Cloud-Based FPGA Development, Deployment, and Use-Cases for Amazon F1 Instances (1:30pm to 3:30pm)

FPGA-accelerated computing has existed for nearly three decades, but has seen slow adoption relative to GPU computing. This has been due in part to the difficulty of programming FPGA-based applications, and due also to the challenge of introducing programmable hardware platforms into existing datacenter environments. Unlike GPUs, FPGAs have not been generally available to developers via a standardized platform such as a gaming PC or a graphics enabled laptop. The convergence of improved FPGA programming tools and a readily available cloud-based FPGAs provides researchers, application developers, and startups with a well-tested, standardized, and accessible FPGA platform. This session introduces the F1 FPGA cloud-based instances, walks through a typical development and deployment process, and highlights a number of use-cases in domains including genomics, video processing, text search, and financial computing.

Workshop Outline:

• Amazon F1 Instance Architecture and Use-cases
• F1 Hardware Development Kit
• F1 Programming Flow
• F1 Developer Experience:
• Maxeler Dataflow Programming for F1 Instances
• Accelize QuickPlay for FPGA Design and Deployment in the Cloud
• Monetizing FPGA Applications in the Cloud via AWS Marketplace

Workshop organizer: David Pellerin

David Pellerin is Business Development Principal for High Performance Computing at AWS, with a focus on high-scale applications in engineering, manufacturing, financial services, life sciences, media, and energy. Prior to joining AWS, Mr. Pellerin had a career in electronic design automation and FPGA-accelerated computing. He has experience with circuit simulation and synthesis, grid and cluster computing, and embedded systems. Mr. Pellerin is the author of five Prentice Hall technical books including Practical Design Using Programmable Logic and Practical FPGA Programming in C.

25th FCCM Retrospective Session -- How did we get here? ...and what does that tell us about where we're going next? (4:00pm to 6:30pm)

Pioneering work in this community 20--25 years ago set the stage for today's reconfigurable environment where Microsoft and Amazon have FPGAs in the cloud, Intel and Altera are one company, and FCCMs are programmed in C and OpenCL. We take a look back at the pioneering efforts and their prespective on the reconfigurable future. How did the last 20--25 years play out like they predicted, and how did it play out differently. What should we learn from this today to carry it forward in the years and decades to come?

Panelists:

• Mark Shand, Waymo LLC: "A 50K gate silicon foundry, with 50 ms turn-around time: A PAM retrospective."
• Maya Gokhale, Lawrence Livermore National Lab: "Parallel Programming Models for Custom Computing Arrays: Lessons from FCCM Roots"
• Stan Williams, HP Enterprise: "The lessons of Teramac revisited"
• John Wawrzynek, UC Berkeley: "Garp: A Reconfigurable Coprocessor---Still a good idea 20 years later?"

FPGA and Reconfigurable Computing Hall-of-Fame Induction

We will recognize the induction of two seminal papers from this community into the 2017 Class of the ACM/TCFPGA FPGA and Reconfigurable Computing Hall of Fame (http://hof.tcfpga.org):

• Programmable Active Memories: Reconfigurable Systems Come of Age. Jean E. Vuillemin, Patrice Bertin, Didier Roncin, Mark Shand, Herve Touati, and Philippe Boucard. IEEE Transactions on VLSI Systems, vol. 4, no. 1, pp. 56-69, 1996.
• A Defect-Tolerant Computer Architecture: Opportunities for Nanotechnology. James R. Heath, Philip J. Kuekes, Gregory S. Snider, and R. Stanley Williams. Science, vol. 280, no. 5370, pp. 1716-1721, 1998.

Workshop organizer: André DeHon