Research On Heterogeneous Multi-core Hardware Acceleration Technology For Satellite Meteorological Information Processing

Reception and processing of satellite meteorological data is not only the core of the Satellite Meteorological Boardcasting System(SMBS),but also an important project in meteorological marine environment support,which plays a fundamental and supporting role.The active SMBS relies on large and high-power-consumption desktop PCs to process meteorological information.Besides,the low efficiency of meteorological data access and service prevents the SMBS from further evolution.In recent years,heterogeneous multicore technologies have made great progress in fields of application acceleration and energy efficiency improvement.On one hand,this thesis constructed a model for heterogeneous So Cs facing specific applications,and studied the hardware acceleration technology of satellite meteorological information processing based on the model.On the other hand,in order to improve the real-time performance and energy efficiency of both meteorological data reveiving and processing,this thesis designed and implemented a heterogeneous multicore acceleration prototype system for satellite meteorological information processing.First of all,this thesis analyzed requirements of the prototype system,and brought out both its functional and performance metrics.And then,overall design of the prototype system was given in this thesis.Besides,overall design of a rapid operation unit of meteorological data,a meteorological data storage system and a meteorological micro-server,which formed the prototype system,was given as well.Secondly,this thesis designed a quadruped model of heterogeneous multicore So Cs,modeled the implementation problem of specific application on a heterogeneous multicore So C as a task mapping problem on heterogeneous multicore So C,and then provided a solution for the mapping problem based on genetic algorithm.Furthermore,a rapid operation unit of meteorological data was designed based on the model.Thirdly,this thesis designed a meteorological data storage system based on file-merging and database,and a lightweight embedded meteorological micro-server based on an open-source server framework.Finally,the prototype system was implemented on the Xilinx ZCU106 hardware platform.The prototype system was tested and verified.Experimental results suggested that the prototype system met both the functional and performance metrics.