Arbitrary diagonal linear equation set parallel solving method and system based on multiple file streams

Abstract

The invention provides an arbitrary diagonal linear equation set parallel solving method and system based on multiple file streams, and the method comprises the four steps: (1) dividing a banded equation into a plurality of initial sub-regions, and independently storing the data information of each initial sub-region in a corresponding file; (2) simultaneously and independently reading the data ofeach initial sub-region system equation by each main core, forming each assembled sub-region system equation by each slave core, and performing condensation polymerization to eliminate internal variables of each assembled sub-region system equation; (3) solving an interface equation in parallel by each main core; and (4) performing back substitution on each initial subarea according to a solvingresult of each main kernel interface equation to solve an internal variable; and if the iteration is finished, terminating, otherwise, starting from the second step again. On the basis of achieving distributed data storage, communication localization is achieved to the maximum extent through hierarchical communication, and the parallel computing efficiency of any diagonal linear equation set on alarge scale can be improved by fully utilizing the structural characteristics of a high-performance computing system.

Description

technical field [0001] The invention relates to a method for solving parallel diagonal equations, in particular to a method and system for parallel solving arbitrary diagonal linear equations based on multi-file streams. Background technique [0002] As a powerful tool for engineering numerical analysis, the finite element method is widely used in aerospace, energy exploration and development, transportation and other fields. It has received more and more attention in engineering design and analysis. effective approach to the problem. However, for large and complex systems, due to the huge calculation scale and complex solution, it is difficult to overcome the problems of large memory space and long solution time by traditional serial finite element analysis methods. The use of parallel computers to research and develop corresponding parallel algorithms provides a practical method for solving such problems. [0003] In terms of parallel algorithms, it is mainly divided int...

AI Technical Summary

Problems solved by technology

However, this method has received great limitations in practical applications. On the one hand, the bandwidth of the band equation of the practical problem is not fixed, so the method cannot be solved; on the other hand, because the method does not fully consider the current domestic high-performance computer system Structural characteristics, resulting in frequent communication between different sub-areas, and the resulting parallel computing efficiency is not ideal

Images