The LogP and MLogP models for parallel image processing with multi-core microprocessor

Abstract

Despite the advancement and availability of the multiple core microprocessors, it remains an issue on how to fully utilize this relatively new computing platform to achieve optimal performance for a parallel algorithm. There are limitations to the existing theoretical model in analyzing parallel algorithms for multi-core microprocessor systems. The proposed Multi-core LogP (MLogP) model is a more realistic model for parallel computing with multi-core microprocessor. The MLogP model is a variant of the popular LogP model for parallel computation. Experiment with parallel image processing algorithms were used to determine the abilities of LogP and MLogP models in predicting the performance of parallel image processing algorithms on a Intel Core2 Quad 2.44 GHz microprocessor. © 2010 ACM.