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:Exhibit Hall
DTSTART;TZID=America/Los_Angeles:20260728T175900
DTEND;TZID=America/Los_Angeles:20260728T175900
UID:dac_DAC 2026_sess306_WIP3309@linklings.com
SUMMARY:Terastal: Layer-Variant-based Scheduling for Real-Time Multi-DNN W
 orkloads on Heterogeneous Accelerators
DESCRIPTION:Sing-Yao Wu, Fengshuo Song, and Eli Bozorgzadeh (University of
  California, Irvine)\n\nHeterogeneous DNN accelerators map layers in real-
 time multi-DNN workloads to preferred accelerators to minimize latency. Ho
 wever, under skewed workloads, cross-accelerator layer latency difference 
 causes deadline misses. We introduce layer variants, customized layer impl
 ementations to reduce latency on non-preferred accelerators. We then prese
 nt Terastal, a real-time layer-variant design and scheduling framework for
  heterogeneous DNN accelerators. Terastal uses efficient network architect
 ures to design variants, performs virtual budget assignment to guide varia
 nt design and scheduling, and dynamically applies variants online to minim
 ize deadline violation and accuracy loss. Terastal reduces deadline miss r
 ate per model by 34.1% versus DREAM scheduler, with minimal accuracy loss.
 \n\nTrack: Student\n\n
END:VEVENT
END:VCALENDAR