BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20211208T164410Z
LOCATION:3014
DTSTART;TZID=America/Los_Angeles:20211208T111000
DTEND;TZID=America/Los_Angeles:20211208T113000
UID:dac_DAC 2021_sess156_RESEARCH459@linklings.com
SUMMARY:MOVED TO VIRTUAL: Scaling up HBM Efficiency of Top-K SpMV for Appr
 oximate Embedding Similarity on FPGAs
DESCRIPTION:Research Manuscript\n\nMOVED TO VIRTUAL: Scaling up HBM Effici
 ency of Top-K SpMV for Approximate Embedding Similarity on FPGAs\n\nParrav
 icini\n\nTop-K SpMV is a key component of similarity-search on sparse embe
 ddings. This sparse workload does not perform well on general-purpose NUMA
  systems that employ traditional caching strategies.\nInstead, modern FPGA
  accelerator cards have a few tricks up their sleeve. We introduce a Top-K
  SpMV FPGA design that leverages reduced precision and a novel packet-wise
  CSR matrix compression, enabling custom data layouts and delivering bandw
 idth efficiency often unreachable even in architectures with higher peak b
 andwidth.\nWith HBM-based boards, we are 100x faster than a multi-threaded
  CPU implementation and 2x faster than a GPU with 20\% higher bandwidth, w
 ith 14.2x higher power-efficiency.\n\nTopic: Design\n\nKeyword: Approximat
 e Computing for AI/ML\n\nVirtual Program: Presented In-Person
END:VEVENT
END:VCALENDAR