BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//pretalx//cfp.riscv-europe.org//eu-summit-2026//talk//CH97A7
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-CH97A7@cfp.riscv-europe.org
DTSTART;TZID=CET:20260609T154000
DTEND;TZID=CET:20260609T155000
DESCRIPTION:With the rapid evolution of RISC-V extensions such as Vector\, 
 Matrix\, and other custom instructions\, RISC-V platforms are becoming cap
 able of executing modern AI models. However\, achieving high-performance d
 eployment while maintaining a unified software stack across diverse extens
 ions remains a key challenge. **This paper introduces Buddy Compiler\, an 
 end-to-end AI compiler designed to provide a unified interface for AI mode
 l integration\, multi-level compilation optimizations\, and extensible cod
 e generation targeting diverse RISC-V extensions.** Buddy Compiler adopts 
 a multi-level architecture consisting of a frontend\, middle-section\, and
  backend\, enabling reusable high-level optimizations while supporting spe
 cialized backends for RISC-V architectures. The frontend provides a graph 
 infrastructure that interfaces with mainstream AI frameworks and converts 
 imported models into a unified representation expressed with high-level ML
 IR dialects. The middle-section is built on MLIR and performs multi-level 
 compilation optimizations\, including operator fusion\, memory access opti
 mization\, and vectorization. The backend implements dedicated MLIR dialec
 ts for RISC-V extensions\, such as RVV\, IME\, AME\, and Gemmini\, and per
 forms target-specific code generation. Through multi-level compilation\, B
 uddy Compiler and its runtime system enable efficient deployment of AI mod
 els on RISC-V platforms\, achieving performance comparable to manually opt
 imized implementations such as llama.cpp.
DTSTAMP:20260522T163507Z
LOCATION:Poster Island C
SUMMARY:End-to-End AI Compilation for RISC-V: A Multi-Level Optimization Ap
 proach - Hongbin Zhang
URL:https://cfp.riscv-europe.org/eu-summit-2026/talk/CH97A7/
END:VEVENT
END:VCALENDAR