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:20260202T201248Z
LOCATION:Second Floor Atrium
DTSTART;TZID=America/Chicago:20251120T080000
DTEND;TZID=America/Chicago:20251120T170000
UID:submissions.supercomputing.org_SC25_sess533_post291@linklings.com
SUMMARY:Massively Parallel Bayesian Inference Framework for GPU Supercompu
 ters: Application to Estimation of Coseismic Fault Slip
DESCRIPTION:Kai Nakao, Tsuyoshi Ichimura, and Kohei Fujita (The University
  of Tokyo)\n\nWe present a massively parallel Bayesian inference framework
  for GPU supercomputers, demonstrated in coseismic fault slip estimation. 
 Bayesian inference, a robust method for inverse analysis, often relies on 
 Monte Carlo sampling with over 100,000 forward simulations, making large-s
 cale applications computationally intensive. A previous state-of-the-art i
 mplementation for the CPU-based supercomputer Fugaku was unsuitable for GP
 Us due to numerous small, imbalanced computations. We redesigned the algor
 ithm to enforce uniform, dense computation and employed Multi-Process Serv
 ice (MPS) to maximize GPU utilization. On a single node of the GPU-based s
 upercomputer Miyabi with an NVIDIA GH200 Grace Hopper Superchip, the metho
 d achieved 13.40 TFLOPS (20% of Tensor Cores FP64 peak) and scaled to 128 
 nodes with 92.3% efficiency. Compared with the original CPU implementation
  on Fugaku, it achieved a 42.1-fold speedup per node and reduced energy-to
 -solution to 18.8%. The methodology provides a general guide for porting B
 ayesian inference and similar applications to GPU-based environments.\n\nT
 ag: Research & ACM SRC Posters\n\nRegistration Category: Technical Program
  Reg Pass\n\nSession Chairs: Kento Sato (RIKEN Center for Computational Sc
 ience (R-CCS)); Chris Schlipalius (Pawsey Supercomputing Research Centre; 
 Commonwealth Scientific and Industrial Research Organisation (CSIRO), Aust
 ralia); and Anja Gerbes (Georg-August-Universität Göttingen)\n\n
END:VEVENT
END:VCALENDAR