Before working at DARPA, Andreas Olofsson founded Adapteva and created the Epiphany many-core processor. A number of variants of this chip, with 16, 64 and 1024 cores have been designed which are all very energy efficient, cheap and powerful which makes them very interesting for use in HPC and embedded systems alike. Adapteva also produced the Parallella, a single board computer for less than $99, which combines an Epiphany chip with standard CPU. Andreas is therefore at the cutting edge of chip design technology and a strong advocate of the fact that you don’t necessarily need the money and resources of extremely large chip companies to make innovative hardware.
Keynote: Building a universal hardware compiler
As the complexity of chips has rapidly increased in line with Moore’s law predictions, we have seen an explosion in the cost and time require to design advanced electronic systems. DARPA is addressing these challenges through two new research programs: IDEA (Intelligent Design of Electronic Assets) and POSH (Posh Open Source Hardware). Together, the two programs form the foundation of an intelligent hardware compiler. The aim of the research effort is to create a hardware compiler capable of automatic generation of production ready manufacturing data directly from source code and schematics. In essence, the program is trying to create the hardware equivalent of a software compiler. Achieving this ambitious goal will require advancing the state of the art in machine learning, optimization algorithms, and expert systems. This session will review the POSH and IDEA research programs and provide analysis of its potential impact on the current HPC ecosystem.