Lucas Timotheo Sanches, a postdoctoral researcher at Louisiana State University in Baton Rouge, was about 12 years old when his father, an electrical engineer, started to teach him how to program computers. That experience and, later, some inspirational teachers led Sanches to become enamored with physics in high school. 

He then learned to combine his love of general relativity, computer science, and high-performance computing. He has since earned a Ph.D. in physics and continues to “play around” with computers. So, when he learned about the Argonne Training Program on Extreme-Scale Computing (ATPESC), he applied. 

He was delighted to find that the hands-on tutorials at ATPESC provided good, practical learning by example.

“We actually did things instead of just listening to the speakers,” he said.

Many early career scientists, such as Sanches, are eager to learn how to use the world’s most powerful supercomputers. They also anticipate acquiring essential skills for their careers in the computational science and engineering worlds by participating in ATPESC.

The U.S. Department of Energy’s (DOE) Argonne National Laboratory plays a key leadership role in training the future high-performance computing (HPC) workforce. The annual ATPESC offers two weeks of intensive training on key skills, approaches, and tools to carry out research on current and next-generation supercomputers. More than 900 attendees have participated in ATPESC since it was launched in 2013. This year’s program was held in late July and early August in St. Charles, Illinois. 

ATPESC offered curriculum that covers all aspects of supercomputing, including hardware, software, frameworks, visualization, artificial intelligence (AI), machine learning, and data-intensive computing. 

ATPESC covers a broad range of topics, and its purpose is, in fact, to fill in the knowledge gaps of the attendees, said Raymond Loy, ATPESC director and Argonne Leadership Computing Facility (ALCF) lead for training, debuggers, and math libraries. The ALCF is a DOE Office of Science user facility.

Each day brings both different material and a different subset of attendees who have a strong focus on that topic. That all depends on their backgrounds, their current research areas, and their inherent curiosities about the topic.

“Given the current importance of AI, I'm sure there will be some adjustments to increase coverage of that area in the future,” said Loy.

This year's group of 74 attendees had many doctoral students, followed by postdocs and early career researchers, including several from industry.

“As HPC has come to the forefront of science, there has been a trend over time of applicants being qualified earlier in their careers in comparison to the early days of ATPESC,” said Loy. 

Rui Fang, a graduate teaching assistant in the University of Pittsburgh’s Mathematics Department, has worked on computational fluid dynamics and big data projects and sought to use high-performance resources to complete tasks. It wasn’t until some friends nudged her to apply to ATPESC that she discovered the high-quality speakers and all the resources that were available to her.

“I also enjoyed meeting people, the networking, that just happened to connect me with someone who felt that my research could be useful to them,” she said. “This has turned out to be one of the best summer programs I have ever attended, with people from different backgrounds who support each other.”

Ilker Topcuoglu, a computational scientist and aerospace engineer at DOE’s National Renewable Energy Laboratory in Golden, Colorado, saw what his work colleagues learned from ATPESC and how they used it in their professions.

“Some things about the program are not really obvious, but you learn which other pathways you can take to turn your project or idea into reality,” said Topcuoglu. 

Topcuoglu also learned by talking with others at the program about hardware and the more concealed elements of programming models. Each day brought new information. 

“The program teaches you about all layers of scientific computing that you may not necessarily be aware of,” he said.

Jyotsna Rajaraman, a doctoral student at the University of Illinois Chicago and a research aide at Argonne, especially enjoyed hearing what her peers were learning about and what they were doing on large-scale machines. ATPESC also helped her to learn more efficiently. 

“The debugging software and high-performance toolkit were especially helpful,” she said.

Sagnik Singha’s father was a physics major in college and had a great understanding of particle physics. When his father came home each day, he would teach his son about what he learned. It inspired the younger Singha in so many ways; he went on to earn a master’s degree and doctorate in mechanical engineering at Texas Tech University. He is now a research associate at the university’s High-Performance Computing Center.

He also aimed to continue his education outside of the university by taking his colleagues’ advice to apply for ATPESC.

“The depth and breadth of topics covered is admirable and this is as hands-on as it gets.” Sagnik said.

While dozens of experts from Argonne and other national laboratories in the field of HPC provided lectures, many offered tours of Argonne's facilities or hands-on demonstrations, Singha noted.

“But when you hear them speak, you learn how to push the boundaries because they were pushing the boundaries,” he said. “You also learn about the pitfalls and how they circumvented them. It was great to know what their interpretations were and what they think they see ahead.”

Lev Gorenstein enjoyed being a first-time ATPESC speaker. He is a solutions architect at Globus, a research cyberinfrastructure service provided by the University of Chicago that offers secure and reliable file transfer, sharing, and data management automation for researchers. 

“From preparation to execution, everything was very well thought-out, unexpected eventualities handled quickly and efficiently, and everything felt like a well-oiled machine. And the overall atmosphere … talented students, great speakers, just an absolute intellectual paradise,” said Gorenstein.

John Holmen, an HPC engineer at DOE’s Oak Ridge National Laboratory’s Oak Ridge Leadership Computing Facility (OLCF), was himself an ATPESC attendee in 2015 and returned this year as a presenter. The OLCF is a DOE Office of Science user facility.

“It was a lot of fun to return as a presenter as I knew how foundational of an experience it can be for attendees and was able to share my excitement for them firsthand,” said Holmen. “Overall, the presentation went well with lots of great and thought-provoking questions. The experience was quite rewarding as you could tell by the questions and general enthusiasm that participants were excited to learn.”