The slowing pace of commodity microprocessor performance improvements combined with ever-increasing chip power demands has become of utmost concern to computational scientists. As a result, the high performance computing community is examining alternative approaches that address the limitations of conventional large-scale computing systems. This minisymposium examines the potential of semi-custom supercomputing platforms that utilize embedded-systems technology. We will explore how these specialized, application-centric systems may be suitable for a wide range of leading numerical methods. Furthermore, this approach to system architecture holds the promise of multi-petaflop computing at a fraction of the cost and power consumption of conventional HPC design.
Organizer:
John Shalf
Lawrence Berkeley National Laboratory / NERSC
Chris Rowen
Tensilica Inc.
Leonid Oliker
Lawrence Berkeley National Laboratory