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:20260402T024534Z
LOCATION:Level 2 Lobby
DTSTART;TZID=America/Los_Angeles:20250623T180000
DTEND;TZID=America/Los_Angeles:20250623T190000
UID:dac_DAC 2025_sess287_LBR106@linklings.com
SUMMARY:Late Breaking Results: BLAST: Bisection-Free Learning Approach for
  Statistical Timing Characterization
DESCRIPTION:Kai Jing, Tao Bai, Zeyuan Deng, Junming Jiao, and Peng Cao (So
 utheast University)\n\nStatistical timing characterization for standard ce
 lls faces significant computational challenges due to the laborious bisect
 ion analysis for setup/hold constraint of sequential cells. To address thi
 s issue, we propose a Bisection-Free Learning Approach for Statistical Tim
 ing Characterization (BLAST) by extracting inherent delay for data path an
 d clock path in sequential cells as specific features. Multi-task learning
  is implemented with a multi-gate mixture-of-experts (MMoE) model to explo
 it the profound interdependency between setup and hold constraint for diff
 erent timing arcs, where the active learning strategy is incorporated to i
 mprove learning efficiency. Experimental results under 135 PVT corners wit
 h TSMC 12nm process demonstrate that the proposed BLAST achieves considera
 ble acceleration by avoiding the iterative bisection search in for statist
 ical constraint prediction with 76.9% runtime reduction compared to the co
 mmercial tool. Excellent prediction accuracy is achieved for various flip-
 flops by BLAST with the relative root mean square error (rRMSE) of 2.21% a
 nd worst-case absolute error (WCAE) of 0.82 ps.\n\n
END:VEVENT
END:VCALENDAR