|Name:||BoF 15: Reconfigurable Computing Architectures for HPC & HPDA|
|Time:||Tuesday, June 20, 2017
03:45 pm - 04:45 pm
|Breaks:||04:45 pm - 05:15 pm Coffee Break|
|Speaker:||Hans-Christian Hoppe, Intel|
|Marie-Christine Sawley, Intel|
|Abstract:||Use of reconfigurable compute units (mainly FPGAs) for HPC workloads has seen ups and downs, with the last peak of interest about a decade ago. Significant advances have been made since in the field of FPGAs and their system integration: new generations provide highly efficient FP units, fast cache coherent interconnects to CPUs were announced, and capacities in terms of functional units and high-speed I/O links has significantly increased. On the SW side, the momentum around OpenCL has been lowering the entry barriers for application developers. Both trends have conspired with the need to drive down energy consumption, and a new wave of interest has emerged in using reconfigurable compute units.
This BoF assembles a distinguished panel of speakers, who will give an up-to-date view of existing proof points for using reconfigurable computing in relevant application fields, the specific potential of current and next-generation reconfigurable platforms and the remaining limitations to a wider take-up. These short presentations will be followed by a panel-style discussion led and minuted by the BoF organizers.
Speakers will come from the academic user community (it is planned to have representatives from CERN and Barcelona Supercomputing Center presenting their FPGA-based application work), vendors of platforms and solutions (like Intel, Xilinx and Maxeler Technologies), and system software/programming model developers.
The presentations and the minutes from the discussion with the audience will be made publicly available.
Targeted AudienceThis BoF will be of prime interest to developers of HPC and HPDA applications and programming models/systems and to architects/designers of future compute systems. The first group will learn about specific success stories in using reconfigurable computing, about the performance and energy potential of such systems, and about the challenges applications face to harness their capabilities. The second group will gain insight into relevant use cases and their specific requirements, and on how these do leverage existing system features. Both should help them to create systems that provide best value to application developers.