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:20260202T201805Z
LOCATION:263-264
DTSTART;TZID=America/Chicago:20251120T111400
DTEND;TZID=America/Chicago:20251120T113600
UID:submissions.supercomputing.org_SC25_sess300_pap172@linklings.com
SUMMARY:MISA-AKMC: Achieve Kinetic Monte Carlo Simulation of 20 Quadrillio
 n Atoms on GPU Clusters
DESCRIPTION:Shunde Li (Computer Network Information Center, Chinese Academ
 y of Sciences; University of Chinese Academy of Sciences, Beijing); Zhijie
  Pan (Hangzhou Dianzi University); Ningming Nie (Computer Network Informat
 ion Center, Chinese Academy of Sciences); Jue Wang (Computer Network Infor
 mation Center, Chinese Academy of Sciences; University of Chinese Academy 
 of Sciences, Beijing); He Bai (Computer Network Information Center, Chines
 e Academy of Sciences); Genshen Chu (University of Science and Technology 
 Beijing; Engineering Research Center of Intelligent Supercomputing, Minist
 ry of Education); Yan Zeng (Hangzhou Dianzi University); Xinfu He (China I
 nstitute of Atomic Energy); Yangang Wang (Computer Network Information Cen
 ter, Chinese Academy of Sciences; University of Chinese Academy of Science
 s, Beijing); Changjun Hu (University of Science and Technology Beijing); a
 nd Xuebin Chi (Computer Network Information Center, Chinese Academy of Sci
 ences; University of Chinese Academy of Sciences, Beijing)\n\nThe Atomic K
 inetic Monte Carlo (AKMC) method provides insights into the macroscopic be
 havior of materials through atomistic-level simulations and finds broad ap
 plications in materials science innovation. Improving simulation scale and
  performance remains a consistent focus in the development of parallel AKM
 C software. We port the AKMC software to GPU clusters. To alleviate the me
 mory pressure in large-scale complex system simulations, we redesign the d
 ata layout and propose the lattice data compression and vacancy data decom
 pression algorithms. Additionally, we propose a multi-level pipeline schem
 e combined with an on-demand communication forwarding and merging strategy
  to reduce data transfer and communication overhead. Compared to state-of-
 the-art KMC software, MISA-AKMC\footnotemark achieves a 10.41-fold improve
 ment in computational throughput and a 52.07-fold expansion in simulation 
 scale. We implement the first true micrometer-scale AKMC simulation involv
 ing 20 quadrillion atoms on GPU clusters. MISA-AKMC achieves 96.03% parall
 el efficiency in weak scaling and 85.29% in strong scaling on 16,000 GPUs.
 \n\nTag: Applications\n\nRecording: Livestreamed, Recorded\n\nRegistration
  Category: Technical Program Reg Pass\n\nSession Chair: Hans-Joachim Bunga
 rtz (Technical University of Munich)\n\n
END:VEVENT
END:VCALENDAR