Moderator: Franck Cappello (Argonne National Laboratory)
Panelists: Taisuke Boku (University of Tsukuba), Christian Plessl (Paderborn University), Kentaro Sano (RIKEN Center for Computational Science (R-CCS)), Kazutomo Yoshii (Argonne National Laboratory), Martin Herbordt (Boston University), Jeff Vetter (Oak Ridge National Laboratory)
Abstract: Reconfigurable computing has been recently introduced in clouds and extreme-scale data centers and shows outstanding performance for specialized applications such as search engine indexing and deep learning. The end of Moore's law, the convergence between HPC and big data, the recent introduction of AI and streaming as execution patterns in HPC, and the availability of devices with exceptional floating-point computing capabilities make reconfigurable computing more attractive than ever for HPC. The SC18 “Reconfigurable computing: will it make it this time?” panel exposed the main challenges to apply reconfigurable computing for HPC and concluded that more progress was needed to demonstrate its relevance and usefulness. In the last 12 months, several HPC platforms with FPGAs have been deployed for production and experimentation. This year's panel will present success stories about HPC applications that have been ported on these platforms and how they have been programmed, deployed, and optimized.
Back to the Panel Archive Listing