Monday April 26 4:00 - 4:50 PM Kelley 1001
Joshua Fryman Intel Labs Computing in 2018 -- The Exascale Problem In March 2010, DARPA announced their Exascale program challenge, a project to present realized double precision ExaFLOPS in a total of 20 Megawatts by the year 2018. This represents over a 100x scaling factor shift in both computational capability as well as power efficiency assuming traditional CMOS scaling. With the march toward 5 nm fabrication, traditional scaling rules are inaccurate, increasing the burden on the Exascale effort. The future is one of power-limited devices at every level, requiring a shift in computer architecture. As a result of that shift, the programming and systems landscape requires a level of flexibility to accommodate the hardware designs while providing better abstractions to the software developer. At every level of computing hardware and software, resiliency and tolerance support will be critical. This talk will frame the key parameters behind the Exascale challenge, and briefly explore their implications across the computing landscape: fabricat ion, architecture, languages, and more. Biography Joshua Fryman received his BS from the University of Florida. Joshua spent many years in industry working on embedded system hardware and software before obtaining his PhD from Georgia Tech. Joshua then joined Intel Labs, focusing on corporate research and development efforts. Joshua has been involved with many projects at Intel, including the Terascale effort, the data parallel Ct/Firetown compiler, the Larrabee ISA and microarchitecture, and the Exascale effort. _______________________________________________ Colloquium mailing list [email protected] https://secure.engr.oregonstate.edu/mailman/listinfo/colloquium
