BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//cfp.riscv-europe.org//eu-summit-2026//talk//WHMPM8
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-WHMPM8@cfp.riscv-europe.org
DTSTART;TZID=CET:20260609T131000
DTEND;TZID=CET:20260609T132000
DESCRIPTION:Inference engines are specialized software systems designed to 
 execute pre-trained Machine Learning models. ONNX Runtime (ORT) emerges as
  a leading open-source inference engine for the Open Neural Network Exchan
 ge (ONNX) format\, allowing models to be deployed regardless of the framew
 ork in which they were trained.\n        While ORT provides a flexible arc
 hitecture for deploying models across diverse hardware\, it currently lack
 s architecture-specific optimizations for RISC-V. Consequently\, computati
 onally intensive tasks such as the convolution operation—which accounts 
 for the majority of inference time in Convolutional Neural Networks (CNNs)
 —suffer from hardware underutilization by relying on standard scalar ins
 tructions. In this paper\, we address this gap by proposing an optimized c
 onvolution implementation leveraging the RISC-V Vector Extension (RVV) and
  integrating it as a custom Execution Provider in ORT. We evaluate our sol
 ution on a Banana Pi BPI-F3 board across six standard reference CNN models
 . Experimental results show that our RVV-accelerated implementation achiev
 es speedups of up to 3x compared to the official scalar ORT release\, sign
 ificantly improving CNN inference performance on RISC-V platforms.
DTSTAMP:20260522T163129Z
LOCATION:Poster Island C
SUMMARY:ONNX Runtime Convolution Acceleration on RISC-V via RVV - Jose Sanc
 hez-Yun
URL:https://cfp.riscv-europe.org/eu-summit-2026/talk/WHMPM8/
END:VEVENT
END:VCALENDAR