| | GSRC Student Profile:
Research Overview: Graduate Student
My research interest is in providing compiler support for implicitly parallel programming. The underlying philosophy of implicitly parallel programming is to allow a developer to write a parallelizable computation separately from the mechanism by which the computation executes in parallel. By separating algorithms from architecture-specific execution strategies, implicitly parallel programming seeks to improve programmability, debugability, performance tuning, and portability of parallel codes.
I am working on tackling the so-called "memory wall" by providing information about memory use and memory dependences within the compiler through a combination of source-level annotations and compiler analyses. I have implemented an elimination-style array access region analysis within the IMPACT framework. Currently, I am developing a compiler intermediate representation to retain dependence and pointer aliasing information while supporting high-level, parallelizing code transformations.
| |
