Overview of Research Interest

My research interests cover the areas of design methods and tools for embedded systems. In specific, they include performance evaluation tools, embedded processors, system architectures, programming models, domain-specific modeling languages, real-time operating systems, and retargetable compilers.

Currently I am actively pursuing research in these directions: high-level modeling of digital systems, functional test generation of microprocessors, system level design of embedded multiprocessor systems, and fast system simulation techniques using multi-processing.

Processor Modeling and Description

Traditionally in the field of architecture description language (ADL), researchers have been using either ad-hoc models or structural models to specify processors. These approach generally work for a limited architecture range and for a limited number of tools. My Ph.D. thesis work resulted in a flexible, formal, efficient and executable model for processor cores, which I called the Operation State Machine (OSM) model. Based on the OSM model, I designed the Mescal Architecture Description Language (MADL) for specifying processors. MADL has been used to generate tools including cycle-accurate simulators (CAS), instruction-set-simulators (ISS) and compiler components. The creation of OSM and MADL represents an significant advancement in the area. Compared to previous work, they are much more flexible in their support for different architectures and tools. Their details are best documented in my thesis. The MADL compiler, with a few examples including an ARM simulator, can be downloaded from this link.

Fast System Simulation

Instruction set simulators (ISSs) are frequently-used validation tools for embedded computer systems. The performance of an ISS is mainly driven by the simulation technique, and the speed of the host workstation. At a time when workstations are no longer improving as dramatically as in the 1990's, the simulation technique deserves more attention. My research aims to produce very fast and flexible simulation techniques and open-source simulators. The techniques that I have been exploring include both dynamic-compilation and using multi-processing. My group has implemented some simulators demonstrating the techniques. We released the SimIt-MIPS demonstrating these techniques at http://simit-mips.sourceforge.net.

High-level Modeling of Digital Systems

It is commonly understood that higher abstraction level favors productivity. The high-level synthesis field has made significant progress in the past two decades in converting sequential programs in high level languages to hardware. One remaining challenge are the modeling of communication among concurrent processes. My group used rendezvous, a classic synchronous communicatino primitive, to model the communication and synchronization among processes. We have made some very interesting discovery in the field. A full introduction of the work is located at the Lyra page.

To Prospective Graduate Students

I am seeking talented graduate students with EE, CS or math background. If you are interested in my research work and can demonstrate strong research potential, please feel free to contact me for openings. Please understand that I do not respond to generic emails.

HOME