Construction And Optimization On Multi-core Parallel Computing Class For Image Algebraic

Multi-core architecture is the important milestone of the development of processor. The invention fulfils the needs of high performance, and is widely useful on such heavy computation situation as oil and gas exploration, weather forecast, virtual reality, artificial intelligence and so on, the new architecture will enhance efficiency significantly. Multi-core is getting popular and popular, and how to make full use of its advantage, keep pace with hardware and research the software which is accordance with multi-core tend to be a focus these days. It's considered that HPC on multi-core gains great value for the study.This subject lays the construction and optimization of parallel class for image algebraic, and work over multi-core platforms and parallel technology, seek for efficient, simplified and extensible blue print for parallel computing on multi-core. The paper begins with the analysis and comparison of multi-core architecture, common parallel algorithm and useful software tools. Secondly, the paper takes apart the primary functions of Mathematical morphology such as erosion, dilation, open, close and thinning, and searches the parallelism inside, and designs the multi-core parallel version according to analysis above. Thirdly, the class is tested by typical applications, and improved the performance by parallelization and compiler optimization based on vTune and Thread Profiler. Finally, the thesis gained objective performance evaluation in accordance with Amdahl law and Gustafson law.Nowadays image monitoring technology is used in various fields more and more popular, as well as image gathering turns to be high accuracy, great width and high frequency, those adds the urgent demand of the high performance image processing. This subject realizes the high performance image algebraic operation base on multi-core, enhances the efficiency of image processing which is combined with image algebraic operators, and gains high value of applications. The innovative points in this subject embodied in technological applications. One characteristic innovate is the image algebraic class which is transformed from the traditional serial version to parallel version on multi-core, the other is the performance analysis of parallel code by the newest technology, which realizes algorithm optimization, code level optimization and compiler level optimization, enhances the operation efficiency and the adaptation of hardware development. The technical route used in the subject has the universality, may promotes to the parallelization of other function libraries.As multi-core on IA is emerging, there's less study on multi-core architecture, and only a few relative resource and applications. This thesis carried on a preliminary exploration to construction, test, and optimization on multi-core parallel class for image algebraic, many research topics need to be expanded in the future.