Concurrent graph data preprocessing method based on FPGA

Abstract

A concurrent graph data preprocessing method based on an FPGA comprises the steps that source graph data blocks and graph data characteristics are extracted from a graph data format information inputset and an algorithm information input set, a graph data and graph algorithm combination matrix is generated, and graph data format pre-conversion is conducted; according to the power law of the graphdata, a process allocation mode is determined through streaming drive scheduling, and then matched data partitioning and parallel design are performed according to attribute parameters of an FPGA accelerator in a heterogeneous platform; therefore, the implementation process of data preprocessing and scheduling optimization of the whole parallel graph processing process on the FPGA is realized. Bycombining the concurrent scheduling strategy of the GPU and FPGA acceleration platforms, the resource utilization rate and the overall performance are remarkably improved through overall preprocessing and scheduling optimization after the optimal data format is selected, and the high efficiency of the graph calculation process is guaranteed.

Description

technical field [0001] The invention relates to a technology in the field of embedded system data processing, in particular to an FPGA-based concurrent graph data preprocessing method. Background technique [0002] In the environment of large-scale graph computing, the processing of concurrent graph query and graph analysis often has the problem of high latency caused by the data structure not being suitable for the current algorithm. Existing solutions can consider solutions from the aspects of single graph computing efficiency, designing graph data structures that are easy to add, delete, modify, and query, and optimizing concurrent scheduling resource sharing, but ignore the performance difference of a graph algorithm when processing different graph data Large and the same kind of graph data will also encounter performance bottlenecks when processed by multiple graph algorithms. In practical applications, a single data structure is often not suitable for most of the concu...

AI Technical Summary

Problems solved by technology

Existing solutions can consider solutions from the aspects of single graph computing efficiency, designing graph data structures that are easy to add, delete, modify, and query, and optimizing concurrent scheduling resource sharing, but ignore the performance difference of a graph algorithm when processing different graph data Large and the same kind of graph data will also encounter performance bottlenecks when using multiple graph algorithms to process. In practical applications, a single data structure is often not suitable for most of the concurrent graph processing problems encountered.

Images