Natural Sciences
Extreme Computing
Twitter
Facebook
Forward
The human brain can absorb piles of numbers. But it can’t easily convert them into an image of, say, wind speed patterns in a hurricane.
It takes powerful computers running complex software to do those conversions. But that presents an ongoing problem: The computer programs that perform these transformations need to be constantly updated to keep up with ever-evolving technical standards.
At the UO, “keeping up with the pace of change” could be the mantra of a special division in the computer and information science department. It’s a place where undergraduates work alongside graduate students in what is essentially a research-and-development lab.
Undergraduates help design and test programming and other tools that will support the new age of computing—high-performance computing and the visualization of big data. This is the Computing and Data Understanding at Extreme Scale research group, or CDUX, the brainchild of associate professor Hank Childs.
“Our undergraduate researchers play essential roles in the program,” Childs said. “It’s a great thing for them to have on their résumés as they look for jobs.”
Rinse and Repeat
Although today’s computers are faster and smarter than ever before, their speed has essentially peaked. To push through that barrier and get the most out of computers’ “brains,” programs need to perform multiple tasks simultaneously rather than sequentially.
Elliott Ewing
Elliott Ewing and Kirsten Dawes—two of four undergraduates on the 15-person CDUX team—learned to rewrite software programming so that a popular tool in the CDUX lab can run efficiently in this new, multitasking environment.
Ewing and Dawes essentially looked under the hood of VisIt, a free visualization and analysis tool created in 2000 by Childs and a team of programmers at the California-based Lawrence Livermore National Laboratory. VisIt processes information-heavy simulation data and produces comprehensible, visual results. It’s been downloaded hundreds of thousands of times, and is used on many of the world’s top computers.
The undergraduates worked with the sections of coding necessary to make the tool compatible with computers running parallel tasks. The work involved scouring VisIt coding for what Ewing called “trouble spots,” then testing new coding in those sections and also on other types of computers. Then doing it over and over again, until all the bugs were gone.
“You rinse and repeat,” Ewing said, “until the process comes out spotless.”
Ewing and Dawes ultimately made changes to the VisIt infrastructure that are now in use by scientists around the globe. It’s an understatement to say the two are proud of their contribution.
“A lot of undergraduates don’t get to work in an industry type of setting,” Ewing said. “There are research opportunities and industry opportunities, and this is a cool mixture of both of them.”
Inside a Hurricane
For undergraduates, one of the payoffs of CDUX is the access to graduate students.
Garrett Morrison
Garrett Morrison teamed up with master’s student Vincent Chen to write programming that enables a visualization tool to produce ultrarealistic- looking graphics to work with VisIt.
The duo is working on a collection of software called GraviT, which illustrates big-data challenges such as the movement and speed of hurricane winds. But new coding was needed for GraviT and VisIt to work together.
Morrison and Chen were the code doctors for this painstaking project, working side by side as they identified and excised faulty sections of programming and replaced it—line by line—with new coding. They met regularly to brainstorm and make sure they were heading in the same direction with solutions.
At the heart of their project was the scientific process of trial and error.
“You try something. It doesn’t work. You try something else. It works a little bit,” Morrison said. “It’s an iterative process.”
The team had to ensure that their programming changes would one day enable GraviT to handle the biggest data sets imaginable—information that captures what transpires when a star explodes, or illustrates the stresses on metals as they’re bent.
Although GraviT is still a work in progress, Morrison and Chen—working with researchers from a handful of other institutions across the country—have written roughly 500 lines of code that gave VisIt its new grip on GraviT’s graphics.
They’ve used GraviT to render what happens inside a tornado and to create images of a hurricane that show the wind speed, precipitation and air temperature throughout the storm.
“When you finally produce something that’s working,” Morrison said, “it’s a moment to be proud of.”
—Jim Murez