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:20260522T150117Z
LOCATION:C141/143/149
DTSTART;TZID=America/Chicago:20181115T110000
DTEND;TZID=America/Chicago:20181115T113000
UID:submissions.supercomputing.org_SC18_sess207_pap256@linklings.com
SUMMARY:ParSy: Inspection and Transformation of Sparse Matrix Computations
  for Parallelism
DESCRIPTION:Kazem Cheshmi (University of Toronto), Shoaib Kamil (Adobe Res
 earch), Michelle Mills Strout (University of Arizona), and Maryam Mehri De
 hnavi (University of Toronto)\n\nIn this work, we describe ParSy, a framew
 ork that uses a novel inspection strategy along with a simple code transfo
 rmation to optimize parallel sparse algorithms for shared memory processor
 s. Unlike existing approaches that can suffer from load imbalance and exce
 ssive synchronization, ParSy uses a novel task coarsening strategy to crea
 te well-balanced tasks that can execute in parallel, while maintaining loc
 ality of memory accesses. Code using the ParSy inspector and transformatio
 n outperforms existing highly-optimized sparse matrix algorithms such as C
 holesky factorization on multi-core processors with speedups of 2.8× and 3
 .1× over the MKL Pardiso and PaStiX libraries respectively.\n\nTag: Linear
  Algebra, Memory, MPI, OpenMP, Programming Systems, Tools\n\nRegistration 
 Category: Tech Program Reg Pass\n\nSession Chair: Naoya Maruyama (NVIDIA C
 orporation)\n\n
END:VEVENT
END:VCALENDAR