Check out the new USENIX Web site.
Student GrantsUSENIX

Bradford Chamberlain, University of Washington (Seattle)

$ 33,995 - 9/99

Brad was born in Annapolis, son of a Naval Academy mathematics professor. He graduated in Computer Science from Stanford, and is now in his seventh and final year of UW's PhD program. Though that is a relatively long time, Chamberlain has used the time very wisely, working on multiple projects, including graph partitioning [1], 3D graphics [2] and the design of the ZPL parallel array language [3]. But, his greatest accomplishment has been team leader and the main intellectual force behind the implementation of the ZPL compiler and runtime system [4].

The Usenix Scholarship is to fund Chamberlain to complete the portability and performance studies of this idea, and write up the work in his thesis during the 2000 calendar year.


Extending ZPL to Support Parallel Sparse Arrays
ZPL is an array-based parallel language developed at the University of Washington during the past several years. ZPL has proven to be successful for dense parallel comutations, providing a clear and concise syntactic form, yet performing comparably with (or better than) handcoded C+MPI programs. ZPL is unique among parallel programming languages in that it provides programmers with a source-level means of evaluating the parallel implementation of their algorthms, allowing them to code in a way that minimizes overheads and maximizes parallelism. Our ZPL compiler has recently been extended to support hierarchical computations such as multigrid problems.

In this work, we will begin work on ZPL's successor language, A-ZPL (Advanced ZPL) by extending traditional ZPL to support sparse arrays. Our plan for doing so will preserve ZPL's clean, concise syntax and performance model, and will allow for the seamless interaction of sparse and dense arrays. By supporting these concepts in a manner that is completely orthogonal to existing ZPL concepts, we will be able to cleanly express crucial sparse hierarchical algorithms such as Adaptive Mesh Refinement (AMR) and the Fast Multipole Method (FMM). The ZPL compiler is already available for download at our website. Once this work is complete, the existing compiler will be augmented to support sparse arrays.

Please click here for a current update of this USENIX funded project.

UENIX Conference
We will submit to either the 2000 USENIX Annual Technical Conference, or the 4th Annual Linux Showcase and Conference, Atlanta.


?Need help? Use our Contacts page.
First posted: 22 Feb 1999 jr
Last changed: 03 Jan 2001 becca
Grants Index
Students Index