多态阵列处理器的并行计算研究

发布时间：2018-06-19 05:25

本文选题：多态阵列处理器 + 众核　；参考：《西安邮电大学》2014年硕士论文

【摘要】：通过在单个芯片上集成成千上万个简单处理核来获得高性能和高吞吐量,已成为目前计算机体系结构研究的热点。虽然众核芯片上的单个简单处理核不具备复杂控制逻辑,在开发指令级并行方面存在很多难点,但是同一芯片上的众多简单处理核协同工作能获得复杂多核所不能获得的指令级并行度。本文通过对流行的计算机体系结构进行了详细分析,针对其在并行计算开发方面的难点,深入研究了众核多态阵列处理器--PAAG(Polymorphic Array Architecture for Graphics and image processing)在并行计算方面的特性,通过结合线程级并行、数据级并行以及指令级并行等多种并行计算模式来挖掘程序中的并行性。本文以具有广泛应用的图像处理为主要研究对象,通过对常用的图像处理算法进行深入研究和分类,挖掘其中的并行处理共性,并以PAAG为硬件开发平台,充分挖掘图像处理领域中大规模并行计算的能力,从时空满载的目标角度来使计算任务最大并行化,以此充分利用众核上的硬件资源,使图像处理程序在多态阵列处理器中获得尽可能高的加速比。本文最后推导了PAAG阵列处理器在图像处理算法中的加速比计算公式,通过实验数据说明了众核PAAG阵列处理器在并行计算方面所具有的优势,并总结出在众核PAAG阵列处理器中进行深度并行计算的方法。
[Abstract]:The integration of thousands of simple processing cores on a single chip to achieve high performance and high throughput has become a hot topic in the research of computer architecture. Although a single processing core on a multi-core chip does not have complex control logic, there are many difficulties in developing instruction level parallelism. However, many simple processing cores on the same chip can work together to obtain instruction level parallelism that complex multi-core can not achieve. In this paper, the popular computer architecture is analyzed in detail, and the characteristics of polymorphic Array Architecture for Graphics and image processing) with multi-core polymorphic array processor in parallel computing are deeply studied in view of its difficulties in parallel computing development. The parallelism in programs is mined by combining thread level parallelism, data level parallelism and instruction level parallelism. This paper takes the widely used image processing as the main research object, through the in-depth research and classification of the commonly used image processing algorithms, mining the common parallel processing, and taking PAAG as the hardware development platform. Fully mining the ability of large-scale parallel computing in the field of image processing to maximize parallelization of computing tasks from the point of view of space-time full load, so as to make full use of the hardware resources on multiple cores. Enables image processors to achieve the highest speedup possible in polymorphic array processors. Finally, the speedup calculation formula of PAAG array processor in image processing algorithm is deduced. The advantages of multi-core paag array processor in parallel computing are illustrated by experimental data. The method of deep parallel computing in multi-core PAAG array processor is summarized.
【学位授予单位】：西安邮电大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP332

【参考文献】