on sessions. Instruction focused on programming techniques using OpenACC, an application programming interface designed for accelerated architectures.

“The information on OpenACC was really helpful,” said Hilda Klasky, a software engineer in the ORNL Modeling and Simulation Group whose work focuses on programming for CPUs. “I wasn’t that familiar with OpenACC but knew parallelism can improve application performance, so I wanted to explore how to program for GPUs.”

Foertter said the OLCF plans to host both the 5-day and mini hackathons again next year and may also develop a series of webinars for participants who want to join remotely.

“Not everyone can take time out of their schedule to attend in person, so we want to offer training that is flexible and accessible,” Foertter said.

Visit the OLCF website for more information on upcoming training events.

Oak Ridge National Laboratory is supported by the US Department of Energy’s Office of Science. The single largest supporter of basic research in the physical sciences in the United States, the Office of Science is working to address some of the most pressing challenges of our time. For more information, please visit