Here is the abstract:
My objective was to build on the work started by Hossain Deldari in his PhD thesis [9] about BSP, i.e. the Bulk Synchronous Parallel Model [12], on the following topics (all extracted from his
Future Worksection):
- Using the SAM (Split and Merge) skeleton to convert a raster image into a linear quadtree, i.e. a quadtree without pointers (based on Gargantini's algorithm [11]).
- Combining data parallelism and task parallelism (pipeline skeleton) [29] in order to compute a class of problems such as 1D-FFT and 2D-FFT [32], since the performance model for this kind of computation had not yet been addressed in research.
- Implementing a Set Membership Classification skeleton [30] in C++ and on BSP. This had been done on the WPRAM with the tree kept in global memory but on BSP the CSG tree needs to be distributed among the processes.
Deliverables (all available for download):
LaTeX or C/C++ source code, Makefiles, benchmarks, etc. are available upon request.
| Tell me (replace localhost by agt-the-walker.net) if you have comments / suggestions / etc. about this page, since feedback is always appreciated ! | |
| Last Update: 2006/01/18 23:31:29 |