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:227-228
DTSTART;TZID=America/Chicago:20211118T103000
DTEND;TZID=America/Chicago:20211118T110000
UID:submissions.supercomputing.org_SC21_sess181_pap296@linklings.com
SUMMARY:TensorKMC: Kinetic Monte Carlo Simulation of 50 Trillion Atoms Dri
 ven by Deep Learning on a New Generation of Sunway Supercomputer
DESCRIPTION:Paper\n\nTensorKMC: Kinetic Monte Carlo Simulation of 50 Trill
 ion Atoms Driven by Deep Learning on a New Generation of Sunway Supercompu
 ter\n\nShang, Chen, Gao, Lin, Wang...\n\nThe atomic dynamics Monte Carlo (
 AKMC) method has played an important role in the multi-scale physical simu
 lation; it is the bridge connecting the micro and macro worlds. Its accura
 cy is limited, however, by the empirical potentials. In this work, we prop
 ose a vacancy-centered tabulation algorithm to efficiently integrate ab in
 itio fitted neural network potentials (NNPs) with AKMC. We port this to SW
 26010-pro and optimize the neural network work potentials with big fusion 
 strategy for the new Sunway heterogenous computing units. We further optim
 ize the memory usage to make TensorKMC capable of simulating up to 50 tril
 lion atoms. TensorKMC can achieve excellent strong scaling performance usi
 ng up to 24,960,000 cores, and linear weak scaling using up to 27,456,400 
 cores.\n\nTag: Algorithms, Applications, Machine Learning and Artificial I
 ntelligence, Numerical Algorithms\n\nRegistration Category: Tech Program R
 eg Pass
END:VEVENT
END:VCALENDAR