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:20230124T171519Z
LOCATION:C144-145
DTSTART;TZID=America/Chicago:20221118T095000
DTEND;TZID=America/Chicago:20221118T100000
UID:submissions.supercomputing.org_SC22_sess445_ws_ia107@linklings.com
SUMMARY:Blocking Sparse Matrices to Leverage Dense-Specific Multiplication
DESCRIPTION:Workshop\n\nBlocking Sparse Matrices to Leverage Dense-Specifi
 c Multiplication\n\nSylos Labini, Vella, Bernaschi, Silvestri, Nutt\n\nRes
 earch to accelerate matrix multiplication, pushed by the growing computati
 onal demands of deep learning, has sprouted many efficient architectural s
 olutions, such as NVIDIA’s Tensor Cores. These accelerators are designed t
 o process efficiently a high volume of small dense matrix products in para
 llel. However, it is not obvious how to leverage these accelerators for sp
 arse matrix multiplication. A natural way to adapt the accelerators to thi
 s problem is to divide the matrix into small blocks, and then multiply onl
 y the nonzero blocks. In this paper, we investigate ways to reorder the ro
 ws of a sparse matrix to reduce the number of nonzero blocks and cluster t
 he nonzero elements into a few dense blocks. While this pre-processing can
  be computationally expensive, we show that the high speed-up provided by 
 the accelerators can easily repay the cost, especially when several multip
 lications follow one reordering.\n\nSession Format: Recorded\n\nTag: Accel
 erator-based Architectures, Algorithms, Architectures, Big Data, Data Anal
 ytics, Parallel Programming Languages and Models, Productivity Tools\n\nRe
 gistration Category: Workshop Reg Pass
END:VEVENT
END:VCALENDAR