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:20230124T170803Z
LOCATION:C1-2-3
DTSTART;TZID=America/Chicago:20221115T083000
DTEND;TZID=America/Chicago:20221115T170000
UID:submissions.supercomputing.org_SC22_sess273_rpost184@linklings.com
SUMMARY:New Accelerated Mixed-Precision Parallel Algorithms for Hermitian
Eigenvalue Problem
DESCRIPTION:Posters, Research Posters\n\nNew Accelerated Mixed-Precision P
arallel Algorithms for Hermitian Eigenvalue Problem\n\nTsai, Luszczek, Don
garra\n\nThe multi-precision methods commonly follow approximate-iterate s
cheme by first obtaining the approximate solution from a low-precision fac
torization and solve. Then, they iteratively refine the solution to the d
esired accuracy that is often as high as what is possible with traditional
approaches. While targeting symmetric/Hermitian eigenvalue problems of t
he form Ax=(lambda)x, we revisited the SICE algorithm by applying the Sher
man-Morrison formula on the diagonally-shifted tridiagonal systems, we pro
pose an updated SICE-SM algorithm. We exploited asynchronous scheduling t
echniques to take advantage of the new computational graph enabled by the
use of mixed-precision in the eigensolver. By incorporating the latest two
-stage algorithms from the PLASMA and MAGMA software libraries for numeric
al linear algebra, we achieved up to 3.6x speedup using the mixed-precisio
n eigensolver with the blocked SICE-SM algorithm for iterative refinement
when compared with full double complex precision solvers for the cases wit
h a portion of eigenvalues and eigenvectors requested.\n\nRegistration Cat
egory: Tech Program Reg Pass, Exhibits Reg Pass
END:VEVENT
END:VCALENDAR