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:20211207T054809Z
LOCATION:220-221
DTSTART;TZID=America/Chicago:20211118T110000
DTEND;TZID=America/Chicago:20211118T113000
UID:submissions.supercomputing.org_SC21_sess161_pap420@linklings.com
SUMMARY:STM-Multifrontal QR: Streaming Task Mapping Multifrontal QR Factor
ization Empowered by GCN
DESCRIPTION:Paper\n\nSTM-Multifrontal QR: Streaming Task Mapping Multifron
tal QR Factorization Empowered by GCN\n\nLin, Yang, Wang, Tsai, Li\n\nMult
ifrontal QR algorithm, which consists of symbolic analysis and numerical f
actorization, is a high-performance algorithm for orthogonal factorizing s
parse matrix. In this work, a graph convolutional network (GCN) for adapti
vely selecting the optimal reordering algorithm is proposed in symbolic an
alysis. Using our GCN adaptive classifier, the average numerical factoriza
tion time is reduced by 20.78% compared with the default approach, and the
additional memory overhead is approximately 4% higher than that of prior
work. Moreover, for numerical factorization, an optimized tasks stream par
allel processing strategy is proposed, and a more efficient computing task
mapping framework for NUMA architecture is adopted in this paper, which i
s called STM-Multifrontal QR factorization. Numerical experiments on the T
aiShan Server show average performance gains of 1.22x over the original Su
iteSparseQR. Nearly 80% of datasets have achieved better performance compa
red with the MKL sparse QR on Intel Xeon 6248.\n\nTag: Reproducibility Bad
ge, Algorithms, Architectures, Numerical Algorithms\n\nRegistration Catego
ry: Tech Program Reg Pass\n\nReproducibility Badges: Artifact Available, A
rtifact Functional, Results Reproduced
END:VEVENT
END:VCALENDAR