BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//cfp.riscv-europe.org//eu-summit-2026//speaker//DCQFDS
BEGIN:VTIMEZONE
TZID:CET
BEGIN:STANDARD
DTSTART:20001029T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20000326T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:pretalx-eu-summit-2026-W3ANF8@cfp.riscv-europe.org
DTSTART;TZID=CET:20260611T130000
DTEND;TZID=CET:20260611T131000
DESCRIPTION:The RISC-V ecosystem is evolving toward AI-oriented computing\,
  with matrix-oriented proposal directions such as AME\, VME\, and IME attr
 acting increasing attention. In LLM inference\, matrix multiplication cons
 titutes one of the dominant computation patterns\, and quantized matrix mu
 ltiplication is widely adopted by many accelerators to improve efficiency.
  In this setting\, the practical value of matrix-oriented proposals depend
 s not only on the instruction capabilities they provide\, but also on how 
 effectively representative operators can be mapped onto realistic executio
 n flows. This work presents an operator-level profiling study of a current
 ly discussed AME proposal for RISC-V AI. We first design representative ma
 trix operators for quantized LLM-style workloads\, then develop a gem5-bas
 ed platform with support for the AME proposal\, and profile matrix multipl
 ication on this platform. Based on these observations\, we further analyze
  scaled matrix multiplication as an extended operator flow and discuss a p
 ossible scaled matrix multiplication instruction strategy as a future opti
 mization direction.
DTSTAMP:20260522T162354Z
LOCATION:Poster Island B
SUMMARY:Profiling and Optimizing AME for Matrix Multiplication - Xinlei Zha
 o
URL:https://cfp.riscv-europe.org/eu-summit-2026/talk/W3ANF8/
END:VEVENT
END:VCALENDAR