Lab Introduction

Research Area

Algorithm/HW/SW Co-design for Intelligent SoCs (Systems-on-Chips)

  • Efficient Deep Learning Architectures

  • Hardware-Friendly Deep Neural Networks

  • Architectures and Design Tools for In-Memory Computing Accelerators

  • Architectures and Compilers for Reconfigurable Computing

  • Electronic Design Automation (EDA) and System-Level Design

Short Description

At Intelligent Computing & Codesign Lab (ICCL) in UNIST, we are mainly concerned with the problems of optmizing system performance and power at the intersection of software and hardware. At the core of information technology is computation, which is realized through the combination of hardware and software. But today both hardware and software are seeing fundamental changes, forecasting a paradigm shift in how we understand and realize computation. This shift is caused first by the changes and advances in hardware technology, in particular, semiconductor device technology, forcing us to seek new boundaries between hardware and software, such as application-specific processors, hardware accelerators, and even reconfigurable processors. The second cause of the shift is the emergence and pervasive use of deep learning systems. Deep learning hardware is not programmed in the conventional way such as sequential or parallel programming, but what is called “training”, or iterative application of relevant data through the learning pathway of the system. Thus deep learning systems provide a real possibility of new computer architectures that are not based on von Neumann abstraction but more like the human brain. At ICCL, as we are doing research involving both hardware and software components of a system, we are in a unique position of being able to perform original research that can pave the way for the paradigm shift in the core definition of computing.

Korean Version

우리 연구실은 기본적으로 소프트웨어와 하드웨어의 경계에서 시스템 성능이나 전력 등을 최적화할 수 있는 다양한 기법들을 연구하고 있습니다. 특히나 오늘날의 컴퓨터는 두가지 중요한 패러다임의 변화를 준비하고 있습니다. 하나는 꾸준히 하드웨어 기술이 발전함에 따라, 하드웨어와 소프트웨어의 경계를 허물거나 기존의 폰노이만 컴퓨터 패러다임을 변경하는 연구 수요와 관심이 매우 높아지고 있다는 것이고, 다른 측면에서는, 최근에 등장한 딥러닝을 중심으로 기존의 순차적 또는 병렬 프로그래밍 방식이 아닌 데이터가 곧 프로그램이 되는 새로운 프로그래밍 패러다임을 제시하는 차세대 컴퓨팅 시스템에 대한 연구가 요구된다는 것입니다. 소프트웨어와 하드웨어 분야의 중심에 있음으로써 우리 연구실은, 컴퓨터가 직면한 두가지 패러다임의 변화를 리드하는 경쟁력있는 연구가 가능합니다.