A structured grid load balancing method based on LPT local optimization

Abstract

The invention discloses a structured grid load balancing method based on LPT local optimization, and aims to overcome the defects of an existing load balancing method and improve the load balancing rate and the calculation speed. The technical scheme is based on a genetic algorithm on the whole, and includes totally eleven steps of configuring parameters, initializing populations, calculating theadaptability, carrying out the local optimization on the LPTSize / 2 chromosome fragments with the maximum fragment value and the LPTSize / 2 chromosome fragments with the minimum fragment value in each chromosome by adopting an LPT method, carrying out the condition judgment, carrying out update judgment, performing population updating, selecting operators, crossing the operators and varying the operators, and outputting the best load balancing mode. According to the invention, a plurality of chromosome fragments with maximum and minimum chromosome fragment values in each chromosome are locally optimized; the fitness of the whole chromosome is improved, the population updating is carried out, the population is not prone to premature, a program is not prone to ending too early, a globally optimal solution can be obtained, and the parallel computing load balance rate of the whole structured grid is improved.

Description

technical field [0001] The invention relates to a load balancing method for improving structured grid parallel computing, especially a parallel load balancing method based on genetic algorithm and LPT (Largest Processing Time, maximum processing time) local optimization. Background technique [0002] Computation has been juxtaposed with theory and experiment as the three main research methods for human beings to understand the world, and it is mainly used to solve problems that are impossible to conduct experiments or that are too expensive to conduct experiments. In recent decades, with the in-depth understanding of physical laws and the needs of engineering applications, engineering computing has developed into a specialized discipline, which has been widely used in aerospace, automobiles, environmental engineering, materials, physics and ships, etc. . The engineering calculation process is mainly to iteratively calculate the characteristic quantities on the grid, and the...

AI Technical Summary

Problems solved by technology

[0006] (1) Too fine a block will lead to a large increase in additional communication overhead and reduce the efficiency of parallel computing

[0007] (2) Too many sub-regions (blocks) will increase the number of iterative solutions and affect the calculation efficiency, which is determined by the characteristics of the region decomposition algorithm itself

[0008] (3) Engineering calculation is not only a science, but also an art. It is not only related to the numerical calculation method, but also related to the grid. Inappropriate block division may lead to calculation divergence

[0081] (1) The deterministic method is fast, but the load balancing effect is uncertain, sometimes the effect is particularly good, but most of the effect is not good

[0082] (2) The intelligent optimization algorithm is relatively primitive, and the problems of poor robustness and low computational efficiency of the corresponding intelligent optimization algorithm are not considered

[0083] (3) The effect of the hybrid algorithm is better, but there are still problems of low computational efficiency and unobvious effect of load balancing optimization

The population generated by the deterministic method often leads to the failure of the intelligent optimization algorithm to obtain a good solution.

Images