BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//cfp.riscv-europe.org//eu-summit-2026//talk//TX9SGW
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-TX9SGW@cfp.riscv-europe.org
DTSTART;TZID=CET:20260611T170000
DTEND;TZID=CET:20260611T171500
DESCRIPTION:Llama.cpp is a widely used open-source platform for running Lar
 ge Language Models (LLMs) on CPUs\, but its support for RISC-V remains lim
 ited compared to x86 and ARM. Many floating-point and quantized kernels la
 ck RISC-V Vector (RVV) implementations\, restricting the performance of ex
 isting hardware. This work improves the upstream RISC-V performance by vec
 torizing core floating-point kernels and extending support across multiple
  quantization types\, enabling first-class support for RVV in Llama.cpp. V
 LEN-aware data repacking is introduced to accelerate GEMM and GEMV kernels
  for both floating point and quantization types. The optimized kernels are
  validated across VLENs up to 1024-bit\, with benchmarking on Banana Pi BP
 I-F3 (256-bit VLEN) demonstrating considerable performance gains over upst
 ream Llama.cpp. This work is supported by the RISC-V Software Ecosystem (R
 ISE)\, with the vectorized kernels being upstreamed to Llama.cpp along wit
 h the test infrastructure.
DTSTAMP:20260522T163251Z
LOCATION:Plenary
SUMMARY:Optimizing Llama.cpp and GGML for RISC-V Vector (RVV) - Taimur Ahma
 d\, Adeel Ahmad
URL:https://cfp.riscv-europe.org/eu-summit-2026/talk/TX9SGW/
END:VEVENT
END:VCALENDAR