Research

Our mission is to advance the state-of-the-art of computing technology and improve the performance and efficiency of systems ranging from low-power edge devices to high-performance server systems. We pursue our goal through innovation and optimization across different levels, including computer architecture, runtime systems, compiler, and algorithm.



Our group was featured in POSTECH Labcumentary

Compiler and Runtime Technology

  • Optimizations for deep learning SW and HW
  • ML based resource management
  • ML driven analysis and transformation
  • IR for parallel and machine-learning workloads

  • Computer Architecture

  • Massively parallel architectures (e.g., GPU)
  • Near-data acceleration for deep learning and big data analytics
  • Heterogeneous memory system with storage-class memory
  • Interconnection networks
  • Secure cache coherence

  • Hardware-Software Co-design

  • Algorithm-hardware co-design for efficient deep learning
  • Systems for autonomous driving

  • DNN Model Acceleration

  • Neural network quantization & pruning
  • NPU/GPU-aware model design & Neural architecture search
  • DNN acceleration on heterogeneuous system
  • Video acceleration based on spatio-temporal correlation