Xinru Tang

ORCID: 0009-0004-6038-3709
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Fluid Dynamics and Mixing
  • Minerals Flotation and Separation Techniques
  • Innovative Microfluidic and Catalytic Techniques Innovation
  • Traffic control and management
  • Neural Networks and Reservoir Computing
  • Optimization and Search Problems
  • Advanced Manufacturing and Logistics Optimization
  • Modular Robots and Swarm Intelligence
  • Simulation Techniques and Applications
  • Heat Transfer and Boiling Studies
  • Advanced Memory and Neural Computing
  • Computer Graphics and Visualization Techniques
  • Real-Time Systems Scheduling
  • 3D Modeling in Geospatial Applications
  • Evacuation and Crowd Dynamics
  • Embedded Systems Design Techniques
  • Data Management and Algorithms
  • Cellular Automata and Applications
  • Scheduling and Optimization Algorithms

Tsinghua University
2004-2024

State Key Laboratory of Chemical Engineering
2004-2005

Spatial architecture is a high-performance that uses control flow graphs and data as the computational model producer/consumer models execution models. However, existing spatial architectures suffer from handling challenges. Upon categorizing their PE models, we find they lack autonomous, peer-to-peer, temporally loosely-coupled capability. This leads to limited performance in intensive programs. A architecture, Marionette, proposed, with an explicit-designed plane. The Control Flow Plane...

10.1145/3613424.3614246 preprint EN cc-by 2023-10-28

Meeting growing demands for low latency and cost efficiency in production-grade large language model (LLM) serving systems requires integrating advanced optimization techniques. However, dynamic unpredictable input-output lengths of LLM, compounded by these optimizations, exacerbate the issues workload variability, making it difficult to maintain high on AI accelerators, especially DSAs with tile-based programming models. To address this challenge, we introduce XY-Serve, a versatile, Ascend...

10.48550/arxiv.2412.18106 preprint EN arXiv (Cornell University) 2024-12-23
Coming Soon ...