Laid-back Numerical Computation Researcher Seeks to Accelerate Supercomputers | UTOKYO VOICES 030
Laid-back Numerical Computation Researcher Seeks to Accelerate Supercomputers
Although he is a numerical computation researcher, Professor Reiji Suda says frankly, “I’m not good at calculating things.”
What he goes on to say, with a smile, makes you doubt your own ears. “Math was also one of my worst subjects. I think my score on the university entrance exam was about 20 out of 100.”
That’s not all he finds difficult. He’s also utterly poor at memorization, and he’s all thumbs. In fact, the very reason he’s not good at calculating things is that the numbers he writes are so messy that he himself reads them wrong while he’s working.
“But computers make up for my weaknesses, and on top of that, they can do things that humans have been unable to. When I watched an NHK computer course, I thought what amazing machines they were!”
Still, Suda was born and raised in a small town far from any city, so was unable to acquire all the knowledge and experience he desired, and the first time he got to touch a computer was when he entered university. He stayed in the Information Studies building where he could use the computers every day until closing time. In his fourth year, he joined the lab and set out on research to expand the potential of computers.
His current specialty is numerical computation. His research is on algorithms to make computer calculations faster. “I cook pretty much daily, but what I pour the most effort into is how to efficiently do several tasks at a time and finish faster,” he says with a laugh. “The same thing motivates my research on numerical computation. You see, thinking about how to order steps efficiently is a deep-seated source of fun for me.”
As computers evolve and become capable of more things, the speed of computation is projected to increase exponentially. For example, weather prediction simulations require an enormous number of calculations, but to increase the accuracy 10,000 times so as to be able to deal with local meteorological changes like torrential rainfall would require 100,000 times the amount of calculation time. Using Suda’s algorithm, however, it can be achieved in around 12,000 times the amount of time. “Creating a faster algorithm will probably not be possible. However, with this algorithm, the greater the load, the bigger the effect.”
Suda is currently working on an algorithm for greater speeds in the next generation of supercomputers that take over from world-class supercomputers like the K computer. What the world is focused on now is a method of increasing overall speed by reducing communication between multiple processors rather than making the computations themselves more efficient. In Japan, Suda’s team was the first to begin such work and remains at the forefront.
Supercomputers became a hot topic once at a government budget screening session where it was asked, “Do they have to be the fastest?” Suda, who contributed greatly to their acceleration, is himself a thoroughly laid-back person. “There isn’t just one correct answer for how to make things faster,” he says. “Ever since I was a child, I haven’t been any good at sports or fighting, and I’ve never had much of a competitive spirit. When a program I write doesn’t work, though, that bothers me no end. But using that frustration to do better next time is what makes someone a professional. Sometimes things are so frustrating that it’s hard to shift gears for a while, but feeling such strong frustration means you’ve found your calling.”
While giving off an air of calmness and complete lack of competitiveness, Suda delightfully takes on aggressive subjects like acceleration and efficiency. What expands the future of supercomputers may be laid-back pros rather than competition.
A pair of self-righting dolls (okiagari-koboshi) purchased when visiting Aizu for a conference. “I was drawn to how they stand back up no matter how many times they’re knocked down, and felt a sense of empathy. It felt a lot like my life, and it made me feel that I want to keep going despite the difficulties that may come.”
[TEXT: Shinri no taikai (“Great Ocean of Truth”)]
“When I was a child, I read a book that contained a quote from Newton saying, ‘I seem to have been only like a boy playing on the sea-shore and diverting myself in now and then finding a smoother pebble or a prettier shell than ordinary, whilst the great ocean of truth lay all undiscovered before me,’ and it made me want to scoop up even a drop of truth myself. That is my so-called ‘original intention.’”
After graduating from the Department of Information Science, Faculty of Science, the University of Tokyo in 1991, Suda acquired a Master’s degree and then became an assistant in the School of Science. In 1996, he earned his doctorate from the school. After working as a lecturer and assistant professor at Nagoya University, Suda became an associate professor with the Graduate School of Information Science and Technology, University of Tokyo in 2002. He has been in his current position since 2010. He researches algorithms to accelerate supercomputers that perform large-scale, high-accuracy simulations from both a theoretical and empirical standpoint. He enjoys listening to classical music and currently has a collection of about 2,800 classical music CDs. Recently he has been enamored of opera after hearing works by Verdi.
Interview date: February 9, 2018
Interview/text: Eri Eguchi. Photos: Takuma Imamura.