Heterogeneous node computing system and method

Abstract

The invention discloses a heterogeneous node computing system and method, the system comprises a CPU module and a GPU module, and the GPU module comprises a PME computing module and a data copying module; the PME calculation module is used for generating an electrostatic potential energy matrix by adopting a long-range force PME algorithm and sending a first data copying instruction to the data copying module; the data copying module is used for obtaining an electrostatic potential energy matrix according to the first copying instruction and sending the electrostatic potential energy matrix and a calculation instruction to the CPU module; the CPU module is used for calculating grid data according to the electrostatic potential energy matrix and the calculation instruction and sending a second data copying instruction to the data copying module; the data copying module is also used for transmitting the grid data to the PME calculation module; and the PME calculation module is also usedfor calculating the stress of each atom in the molecular dynamics PME simulation by adopting a spline difference algorithm according to the grid data. By means of the technical scheme, the calculationefficiency in the molecular dynamics simulation process is improved, and the utilization rate of hardware resources is increased.

Description

technical field [0001] The present application relates to the technical field of high-performance computers, and in particular, to a heterogeneous node computing system and a heterogeneous node computing method. Background technique [0002] The molecular simulation method in the field of high performance computing uses computers to establish molecular models to simulate the structure and dynamic behavior of molecules, so as to obtain various chemical and physical properties of molecular systems. Among them, molecular dynamics simulation (Molecular Dynamics, MD) Has been widely used in scientific research and engineering fields. Molecular dynamics simulations generally use periodic boundary conditions, and the essence of the calculation of electrostatic interactions is that for N-body problems under periodic boundary conditions, the interaction between each two atoms needs to be considered, so , which has a huge amount of computation and requires global communication. In o...

AI Technical Summary

Problems solved by technology

Since the PME algorithm needs to call the Fast Fourier transform (FFT) for calculation, and the FFT library currently implemented on the GPU only has a single-card version, the PME algorithm implemented on the GPU can only be calculated on a single card, which greatly reduces the Restricts the size of molecular dynamics system that can be simulated and calculated

At the same time, during the simulation calculation process of the GPU, the CPU is often idle, resulting in a waste of computing resources.

[0008] In addition, when multiple GPUs perform simulation calculations, communication and data exchange between GPUs are required. For the currently commonly used CPU+GPU heterogeneous computing cluster architecture, the computing efficiency is low, and there are serious constraints on hardware resources. waste

Images