BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CECS - ECPv6.18.4//NONSGML v1.0//EN
CALSCALE:GREGORIAN
METHOD:PUBLISH
X-ORIGINAL-URL:https://www.cecs.uci.edu
X-WR-CALDESC:Events for CECS
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:20240310T100000
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
DTSTART:20241103T090000
END:STANDARD
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
END:VTIMEZONE
BEGIN:VEVENT
DTSTART;TZID=America/Los_Angeles:20251015T140000
DTEND;TZID=America/Los_Angeles:20251015T150000
DTSTAMP:20261027T171519Z
CREATED:20251007T230120Z
LAST-MODIFIED:20261027T171519Z
UID:16512-1760536800-1760540400@www.cecs.uci.edu
SUMMARY:Unleashing Multi-GPU Computing to the Next-Level
DESCRIPTION:Title: Unleashing Multi-GPU Computing to the Next-Level \nSpeaker:  Dr. Xulong Tang \nDate and Time: Wednesday\, October 15\, 2:00pm \nLocation: EH 2430 \nAbstract: \nModern computing platforms increasingly rely on multi-GPU systems to deliver the performance\, bandwidth\, and memory capacity required by data-intensive applications spanning cloud and edge services to large language models (LLMs) and quantum workloads. However\, scaling GPUs beyond single-device encounters fundamental system-level challenges\, including address translation overheads\, complex memory management\, and inefficient multi-tenancy in shared environments. In this talk\, I will present our recent efforts on hardware-software co-design to overcome these bottlenecks and unleash the full potential of multi-GPU computing. I will discuss novel techniques such as lightweight invalidations for page translation\, short-circuiting page table walks\, and multi-instance GPU (MIG) reconfiguration strategies that collectively deliver substantial performance and efficiency gains. Looking forward\, I will highlight emerging research directions in scalable heterogeneous GPU architectures and memory management in GPU virtualization. An especially exciting avenue is applying these insights to LLMs\, where efficient GPU memory use and multi-GPU coordination are critical for KV-cache reuse\, reduced serving latency\, and scalable throughput. Together\, these innovations pave the way for the next generation of GPU systems that can power both large-scale scientific computing and next-generation intelligent applications. \nBiography: \nDr. Xulong Tang is an Associate Professor in the Department of Computer Science at the University of Pittsburgh. His research spans computer architecture and systems\, with a focus on multi-GPU architectures\, scalable memory management\, and heterogeneous computing across high-performance\, edge\, and quantum systems. His group has pioneered advances in GPU memory management\, temporal graph neural network training\, and quantum compilation. Dr. Tang has published extensively in premier venues such as MICRO\, HPCA\, ISCA\, ASPLOS\, PLDI\, ICLR\, and NeurIPS\, and his work has received recognition including an Intel Research Award\, an ICLR Spotlight Award\, and a MICRO Best Paper Nomination. He is also active in the community\, serving as General Co-Chair of ASPLOS 2026 and Vice Chair of the IEEE Micro Top Picks 2025 Selection Committee. \nHosted By: Prof. Sitao Huang \nMore Information
URL:https://www.cecs.uci.edu/event/unleashing-multi-gpu-computing-to-the-next-level/
LOCATION:EH 2430\, Engineering Hall\, University of California\, Irvine\, Irvine\, CA\, 92697\, United States
CATEGORIES:Seminars at CECS
ATTACH;FMTTYPE=image/png:https://cecs.uci.edu/files/2025/10/Xulong_10_15_25-e1793121315619.png
END:VEVENT
END:VCALENDAR