Planet-Scale Extended Reality System

In today's emerging computer technology domains, research on a single stack is no longer adequate to support our fourth wave of technology transformation: we are moving towards a meta-verse era where human perception, hardware design, low-latency low-power constraints and photorealistic context-aware immersion all co-exist in a tiny IoT/edge device, which can communicate with you through natural language processing models, entertain you via sophisticated prediction of your desire, while customising a universe just for your own personal statistics and imagination. Now we not only have to care about the independent component's efficiency such as algorithm, compiler, runtime, operating system, hardware design, etc, but also have to put them together in a cohesive fashion so they efficiently perform together and pose a positive impact on user quality of experience (QoE). For the last several years, we and our collaborators have published some of the first papers (HPCA'17, HPCA'18, HPCA'19, ISCA'19, ASPLOS'21) on future VR system characterisations and system-level design & optimisations (including both multi-accelerator based HMD SoC and cloud server designs) in the field of computer architecture. Ultimately, FSA and our international collaborators aim to prototype the first planet-scale virtual reality metaverse system that is accessible by users who may not have sophisticated mobile hardware and high-speed wireless network.


  • Post0-VR: Enabling Universal Realistic Rendering for Modern VR via Exploiting Architectural Similarity and Data Sharing, HPCA'23
  • Q-VR: System-Level Design for Future Mobile Collaborative Virtual Reality Rendering, ASPLOS'21
  • OO-VR: NUMA Friendly Object-Oriented VR Rendering Framework For Future NUMA-Based Multi-GPU Systems, ISCA'19
  • PIM-VR: Erasing Motion Anomalies In Highly-Interactive Virtual Reality World With Customised Memory Cube, HPCA'19
  • Perception-Oriented 3D Rendering Approximation for Modern Graphics Processors, HPCA'18
  • Processing-in-Memory Enabled Graphics Processors for 3D Rendering, HPCA'17