BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20211207T055402Z
LOCATION:231-232
DTSTART;TZID=America/Chicago:20211115T103000
DTEND;TZID=America/Chicago:20211115T110000
UID:submissions.supercomputing.org_SC21_sess343_ws_h2rc108@linklings.com
SUMMARY:ACCL: FPGA-Accelerated Collectives over 100 Gbps TCP-IP
DESCRIPTION:Workshop\n\nACCL: FPGA-Accelerated Collectives over 100 Gbps T
 CP-IP\n\nHe, Parravicini, Petrica, O’Brien, Alonso...\n\nCollective operat
 ions such as scatter, gather, reduce, etc are utilized broadly to implemen
 t distributed HPC applications and are the target of extensive optimizatio
 n in all MPI implementations as well as dedicated collective libraries by 
 accelerator vendors (e.g. NCCL and RCCL by NVidia and AMD respectively). W
 e present ACCL, an open-source FPGA-accelerated collectives library design
 ed to serve applications running primarily in Xilinx FPGAs. Compared to pr
 evious collective communication solutions for FPGA, ACCL is flexible and e
 xtensible, easily portable, and fast. We evaluate ACCL up to 8 nodes and d
 emonstrate that ACCL outperforms OpenMPI over 100 Gbps TCP-IP for large me
 ssages.\n\nTag: Accelerator-based Architectures, Applications, Architectur
 es, Emerging Technologies, Heterogeneous Systems, Memory Systems, Networks
 \n\nRegistration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR