Current Research (More)
Programmers' productivity is of vital importance in computing today, which has wide-ranging economic and scientific consequences. In the domain of high-performance computing the productivity problem is an especially difficult one to solve. A move towards higher-level scientific languages, such as MATLAB, has the potential to ameliorate the situation as long as compiler techniques can ensure acceptable performance—something that has remained elusive so far. With the ubiquity of multi-core processors, the issue of productivity is no longer limited to the traditional high-performance computing applications. It is time for the lessons learned in high-performance and parallel computing to be applied to a large class of applications and problem domains. This is the focus of my current research.
My current research interests include:
- Compiler techniques to eliminate performance bottlenecks in high-level programming languages with focus on MATLAB and Ruby.
- Automatic parallelization techniques for high-level programming systems, including high-level languages and applications written using high-level component libraries (e.g., generic libraries).
- Program analysis and compiler techniques that pay special attention to memory bandwidth issues in the context of high-level programming systems, which are becoming most critical on the emerging “many-core” processors.
- Automatic program partitioning for heterogeneous parallel systems, such as those involving embedded processors, GPUs, and FPGAs.
Research Blogs
- Arun Chauhan (Thought Compilation)
- Chun-Yu Shei (Chun-Yu's Research Blog)
- Christine Task (Christine's Research Blog)
- Andrew Keep (Ruby colored glasses)