Dr. Rishi Khan

Dr. Khan is the Vice President of Research and Development at ET International, Inc (ETI). His team developed the system software for the Cyclops64 petascale many-core supercomputer (> 1 million cores) and have expertise in large scale irregularly load balanced parallel programs. Dr. Khan leads the research effort of a new class of dynamic adaptive runtime systems based on the “codelet execution model” first developed under DARPA UHPC. ETI’s runtime, SWARM (SWift Adaptive Runtime Machine), has shown promising scalability under a host of different application domains including data parallel, recursive search, tiled distributed linear algebra, adaptive mesh, and graph. In the latter category, ETI’s SWARM implementation of Graph500 has been in the top 10 of the Graph500 challenge. SWARM’s goals are to efficiently exploit parallelism, load balance work, hide latency through asynchrony, minimize data movement, support/exploit heterogeneity,  provide resilience, and minimize power. SWARM is designed to be applicable to the hardware environments of today’s supercomputers and enterprise data centers as well as tomorrows exascale systems and low power embedded systems targeting 50-75 GFLOPS/Watt.