Note to Reporters: A photo of Michelle Mills Strout is available with the news release at http://www.news.colostate.edu/.
Michelle Mills Strout, assistant professor of computer science at Colorado State University, will spend the next five years developing models and tools that enable scientists to more easily develop faster, more precise computational models of the physical world thanks to a $750,000 grant from the U.S. Department of Energy.
Energy Secretary Steven Chu last week announced 69 scientists received the Early Career Research Program grants out of a pool of 1,750 applicants from universities and national laboratories from across the country. Strout is the only one from CSU.
Strout, who works in the high performance computing research group at the university, will receive $750,000 over five years as part of the $85 million in funding awarded. Her proposal, “Separating Algorithm and Implementation via Programming Model Injection,” was funded by the DOE Office of Advanced Scientific Computing Research.
“Researchers such as Dr. Strout are helping to set Colorado State apart as a home for scientists building solutions to real problems,” said Bill Farland, vice president for Research at Colorado State. “Her research could have significant implications for a variety of fields requiring complex modeling with intense computing.”
Strout is already working with fellow CSU faculty member Karolin Luger, a University Distinguished Professor in biochemistry and Howard Hughes Medical Institute Investigator, to more efficiently run molecular dynamic simulations, which can take months to simulate 25-50 nanoseconds of time. A nanosecond is one-billionth of a second.
Luger’s research group examines the structure of chromatin, the fundamental building block that helps DNA condense into chromosomes. Proteins that come into contact with this DNA are flexible so it is necessary to understand more about what parts are moving and range of motion of the various pieces, Strout said. Existing computers could spend months processing data to get the desired results– and the data must be chopped up into pieces to run parallel on different computers, Strout said.
Strout’s research will explore how to program and speed up scientific computations, especially when the structure of the physical entities being simulated is irregular. She will develop injectable programming models for specifying performance-critical algorithms such as dense and sparse matrix computations, task graph computations, stencil computations and lookup tables.
“If you really want to use computer simulation to help guide research, you need simulation development and the simulations themselves to have a decent turnaround and precision,” she said.
Strout plans to hire several graduate students and undergraduate hourly students to assist with the research. Eventually, she expects to create code generators and software libraries that can be shared with scientists around the globe.