|
Duncan Hall, Office 2068
6100 Main Street, MS 132
Houston, TX 77005
e-mail: mary f at rice dot edu
|
Under the advisement of Dr. Ken Kennedy and Dr. John Mellor-Crummey, I developed a compiler for Matlab D, a parallel Matlab language. Matlab D is Matlab plus arrays distributed across processors and a parallel for loop. The Matlab D compiler translates scripts in Matlab D to parallel Fortran calling GASNet communication routines for scalable parallel performance. The compiler supports user-defined data distributions to provide high performance even on scripts with unconventional data-access patterns.
Courses Taken at Rice
- Comp 412: Compiler Construction
- Comp 512: Advanced Compiler Construction
- Comp 515: Advanced Compilation for Vector and Parallel Processors
- Comp 422: Parallel Computing
- Comp 635: Heterogeneous Processors
- Comp 425: Computer Systems Architecture
- Comp 440: Artificial Intelligence
- Comp 481: Automata, Formal Languages, and Computability
- Comp 482: Algorithms
- Comp 311: Programming Languages
In Spring 2007, I led Comp 515, the seminar course on advanced compilation for high-performance systems. I was also the teaching assistant for Comp 412 for two semesters and Comp 421 for two semesters.
Publications
- M. Fletcher, C. McCosh, G. Jin, and K. Kennedy, "Compiling Parallel Matlab for General Distributions Using Telescoping Languages," Proceedings of ICASSP, April 2007. (pdf)
- Matlab D poster from Supercomputing '06
- J. E. Walter, M. E. Brooks (Fletcher), D. F. Little, and N. M. Amato, "Enveloping multi-pocket obstacles with hexagonal metamorphic robots," IEEE International Conference on Robotics and Automation, April 2004, New Orleans. (pdf)
- J. E. Walter, M. E. Brooks (Fletcher), and N. M. Amato, "Filling an obstacle pocket with hexagonal metamorphic robots," invited paper, 8th Conference on Intelligent Autonomous Systems, March 2004, Amsterdam.
- M. E. Brooks (Fletcher), "Algorithms for filling obstacle pockets with hexagonal metamorphic robots," Technical Report, TR03-002, Parasol Lab, Department of Computer Science, Texas A&M University, August 2003. (pdf)
Work Experience
I spent the past summer at Microsoft Research in Redmond, Washington, where I added nested data parallelism to the Accelerator project. Accelerator provides a library of data-parallel functions compatible with the .NET languages and compiles calls to these functions on the fly to code for the target platform (currently GPU's and multicore platforms).
In 2003, I participated in the CRA's Distributed Mentorship Project. The project gives female undergraduate students experience working on computer science research. I would strongly recommend the program to any sophomore or junior in college considering graduate school. It gives you first-hand experience working in a research group alongside graduate students.
My mentor was Jenny Walter from Vassar, and we worked at Texas A&M, where she obtained her PhD. I helped develop algorithms for filling obstacles with hexagonal metamorphic robots. This is the research page for the project.
I am also collaborating on the Habanero project, which addresses multicore challenges in programming languages, compilers, and runtime systems.