Using regression based methods for time-constrained scaling of parallel processor computing applications
MetadataShow full item record
There is a need for an automated method that facilitates time-constrained scaling for applications to encourage the wider use of parallel computation for computationally intense problems. In this thesis, we show for the first time a self-generated focal training method that is able to accurately achieve time-constrained scaling using a focused regression. This is demonstrated with six benchmark applications, but can be extended to any application of interest for which time-constrained scaling is needed.