Advances in technology, and computing in particular, occur on an almost daily basis. Today’s cutting-edge technology could become obsolete in just a few months. One trend in the computing industry is toward an increase in the number of cores on a microchip.
The rationale driving chip manufacturers to develop multicore processors is that these processors provide better performance per watt than single-core processors. Multicores systems require significant sharing of resources such as memory and networks, which can constrain performance and impact power consumption. Not much is known about the power-performance tradeoffs in multicore systems.
Texas A&M University’s Dr. Valerie Taylor is developing infrastructure that will help engineers and application developers better understand how such multicore systems work. She recently received $2.4 million from the National Science Foundation’s Computer Systems Research program to develop Multicore Application Modeling Infrastructure or MuMI (pronounced "mummy").
"Multicore processors are the foundation of next-generation computing systems. The computing industry is only going to increase the number of cores on a chip," said Taylor, who is head of the Department of Computer Science and Engineering.
MuMI will facilitate systematic measurement, modeling, and prediction of performance, power consumption, and power-performance tradeoffs in multicore systems. MuMI will also be used to model, analyze, and optimize the power consumption and performance of key benchmarks and applications of multicore systems such as weather and climate modeling applications, biomolecular simulations and energy simulations.
Collaborating with Taylor on this project are Dr. Shirley Moore from the University of Tennessee in Knoxville and Dr. Kirk Cameron from Virginia Tech.
TEES is the lead site for MuMI, with a funding of $865,000 and Taylor as the sole primary investigator on the TEES grant. TEES is the engineering research agency of the State of Texas and a member of The Texas A&M University System.