A graph computing method and system based on dynamic code generation

Abstract

The present invention proposes a graph calculation method and system based on dynamic code generation, including: constructing an intermediate graph structure containing graph operation primitives according to a graph construction request, and storing the intermediate graph structure in an intermediate graph buffer after associating it with a graph name ;According to the graph algorithm request, generate a graph algorithm structure composed of external code bytecodes, and send it to the graph algorithm buffer; retrieve the intermediate graph buffer and graph algorithm buffer with the execution request, and obtain the intermediate graph structure to be executed, the graph to be executed A triplet composed of the algorithm structure and the parameter list, and retrieve the triplet in the local code cache to obtain the execution object in the local code cache, and execute to obtain the result. The invention injects the generated code in the local code space, eliminating the overhead of data exchange; constructs an intermediate graph structure that can be recompiled, so that the access code of graph data can be compiled and optimized; meanwhile, it increases the cache of the intermediate graph structure and the graph algorithm Caching avoids the preprocessing overhead of graph calculations.

Description

technical field [0001] The invention relates to the field of graph calculation and analysis in computer technology, and in particular to a graph calculation method and system based on dynamic code generation. Background technique [0002] Graph computing is an important technology for analyzing and mining associations in data. Due to the nonlinear characteristics of the graph structure, the development of graph algorithms is difficult, and at the same time, the performance is poor, and the code is not easy to optimize. In order to support efficient graph computing requirements, corresponding computing frameworks emerged as the times require. The graph computing framework encapsulates the graph into a unified data structure and provides basic operators to assist in the development of graph algorithms. [0003] The current graph computing framework is mainly divided into two categories: (1) Provide a fixed graph operator library and use an interpreted language to combine ope...

AI Technical Summary

Problems solved by technology

[0006] 2. The overhead of external code calls is introduced: when the graph algorithm obtains the required attributes and other data, the generated graph calculation code needs to be accessed through the external code interface, which not only introduces the overhead of external calls, but also makes the related external codes need to Exported symbols cannot be optimized inline;

[0007] 3. The graph structure cannot be reused, and the preprocessing cost of the algorithm is high: the graph algorithm needs to build the dependent graph structure data at runtime, and the same graph structure cannot be directly reused between different algorithms

At the same time, due to the introduction of complex code generation logic, the difficulty of developing graph algorithms has increased

Images