A Parallel Partition Implementation Method for SPH Algorithm

Abstract

The invention discloses a parallel partition implementing method for an SPH (Smoothed Particle Hydrodynamics) algorithm. The method comprises the following steps: creating a simulation model for simulation analysis and performing initial setting; calculating the pressure value, stress tensor and artificial viscosity value of each particle corresponding to the simulation model; performing search analysis on the simulation model within each set time step, wherein the search analysis comprises determining each parallel partition and each parallel partition boundary corresponding to the simulation model; calculating a state equation and a control equation corresponding to the simulation model; and outputting a simulation result based on the set termination condition. The method reduces the storage space required for the simulation process, greatly reduces the traffic, then greatly improves the calculation efficiency, and effectively reduces the calculation cost.

Description

technical field [0001] The invention relates to the technical field of computer simulation, in particular to a method for implementing parallel partitions for SPH algorithms. Background technique [0002] Usually, the laws of motion and deformation of the objects we care about can be described by differential equations, which are called physical control equations. In fact, most moving objects are complex in shape, and the physical control equations must be numerically determined with the help of a computer. Calculation, also called numerical simulation or numerical simulation, means to use numerical calculation methods to simulate the movement and deformation of objects under certain conditions according to objective laws. able to come very close. Therefore, numerical simulation calculation has become the third research method next to experimental means and theoretical means in scientific research and engineering applications. It has the advantages of intuitive results, hig...

AI Technical Summary

Problems solved by technology

[0008] The existing SPH parallel methods include the parallel method of shared storage and the parallel method of distributed storage. Among them, shared storage requires all CPUs to share the same block of memory. The advantage of this method is that programming is very simple and does not require code changes. The disadvantage is The scalability is very poor, and the hardware requirements are extremely high; the parallel method of distributed storage is to use the network to connect different hosts, each host is independently calculated, and it performs the necessary communication through the network to simulate the same problem The disadvantage of distributed storage is that the program becomes very complicated, and the advantage is that the scalability is very good, so distributed parallel programs can run on computers with shared storage, but parallel programs with shared storage cannot be used on distributed computers , so the development of distributed parallel SPH programs has a better application prospect

The current distributed parallel SPH method uses a simplified parallelization method in the search process. The most commonly used method is to hand over all particle information to each computing node, but each computing node is responsible for searching for the part of particles that it is responsible for. Therefore, a large amount of redundant information will be given to each node, which may bring a large amount of storage burden. At the same time, at the end of each calculation, it is necessary to communicate all the particle information responsible for each computing node to each computing node, resulting in extremely low efficiency.

In summary, due to the technical limitations of the existing SPH parallel search method, the existing technology can deliver particles to each computing node separately, but cannot guarantee the load balance of each node and clearly determine the search relationship between nodes

Images