Analytical model for evaluating performance multiprocessing for a set of parallel algorithmic structure

Abstract

The effectiveness of parallel processing depends on the type of algorithm parallelization and hardware architecture. It is important to estimate the projected effectiveness of parallelization of the problem for a particular hardware platform, took place at the earliest stage of development. For this purpose, are encouraged to develop an analytical model predictions of the future performance of hardware-software system. For each parallel algorithmic structures (parallel algorithm) of the submission of the main set is the conclusion the most important performance characteristics: acceleration factor, efficiency and scalability. The adequacy of the developed analytical model for predicting the efficacy of the treatment is confirmed in practice.