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:20260202T201257Z
LOCATION:Second Floor Atrium
DTSTART;TZID=America/Chicago:20251121T080000
DTEND;TZID=America/Chicago:20251121T120000
UID:submissions.supercomputing.org_SC25_sess620_post175@linklings.com
SUMMARY:High-Performance Sparse Attention on Tensor Cores: Fused3S and Bey
 ond
DESCRIPTION:Zitong Li (University of California, Irvine)\n\nSparse attenti
 on is a core building block in many leading neural network models, from gr
 aph-structured learning to sparse sequence modeling. It can be decomposed 
 into a sequence of three sparse matrix operations (3S): sampled dense-dens
 e matrix multiplication (SDDMM), softmax normalization, and sparse matrix 
 multiplication (SpMM). Efficiently executing the 3S computational pattern 
 on modern GPUs remains challenging due to (a) the mismatch between unstruc
 tured sparsity and tensor cores optimized for dense operations, and (b) th
 e high cost of data movement. \n\nPrevious works have optimized these spar
 se operations individually or addressed one of these challenges. This post
 er introduces Fused3S, the first fused 3S algorithm that jointly maximizes
  tensor core utilization and minimizes data movement. Across real-world gr
 aph datasets, Fused3S achieves significant speedup over state-of-the-art k
 ernels on H100 and A30 GPUs.\n\nTag: Research & ACM SRC Posters\n\nRegistr
 ation Category: Technical Program Reg Pass\n\nSession Chairs: Kento Sato (
 RIKEN Center for Computational Science (R-CCS)); Anja Gerbes (Georg-August
 -Universität Göttingen); and Chris Schlipalius (Pawsey Supercomputing Rese
 arch Centre; Commonwealth Scientific and Industrial Research Organisation 
 (CSIRO), Australia)\n\n
END:VEVENT
END:VCALENDAR