BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Denver
X-LIC-LOCATION:America/Denver
BEGIN:DAYLIGHT
TZOFFSETFROM:-0700
TZOFFSETTO:-0600
TZNAME:MDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0600
TZOFFSETTO:-0700
TZNAME:MST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20200129T163557Z
LOCATION:505
DTSTART;TZID=America/Denver:20191118T153000
DTEND;TZID=America/Denver:20191118T155500
UID:submissions.supercomputing.org_SC19_sess128_ws_ia114@linklings.com
SUMMARY:Stretching Jacobi: A Two-Stage Pivoting Approach for Block-Based F
 actorization
DESCRIPTION:Workshop\n\nStretching Jacobi: A Two-Stage Pivoting Approach f
 or Block-Based Factorization\n\nThuerck\n\nGiven enough parallel compute u
 nits, the processing time for sparse (in-)complete factorization is determ
 ined by the number of level sets found in the pattern of the matrix. For p
 ositive definite or diagonal dominant matrices, the iterative Jacobi-metho
 d presents itself as a scalable alternative on massively-parallel systems.
  In this work, we present modifications to the Jacobi algorithm, extending
  its applicability to a wider class of matrices.  Our approach cuts the ma
 trix into blocks to allow register-based pivoting in batched CUDA factoriz
 ation kernels and, for the first time on GPUs, also flexible permutations 
 on the block level in an a posteriori threshold full pivoting scheme.  Exp
 eriments show our batched kernels to be on par with partially pivoted appr
 oaches and the resulting sparse factorizations compare favorable in qualit
 y and speed with their traditionally scheduled counterparts.\n\nTag: Works
 hop Reg Pass, Algorithms, Architectures, Irregular Applications\n\nRegistr
 ation Category: Workshop Reg Pass, Algorithms, Architectures, Irregular Ap
 plications
URL:https://sc19.supercomputing.org/presentation/?id=ws_ia114&sess=sess128
END:VEVENT
END:VCALENDAR