BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-LIC-LOCATION:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260730T152640Z
LOCATION:
DTSTART;TZID=America/Los_Angeles:20260728T175200
DTEND;TZID=America/Los_Angeles:20260728T175300
UID:dac_DAC 2026_sess317_RESEARCH513@linklings.com
SUMMARY:AsymCheck: Asymmetric Partitioned Checkpointing for Efficient Larg
 e Language Model Training
DESCRIPTION:Zhangqiang Ming, Yuchong Hu, and Zhiyuan Luo (Huazhong Univers
 ity of Science and Technology); Patrick P. C. Lee (The Chinese University 
 of Hong Kong); Yuanhao Shu (Innovation Research Institute of Cethik Group 
 Co., Ltd.); and Xinjue Zheng, Wenxiang Zhou, and Dan Feng (Huazhong Univer
 sity of Science and Technology)\n\nDistributed large language model (LLM) 
 training faces prevalent failures and requires efficient checkpointing. St
 ate-of-the-art approaches employ partition-based pipelined checkpointing, 
 splitting checkpoints into partitions for concurrent processing. However, 
 existing solutions rely on a fixed partition size, which our analysis reve
 als is suboptimal for LLM training: large partitions cause bandwidth stall
 s during forward passes, while small partitions incur substantial startup 
 overhead during backward passes. We propose AsymCheck, which employs asymm
 etric partitioning: small partitions for forward passes and large partitio
 ns for backward passes, plus selective partition compression and batched f
 lushing optimizations. Evaluation on 64 GPUs shows AsymCheck reduces train
 ing time by 20.1%-48.2% over state-of-the-art methods, approaching no-chec
 kpointing efficiency.\n\nTrack: EDA7-II. Physical Design and Verification,
  Design\n\nSession Chairs: Christian Pilato (Politecnico di Milano); Mohse
 n Imani (University of California, Irvine); Flavio Ponzina (San Diego Stat
 e University); Amin Farshidi (Cadence Design Systems, Inc.); Yonggan Fu (G
 eorgia Institute of Technology, NVIDIA); Dimitrios Tsaras (Huawei, Hong Ko
 ng University of Science and Technology (HKUST)); Anh Nguyen (Fujitsu Limi
 ted); Ruiyang Qin (Villanova University); Jiliang Zhang (Hunan University)
 ; Meng Li (Peking University); and Rassul Bairamkulov (Advanced Micro Devi
 ces (AMD))\n\n
END:VEVENT
END:VCALENDAR