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:710
DTSTART;TZID=America/Denver:20231113T165000
DTEND;TZID=America/Denver:20231113T171000
UID:submissions.supercomputing.org_SC23_sess445_ws_scalah102@linklings.com
SUMMARY:Task-Based Polar Decomposition Using SLATE on Massively Parallel S
 ystems with Hardware Accelerators
DESCRIPTION:Dalal Sukkari and Mark Gates (University of Tennessee, Innovat
 ive Computing Laboratory); Mohammed Al Farhan (King Abdullah University of
  Science and Technology (KAUST)); and Hartwig Anzt and Jack Dongarra (Univ
 ersity of Tennessee, Innovative Computing Laboratory)\n\nWe investigate a 
 new task-based implementation of the polar decomposition on massively para
 llel systems augmented with multiple GPUs using SLATE. We implement the it
 erative QR Dynamically-Weighted Halley (QDWH) algorithm, whose building bl
 ocks mainly consist of compute-bound matrix operations, allowing for high 
 levels of parallelism to be exploited on various hardware architectures, s
 uch as NVIDIA, AMD, and Intel GPU-based systems. To achieve both performan
 ce and portability, we implement our QDWH-based polar decomposition in the
  SLATE library, which uses efficient techniques in dense linear algebra, s
 uch as 2D block cyclic data distribution and communication-avoiding algori
 thms, as well as modern parallel programming approaches, such as dynamic s
 cheduling and communication overlapping, and uses OpenMP tasks to track da
 ta dependencies.\n\nWe report numerical accuracy and performance results. 
 The benchmarking campaign reveals up to an 18-fold performance speedup of 
 the GPU accelerated implementation compared to the existing state-of-the-a
 rt implementation for the polar decomposition.\n\nTag: Algorithms, Heterog
 eneous Computing, Large Scale Systems\n\nRegistration Category: Workshop R
 eg Pass\n\nSession Chairs: Vassil Alexandrov (Hartree Centre, STFC); Jack 
 Dongarra (University of Tennessee, Knoxville; Oak Ridge National Laborator
 y (ORNL)); Christian Engelmann (Oak Ridge National Laboratory (ORNL)); Al 
 Geist (Oak Ridge National Laboratory (ORNL)); and Dieter A. Kranzlmueller 
 (Ludwig-Maxmilians-Universität München, Leibniz Supercomputing Centre (LRZ
 ))\n\n
END:VEVENT
END:VCALENDAR