Advances in material science, biology, climate modeling and other areas hinge on efficient utilization of massive computing power, and that's part of the focus of a project headed by Jeff Vetter of Oak Ridge National Laboratory. To be successful, Vetter and colleagues must devise ways to overcome fundamental limitations of existing high-end computing systems while working to remove technological barriers to producing next-generation machines. They intend to do this by exploring three prototype core technologies: reconfigurable computing, processors in memory architectures and optical processors. Reconfigurable computing uses programmable hardware to literally plant a software algorithm into hardware. Processors in memory architectures attempt to overcome the "memory wall" by moving computation close to data in memory. Optical processors compute with light instead of silicon to improve parallelism. This project builds on ORNL's strong role in evaluating leadership-class computing machines, and researchers expect their effort to pave the way toward petascale computing for a multitude of scientific applications. This project is funded by ORNL's Laboratory Directed Research and Development Program.
Menu