Maximizing volunteer computing systems to solve big problems

March 15, 2022

Diagram shows three main thrusts of Pai's CAREER award
Sreepathi Pai, an assistant professor of computer science, will develop new automatic performance tuning algorithms to maximize the GPU performance in volunteer computing systems to accelerate scientific discovery in fields such as medicine, biology, and astronomy.

Sreepathi Pai’s algorithms will help disparate GPUs work efficiently together

People all over the world donate time on their computers to large-scale volunteer computing systems such as Einstein@Home and Folding@Home.

Why? In order to solve important scientific problems.

“There’s a strong sense of altruism, and the ability to contribute to science motivates people to do this,” says Sreepathi Pai, an assistant professor of computer science at the University of Rochester. His research aims to make it easier to write high-performance programs on increasingly complex machines.

The computational power delivered by these volunteer computing systems rivals some of the most powerful supercomputers in the world. However, volunteer systems consist of hundreds of different kinds of graphics processing units (GPUs) used to speed up scientific calculations. This heterogeneity hinders maximum performance.

Supported by a prestigious National Science Foundation CAREER award, Pai will develop new automatic performance tuning algorithms to maximize the GPU performance in volunteer systems.

“Our primary concern is speed,” Pai says. “So, if you have a program, but it can run 20 percent faster by changing a few parameters on a particular machine, that’s what we want to do.

“By developing autotuning algorithms that can deal with the heterogeneity, providing feedback to programmers about unrealized performance, and working at scale in real volunteer computing systems, this project will enable these systems to maximize GPU performance and accelerate scientific discovery in fields such as medicine, biology, and astronomy.”

Robot turtles and classic electronic games broaden the impact

The Faculty Early Career Development (CAREER) award is NSF’s most prestigious recognition for early-career faculty members. It provides recipients with five years of funding to help lay the foundation for their future research. It also requires recipients to demonstrate how their research can have a broader impact to benefit society.

Pai will broaden the impact of his project in two ways.

Working with the University’s Kearns Center, Pai has already been developing a better way to teach programming to Upward Bound high school students from the Rochester City School District. The students attend the center’s summer STEM programs.

Popular block-based beginner’s coding tools taught. to the students, such as Scratch and Blockly, “are very different from the languages most professional programmers use, which is mostly just text,” Pai says. Last summer, he had success “bridging the gap” by using LOGO, an older, more text-based programming widely known for its use of turtle graphics.

This summer, if supply chain issues don’t interfere, Pai would like to introduce an actual turtle robot that students can program to do various tasks.

“That way students won’t think of computing as something where you just sit at your keyboard, typing, but as something that can have a real effect on the world around them,” Pai says. “I think that could really motivate them.”

Pai also hopes to explore ways to help Rochester’s famed Strong National Museum of Play preserve its collection of original electronic games—including Pong, Space Invaders, Grand Theft Auto, Mario Brothers, Pac-Man and other classics in its Word Video Game Hall of Fame.

“It’s an amazing collection,” Pai says. “They have all the original material, including cartridges, CD roms, and actual arcade machines for many of these games.”

One approach would be to create emulators—which enable one computer system to behave like another system—to save wear and tear on the original games, Pai says.