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:20260422T000712Z
LOCATION:401-402
DTSTART;TZID=America/Denver:20231116T110000
DTEND;TZID=America/Denver:20231116T113000
UID:submissions.supercomputing.org_SC23_sess158_pap613@linklings.com
SUMMARY:High-Performance SVD Partial Spectrum Computation
DESCRIPTION:David Keyes and Hatem Ltaief (King Abdullah University of Scie
 nce and Technology (KAUST)); Yuji Nakatsukasa (Mathematical Institute Univ
 ersity of Oxford); and Dalal Sukkari (University of Tennessee, Innovative 
 Computing Laboratory (ICL))\n\nWe introduce a new singular value decomposi
 tion (SVD) solver based on the QR-based Dynamically Weighted Halley (QDWH)
  algorithm for computing the partial spectrum SVD (QDWHpartial-SVD) proble
 ms. By optimizing the rational function underlying the algorithms only in 
 the desired part of the spectrum, QDWHpartial-SVD algorithm efficiently co
 mputes a fraction (say 1-20%) of the most significant singular values/vect
 ors. We develop a high-performance implementation of QDWHpartial-SVD on di
 stributed-memory manycore systems and demonstrate their numerical robustne
 ss. We perform a benchmarking campaign against their counterparts from the
  state-of-the-art numerical libraries across various matrix sizes using up
  to 36K MPI processes. Experimental results show performance speedups for 
 QDWHpartial-SVD up to 6X and 2X against PDGESVD from ScaLAPACK and KSVD, r
 espectively. We also report energy consumption for these algorithms and de
 monstrate how QDWHpartial-SVD can further outperform PDGESVD in that regar
 d by performing fewer memory-bound operations.\n\nTag: Algorithms, Linear 
 Algebra, Post-Moore Computing\n\nRegistration Category: Tech Program Reg P
 ass\n\nReproducibility Badges: Artifact Available, Artifact Functional, Re
 sults Reproduced\n\nSession Chair: Julien Langou (University of Colorado D
 enver; University of Colorado, Department of Mathematical and Statistical 
 Sciences)\n\n
END:VEVENT
END:VCALENDAR