Xinyu Yang

Xinyu Yang

Auto-generated profile, based on publicly available information

Xinyu Yang

Hong Kong University of Science and Technology, Hong Kong, Hong Kong

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
27 August 2025
Verified
Conference Paper
Open access
XWXinchen Wan
Xinchen Wan
XYXinyu Yang
Xinyu Yang
KXKaiqiang Xu
Kaiqiang Xu
XLXudong Liao
Xudong Liao
YJYilun Jin
Yilun Jin
KCKai Chen
Kai Chen

16 October 2023
Verified
Conference Paper
Open access
MKMyeongseob Ko
Myeongseob Ko
XYXinyu Yang
Xinyu Yang
ZJZhengjie Ji
Zhengjie Ji
HJHoang Anh Just
Hoang Anh Just
PGPeng Gao
Peng Gao
RJRuoxi Jia
Ruoxi Jia

11 August 2026
Verified
Conference Paper
Open access
XYXinyu Yang
Xinyu Yang
KXKaiqiang Xu
Kaiqiang Xu
KCKai Chen
Kai Chen
GPU memory offloading is widely adopted for LLM workloads but shifts the bottleneck to GPU-CPU transfers, which can take up to 90% of the end-to-end inference/training time! Paradoxically, over 60% of PCIe bandwidth remains idle. The root cause is that PCIe links are individually bottlenecked but collectively underutilized. Bursty, phase-driven transfer patterns leave bandwidth idle both within and across jobs. We present TurboBus, which pools PCIe bandwidth across co-located jobs via emerging scale-up fabrics. TurboBus enables any GPU to borrow idle PCIe links from neighboring GPUs, even those belonging to other jobs, while preserving isolation through a privileged daemon. At the core of TurboBus, it streams data through relay GPUs with bounded memory overhead, keeps all links busy through fine-grained PCIe allocation, enables bidirectional transfers leveraging PCIe/NVLink bandwidth asymmetry, and balances fairness and completion time with a size-aware scheduling. We fully implement TurboBus and our experiments show that it reduces first-token latency by up to 40% for on-demand model loading (within 5% of the analytical optimum), achieves up to 1.6x throughput for KV-cache-offloaded inference, and accelerates training by up to 7%, while imposing less than 1% overhead on co-located workloads.