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:3000\, Level 3
DTSTART;TZID=America/Los_Angeles:20250625T153000
DTEND;TZID=America/Los_Angeles:20250625T154500
UID:dac_DAC 2025_sess112_RESEARCH1039@linklings.com
SUMMARY:Blaze: An Efficient Bit-Sparse Attention Architecture With Workloa
 d Orchestration Optimization
DESCRIPTION:Runzhou Zhang, Faxian Sun, Yiming Wang, Kunchen Zou, Zhinan Qi
 n, Jianli Chen, Jun Yu, and Kun Wang (Fudan University)\n\nThe attention m
 echanism is a key component in neural networks, essential for retrieving r
 elevant information in Natural Language Processing (NLP). However, the hig
 h computational complexity and substantial power consumption limits the de
 ployment of attention-based models. To overcome these issues, we introduce
  Blaze, an efficient attention architecture that utilizes both value and b
 it-level sparsity with workload orchestration optimization. Our Approximat
 e-Computing-Based (ACB) mechanism addresses workload imbalance in bit-spar
 se architectures, while the Leading-Booth mechanism further enhances the p
 erformance of attention computations. We also design a reconfigurable comp
 uting engine to support these innovations, improving performance in attent
 ion inference tasks.\n\nTopics: AI\n\nTracks: AI3: AI/ML Architecture Desi
 gn\n\nSession Chairs: Debjyoti Bhattacharjee (imec) and Minah Lee (Georgia
  Institute of Technology)\n\n
END:VEVENT
END:VCALENDAR