BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Events - ECPv6.15.18//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-WR-CALNAME:Events
X-ORIGINAL-URL:https://events.ucsc.edu
X-WR-CALDESC:Events for Events
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-Robots-Tag:noindex
X-PUBLISHED-TTL:PT1H
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:20250309T100000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:20251102T090000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:20260308T100000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:20261101T090000
END:STANDARD
BEGIN:DAYLIGHT
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
DTSTART:20270314T100000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:20271107T090000
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20260123T093000
DTEND;TZID=America/Los_Angeles:20260123T110000
DTSTAMP:20260404T170853
CREATED:20260120T223725Z
LAST-MODIFIED:20260120T223725Z
UID:7272-1769160600-1769166000@events.ucsc.edu
SUMMARY:Sharma\, R. (CSE) - Automatically Evolving GPU Libraries for Performance Portable AI Kernels
DESCRIPTION:GPUs are the workhorses of modern AI\, widely deployed and developed by many vendors including Apple\, Qualcomm\, Intel\, AMD\, and NVIDIA. While these GPUs all offer high compute potential\, programming them effectively is difficult because they differ in performance-critical features like SIMT width\, cache capacity\, and memory bandwidth\, demanding different optimization strategies. Tunable kernels address this by exposing parameters such as tiling dimensions and workgroup sizes\, enabling per-device specialization. Yet this produces static libraries: tuned once\, then frozen\, degrading as new hardware emerges. We propose automatically evolving libraries that expand their tuning knowledge as new hardware emerges\, with minimal impact on user experience. \nTo build such libraries\, we first need to understand the tuning landscape. We address this through GPU Goldmines\, a WebGPU-based framework for exhaustively collecting tuning data across diverse devices. Our tuned matrix multiplication kernels outperform an optimized baseline by 8.4x on average\, while matrix-vector kernels achieve 93% of platform bandwidth. We find that hyper-tuning for a single GPU causes 50% performance degradation on other devices\, whereas data-driven portability methods recover 88% of peak performance. These kernels are fundamental to the prefill and decode phases of LLM inference. We integrate them into llama.cpp as our evaluation platform\, where they outperform CPU and Vulkan backends. \nBuilding on this data\, we are developing Living Libraries to improve performance continuously without disrupting users. This means choosing good parameters upfront\, learning from real-world execution\, and knowing when to keep searching versus when to stop\, though hand-designed parameter spaces remain inherently bounded. To move beyond this\, we extend toward LLM-based kernel evolution\, where language models propose entirely new kernel variants\, opening a less structured but higher potential search space. \nEvent Host: Rithik Sharma\, Ph.D. Student\, Computer Science and Engineering \nAdvisor: Tyler Sorensen & Yuanchao Xu   \n  \nZoom: https://ucsc.zoom.us/j/92739836317?pwd=0ydDzimUFIoaLDUKst96dk27th4lvW.1 \nPasscode: 089560
URL:https://events.ucsc.edu/event/sharma-r-cse-automatically-evolving-gpu-libraries-for-performance-portable-ai-kernels/
LOCATION:Engineering 2\, Engineering 2 1156 High Street\, Santa Cruz\, CA\, 95064
CATEGORIES:Ph.D. Presentations
ATTACH;FMTTYPE=image/png:https://events.ucsc.edu/wp-content/uploads/2026/01/option-3-1.png
GEO:37.0009723;-122.0632371
X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Engineering 2 Engineering 2 1156 High Street Santa Cruz CA 95064;X-APPLE-RADIUS=500;X-TITLE=Engineering 2 1156 High Street:geo:-122.0632371,37.0009723
END:VEVENT
END:VCALENDAR