Supercomputer researcher wins Turing Award for boosting ultra-powerful machines





IBM Summit supercomputer

Oak Ridge National Laboratory’s Summit supercomputer, built by IBM and currently the second fastest machine in the world

Oak Ridge National Laboratory

Jack Dongarra, a researcher at Oak Ridge National Laboratory in Tennessee, won the 2021 Turing Award for software that has unlocked the power of the world’s largest computers. Many previous recipients have indirectly promoted supercomputers, but Dongarra has specialized in code that speeds up scientific computations and runs them on machines with thousands of processors.

“These contributions provided a framework from which scientists and engineers made important discoveries and groundbreaking innovations in areas such as big data analytics, healthcare, renewable energy, weather forecasting, genomics and economics,” the Association for Computing Machinery said Wednesday of its choice to name Dongarra. the winner. The prestigious prize, which has been awarded $1 million from Google since 2014, is often referred to as the computer equivalent of the Nobel Prize.

The award comes as supercomputing is becoming increasingly important in understanding the world and dealing with its challenges. Supercomputers are often used to predict both next week’s weather and the course of climate change in the coming decades. They are being deployed to simulate the high-energy physics of aging nuclear weapons and fusion reactors, which scientists hope will yield abundant electricity in the future. The industry uses the hugely powerful machines to design planes that are more efficient.

Supercomputer researcher Jack Dongarra

Supercomputer researcher Jack Dongarra

Tara Kneiser/Association for Computing Machinery

Good software is critical to taking advantage of the vast supercomputing hardware that today consists of thousands of machines with millions of processing cores. Dongarra wrote algorithms that accelerated the algebraic calculations at the heart of most scientific computer science. He also helped develop a message passing interface, called MPI for short, that is needed to distribute a single computing task across thousands of processors.

Dongarra currently works at Oak Ridge, home to the IBM-built Summit machine that was the fastest supercomputer in the world for two years, until Japan’s Fugaku took the crown. The Energy Department’s lab also houses the Frontier machine, which could bring that crown back when it’s fully operational in 2023.

“I’m still in shock at the news about the award, to be honest,” Dongarra, who has written papers with past Turing winners, said in an interview. “It’s a great honor and I feel humbled.”

Since 1966, ACM has presented the Turing Award to the greatest stars of computer science. The recipients’ research paved the way for every commercial success in the industry, from IBM PCs and Apple iPhones to Google Search and the Call of Duty video game.

Previous Turing Awards have gone to the inventors of the encryption technology that protects communications and e-commerce, the chip design used in every mobile phone, the World Wide Web, the graphical user interface on personal computers, today’s artificial intelligence technology and the Unix operating system. .

Today, Dongarra is intrigued by new trends in supercomputing. One is cloud computing services that allow anyone to rent access to massive systems. Another is artificial intelligence technology, which he says can be useful for quickly finding approaches that can be verified later by traditional means.

In the 1970s, Dongarra began an important software project, Linpack, to help researchers solve linear equations more easily on different computers. He kept a list of how well various machines performed, including famous machines like the Cray-1 supercomputer and everyday minicomputers like Digital Equipment Corporation’s PDP-8.

That list, combined with a related list by Hans Mayer of the University of Mannheim, became the Top500 ranking for supercomputers published twice a year since 1993.

Linpack has shown remarkable longevity, in part because it has been updated many times as supercomputers evolved to include new technologies — for example, clusters of cheaper Linux machines from the 1990s, graphics processing units from companies like Nvidia over the past decade, and now AI accelerators to speed up machine learning tasks.

But for the measure of speed, it’s time Linpack’s reign ends, Dongarra said.

“We need new benchmarks,” he said. Modern machines today are more limited by their ability to commute data internally. Dongarra prefers the more recent HPCG speed test, which also reports the Top500.




Leave a Comment

x