面向新型处理器的数据密集型计算∗

摘要

近年来，随着数据量的不断增大，数据密集型计算任务变得日益繁重。如何能够快速、高效地实现在大规模数据集上的计算，已成为数据密集型计算的主要研究方向。最近几年，研究人员利用新型的硬件处理器对数据密集型计算进行加速处理，并针对不同新型处理器的特点，设计了不同形式的加速处理算法。主要对新型硬件处理器基于数据密集型计算的研究进行了综述。首先概述了新型硬件处理器的特点；然后，分别对新型处理器 FPGA 和 GPU 等硬件进行性能分析，并分析了每种处理器对数据密集型计算的效果；最后提出了进一步的研究方向。%In recent years, with increased data volume, data-intensive computing tasks become increasingly critical. How to efficiently and effectively implement data-intensive computing on large data sets becomes a major research direction for data-intensive computing. Currently, researchers attempt to use new processors to accelerate the data-intensive computing process. Different acceleration approaches could be adopted according to the characteristics of new processors. In this paper, the new processors, well as the algorithms, for data-intensive computing research are surveyed. First, the new features of processors are reviewed. Then the capability of each new processors and their performance over data intensive computing are analyzed. Finally, the future research directions are discussed.