SCS Undergraduate Thesis Topics
|Matthew Trentacoste||Doug James||Generalized Matrix Computation on Graphics Hardware|
The increased versatility of graphics hardware has allowed for general computation to be mapped onto the GPU. The main considerations that must be addressed before widespread adoption of the GPU as another computation processor are the availability of high precision formats and APIs that abstract away the details of the implementation. Recent graphics hardware has support for floating point data throughout the pipeline, so we turn to the second concern of providing an interface to the hardware that hides the specifics of implementing the functionality on graphics hardware, but still maintains performance. We then use this interface to implement non-negative matrix factorization when used for performing feature extraction to demonstrate the strengths of the library when run on current graphics hardware.