|
| |
Welcome to my website
 My name is Albert R. Wynn III
I am a participate in the
Research Alliance in Math and Science (RAMS) Summer Internship
Program. Sponsored by the Mathematical, Information, and Computational Science Division and the Office of
Advanced Scientific Computing Research at the Department of Energy's Oak Ridge National laboratory
in Oak Ridge, Tennessee. The RAMS program is an initiative to allow minorities in Computing and Computational Sciences, Life Sciences, Engineering, and Mathematics the opportunity to experience the breadth of research available
at a national lab.
* Streaming Singular Value Decomposition (SVD)*
In order to maintain US dominance in maritime sensing, there is a continuing need to develop innovative approaches for near real–time remote detection of underwater threats and targets. However, the complexity of the calculations that one needs to perform increases dramatically with the size of sensor arrays being deployed, particularly in littoral environments. This, in turn, results in substantial growth of the underlying computational requirements. Such a growth cannot readily be met with standard hardware, and emerging, revolutionary computing technologies need to be considered. One of the most promising advances builds upon the recent availability of multi-core devices such as the Sony-Toshiba-IBM (STI) CELL processor. The goal of this project is to implement, first on the STI CELL simulator, then on the actual hardware, a fast(streaming) Singular Value Decomposition (SVD) algorithm. SVD–based methods are at the heart of underwater threat and target detection. Their throughput requirements including (1) high fidelity adaptive whitening of signal replicas, (2) data-driven generalized coordinate rotations and individual coordinate scaling are exacerbated because of the large computations. The performance of the streaming SVD method will be compared with that of the conventional SVD approach.
* Sony Toshiba IBM (STI) Cell Processor *
The Sony Toshiba IBM Cell Processor is leapfrog in the development of Computational Technology.
This processor is inherently capable of high parallelism that can nominally result in very high performance (over 200 GFLOPS per node).
|
