Ziyi Zhao (赵子懿) 
PhD Student @ Rice University | Computer Science | System & Memory
Hey there! I’m Ziyi (pronounced \Zee-E\), a 5th-year Ph.D. student at Rice University.
I focus on Memory Systems research, particularly heterogeneous systems that combine different memory types and processors like GPUs and CPUs.
I work with Prof. Scott Rixner and Prof. Alan Cox.
I am currently also a part-time intern at NVIDIA, where I work on GPU page migration algorithms to enhance GPU performance and enable larger models to run efficiently on GPUs with limited memory capacity.
I earned my Bachelor’s degree from Nankai University, where I had the fortune to do research on QEMU with Prof. Xiaoli Gong, Prof. Pen-Chung Yew, and Prof. Wenwen Wang.
- Working on GPU memory systems within the CUDA UVM team, focusing on OS- and runtime-level mechanisms for datacenter workloads. Architected and implemented a selective page migration system that enables transparent data movement between CPU and GPU memory with minimal hardware support, improving GPU execution efficiency without requiring application code changes.
- Delivered up to 17% performance gains on recommendation system workloads compared to hand-tuned memory management, and 66–100× speedups over Unified Virtual Memory (UVM), with additional 1–4× improvements on LLM and HPC workloads.
- Worked on GPU memory systems within the CUDA UVM team, focusing on performance characterization for datacenter-scale workloads. Developed a suite of microbenchmarks to analyze GPU memory access patterns and identify bottlenecks across the memory hierarchy in both hardware and software.
- Performed in-depth analysis of memory access and address translation behavior (e.g., TLB coverage and page table walks), and delivered insights that directly informed architectural decisions for next-generation GPUs.
(See Google Scolar for a complete publication list.)
- Oasis: Persistent Memory Oversubscription for Emerging Applications via Persistent MMAP
[code]
Ziyi Zhao, Scott Rixner
Proceedings of the 18th ACM International Systems and Storage Conference (SYSTOR '25). - Liberator: A data reuse framework for out-of-memory graph computing on GPUs
Shiyang Li, Ruiqi Tang, Jiawen Zhu, Ziyi Zhao, Xiaoli Gong, Wenwen Wang, Jiang Zhang, Pen-Chung Yew
IEEE Transactions on Parallel and Distributed Systems (TPDS) 34.6 (2023): 1954-1967. - Enhancing Atomic Instruction Emulation for Cross-ISA Dynamic Binary Translation
[code]
[slides]
Ziyi Zhao, Zhang Jiang, Xiaoli Gong, Ying Chen, Wenwen Wang, Pen-Chung Yew
2021 IEEE/ACM International Symposium on Code Generation and Optimization (CGO '21). - DQEMU: A Scalable Emulator with Retargetable DBT on Distributed Platforms
[code]
[slides]
Ziyi Zhao, Zhang Jiang, Ximing Liu, Xiaoli Gong, Wenwen Wang, Pen-Chung Yew
The 49th International Conference on Parallel Processing (ICPP '20).
Artifact Evaluation Committee (reviewers who validate the results of research paper)
- CGO '23 Artifact Evaluation Committee Member
- ASPLOS '23 Artifact Evaluation Committee Member
- MICRO '22 Artifact Evaluation Committee Member
Teaching Assistant
- Parallel Computing (COMP 422/534) - Fall 2022, Rice University
- Introduction to Computer Networks (COMP/ELEC 429/556) - Spring 2022, Rice University
- Operating Systems - Spring 2021, Nankai University