Research And Implementation Of Technology Of Parallel Vector Spatial Analysis Operation Optimization On Multi-core Processors

With the development of GPS ect. technology, data presents exponential growth trend, Large scale computational application promotes the development of parallel computer. But the parallel computer is mainly applied in the laboratory, its practical applications in engineering level is not high, especially in GIS, it is not highlight its multi-core processor advantage, therefore, making full use of multi-core processor of computer computation ability to accelerate GIS software execution speed, has become a key problem.In the future, GIS parallel will be the development trend of GIS, in parallel GIS, improve the practical efficiency, reduce computer application performance and machine peak performance gap. This paper basing on the data parallel program key technology and spatial analysis theory mainly researched the following work:spatial index, topological make polygon and buffer analysis, and realized the parallel algorithm of those algorithm in order to improve the actual performance of parallel applications. Based on the data parallel model and spatial analysis technology of parallel computer, can give full play to advantages of parallel computer, improved the performance of spatial analysis effectively, so,has the important academic and academic value and applied value. Specifically, the main content of this paper summarized as follows:1.research on the parallel of spatial index. This paper studies the spatial analysis of the most commonly used spatial index R-tree spatial index and its parallelization. On multi-core CPU computer, realized the R-tree parallel with OpenMP.R tree is a balanced tree similar to B tree, the most commonly used construction method is node insertion method. This paper puts forward a bulk insert algorithm for R tree, and its parallel, in order to make full use of computer hardware compared to the original single nuclear computer advantage. Based on the R tree index test key is inserted into the construction and performance comparison of parallel and serial R tree in the multi-processor environment, using the MAPGIS data format, the data quantity is small, insert and batch construction algorithm is very fast, but in the data than larger, batch has been established shows the superiority, in the experimental data, the bulk insert algorithm and traditional insertion algorithm efficiency has difference of10times in parallel and serial insert, insert, is the same with the increasing amount of data, parallel insertion has the advantage of more and more obvious, in the experimental data of differential distance9times. In the multi-core processor environment, building a parallel R tree has more advantages than the serial construction of R tree for a large amount of data, can improve the performance of spatial analysis.2.study the parallel of topological make polygon. This paper studies the spatial analysis of midline topology make polygon commonly used algorithm, then according to the winged edge structure characteristics of the computer graphics, promised the parallel topology make polygon and do the experiment to analysis the performance of serial and parallel topology make polygon algorithms.Vector data topology construction area is a very basic and important operation in spatial analysis, topological construction area is set which consists of one or more lines on a plane are treatment, process to form a closed polygon. This paper is based on the PMR four fork tree of parallel line topology construction area is studied, according to the environmental characteristics of multi-core processor, parallel algorithm for topology construction area, improve the multi-core processor sharing the performance of storage system. Experiments show that, the data quantity is small, parallel algorithm shows no advantage, but in a relatively large amount of data, parallel algorithm advantage will become more and more obvious; set the number of threads is not the more the better, the number of threads is higher than that of nuclear option number, because the communication between each thread, often will reduce the efficiency, so set the number of threads nuclear number about2times the most suitable.3.do the research on the buffer analysis parallel. This paper studies the vector data of unequal radius buffer analysis, and based on the data parallel theory, proposed the parallel unequal radius buffer analysis. Then this paper carries on the experiment and analysis the performance of the serial and parallel algorithm of unequal radius buffer analysis.Serial analysis of vector data with different radius of buffer, and Study on its multi-core processor environment parallel, to improve the speed of unequal radius buffer. Time is different because the amount of data and the division of the regional situation, if the amount of data is very small, the communication between segmentation and the information exchange used large time, if most of the region need to repeat the calculation, so the calculation time only will not be reduced, but also takes up more space. Conversely, if the amount of data is large, and the segmentation divided more appropriate, elements of repeated calculation is not much, so efficiency will be higher.