Current Projects

  1. Performance Analysis and Optimization for Unified Memory Systems
  2. High Performance Parallel Machine Learning Algorithms
  3. Cross-Layer Power-Bounded High Performance Computing on Emerging and Future Heterogeneous Computer Clusters
    • Coordinator - An Application-Aware Power Coordination Framework for NUMA Multicore Systems. Source code available at github.
    • CLIP - Cluster-Level Intelligent Power Coordination for Power-Bounded Systems.
  4. High Performance Computing on Heterogeneous Systems
    • Software-Based Dynamic GPGPU Co-run and Scheduling Framework.
    • Understanding and Modeling the Performance of Unified Virtual Memory.
  5. Resilient and Energy Efficient Scalable Computing on Heterogeneous Architectures

Past Projects

  1. Marcher - A Heterogeneous High Performance Computing Infrastructure for Research and Education in Green Computing, 2013/8-2018/5
  2. Develop, Improve, and Maintain the Parallel Fine-Grained Soft Error Fault Injector, 2016/8-2018/5.