BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/New_York
X-LIC-LOCATION:America/New_York
BEGIN:DAYLIGHT
TZOFFSETFROM:-0500
TZOFFSETTO:-0400
TZNAME:EDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0400
TZOFFSETTO:-0500
TZNAME:EST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20210402T160552Z
LOCATION:Track 1
DTSTART;TZID=America/New_York:20201113T143000
DTEND;TZID=America/New_York:20201113T153000
UID:submissions.supercomputing.org_SC20_sess220_pec175@linklings.com
SUMMARY:ExaMPI – Keynote: Streaming Messages – A Distributed Memory Progra
 mming Model for Reconfigurable Hardware
DESCRIPTION:Workshop\n\nExaMPI – Keynote: Streaming Messages – A Distribut
 ed Memory Programming Model for Reconfigurable Hardware\n\nDe Matteis\n\nR
 econfigurable architectures, such as Field-Programmable Gate Arrays (FPGAs
 ), are stepping-in modern data centers and HPC clusters to address the nea
 ring end of technology scaling. Modern HPC FPGAs are usually shipped with 
 network connections to facilitate their large scale adoption in cloud serv
 ices and supercomputers. When FPGAs are deployed in distributed settings, 
 however, communication is typically handled either by going through the ho
 st machine, sacrificing performance, or by streaming across fixed device-t
 o-device connections, sacrificing flexibility. In this talk, we present St
 reaming Messages, a communication model for distributed memory programming
  in multi-FPGA systems that unifies explicit message passing with a hardwa
 re-oriented programming model. We introduce the Streaming Message Interfac
 e (SMI), a communication interface specification for high-level synthesis 
 programs specifically designed to capture and expose the semantics of Stre
 aming Messages. Finally, we show the results of a reference implementation
  of SMI that integrates with OpenCL on Intel FPGAs, demonstrating that, us
 ing SMI, programmers can implement distributed and scalable HPC programs o
 n reconfigurable hardware, without deviating from best practices for hardw
 are design.\n\nRegistration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR