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:20260422T000711Z
LOCATION:708
DTSTART;TZID=America/Denver:20231112T161000
DTEND;TZID=America/Denver:20231112T163000
UID:submissions.supercomputing.org_SC23_sess438_ws_corr101@linklings.com
SUMMARY:Mixed-Precision S/DGEMM Using the TF32 and TF64 Frameworks on Low-
 Precision AI Tensor Cores
DESCRIPTION:Pedro Valero-Lara (Oak Ridge National Laboratory (ORNL)), Ian 
 Jorquera (Colorado State University), and Frank Lui and Jeffrey Vetter (Oa
 k Ridge National Laboratory (ORNL))\n\nUsing NVIDIA Tensor Cores has enabl
 ed the significant acceleration of general matrix multiplication for appli
 cations in AI and in high-performance computing. The use of such specializ
 ed accelerators can provide a performance increase between 8x and 20x, alb
 eit with a loss in precision. However, higher precisions are required in m
 any applications. Fortunately, mixed-precision methods can be employed to 
 maintain a high precision while also taking advantage of the performance o
 f lower-precision AI cores. We extend the state of the art by using NVIDIA
 ’s new TF32 framework, which not only burdens some constraints of the prev
 ious frameworks but also provides an equivalent precision and performance 
 by using a much simpler approach. We also propose a new framework called T
 F64 that attempts double-precision arithmetic with low-precision Tensor Co
 res. Although this framework does not exist yet, we validated the correctn
 ess of this idea and achieved an equivalent of 64-bit precision on 32-bit 
 hardware.\n\nTag: Applications, Software Engineering\n\nRegistration Categ
 ory: Workshop Reg Pass\n\nSession Chairs: Ignacio Laguna (Lawrence Livermo
 re National Laboratory (LLNL)); Cindy Rubio-González (University of Califo
 rnia, Davis); and Emmanuelle Saillard (French Institute for Research in Co
 mputer Science and Automation (INRIA))\n\n
END:VEVENT
END:VCALENDAR