Hartree approximation method for achieving acceleration through GPU

Abstract

The invention provides a Hartree approximation method for achieving acceleration through a GPU. An essential procedural framework and an organization flow include the steps that that a basis vector is selected, a tentative single-particle state is established, and Hartree items are generated and stored in a linear memory to serve as initial parameters; solution space of a question is generated; computational grids are divided and input parameters are transmitted to a video memory; a function on an equipment side is called for computing particle-particle interaction; a computed result is copied to the memory from the video memory; a Hartree matrix is generated and matrix diagonalization operation is conducted on a CPU; a new single-particle state is acquired through computation; according to the single-particle state acquired after self-consistent operation, physical quantities such as energy and state vectors of multiple systems are further computed. Compared with the prior art, the computed load characteristics of a Hartree approximation computation method are completely analyzed, various computation resources in a computer are fully used, and rapid simulation computation of multi-body physical and quantum computing chemical questions through the Hartree approximation method is greatly accelerated.

Description

technical field

[0001] The invention relates to the field of high-performance computing, in particular to a Hartree approximation method accelerated by GPU. Background technique

[0002] With the continuous progress of human society and the development of science and technology, people's understanding of nature has become more and more extensive, and gradually deepened into physical systems such as micro and nano. At the same time, the needs of biotechnology, chemical production and pharmaceuticals also urgently require people to accurately grasp the properties of objects at the microscopic level, especially at the electronic, atomic or molecular scale. But unfortunately, the microscopic system often includes tens of thousands of electrons and atoms. From a theoretical point of view, there is no strict analytical solution to the many-body problem. Therefore, solving the many-body problem has always been a difficult problem in the field of physics. In order to break through...

Examples

Embodiment Construction

[0019] A GPU-accelerated Hartree approximation method of the present invention will be described in detail below in conjunction with the accompanying drawings.

[0020] like figure 1 Shown, a kind of Hartree approximation method that utilizes GPU to accelerate of the present invention, its basic program structure and organizational flow are:

[0021] (1) Select the base vector and construct a tentative single-particle state, generate a Hartree term, and store it in a linear memory as an initial parameter;

[0022] (2) Generate the solution space of the problem;

[0023] (3) Divide the calculation grid, transfer the input parameters to the video memory, and put the parameters that need to be frequently accessed in the calculation into the shared memory of the GPU;

[0024] (4) Call the device-side function to calculate the particle-particle interaction, and store the results corresponding to each thread in the video memory;

[0025] (5) Copy the calculated results from the v...

technical field

[0001] The invention relates to the field of high-performance computing, in particular to a Hartree approximation method accelerated by GPU. Background technique

[0002] With the continuous progress of human society and the development of science and technology, people's understanding of nature has become more and more extensive, and gradually deepened into physical systems such as micro and nano. At the same time, the needs of biotechnology, chemical production and pharmaceuticals also urgently require people to accurately grasp the properties of objects at the microscopic level, especially at the electronic, atomic or molecular scale. But unfortunately, the microscopic system often includes tens of thousands of electrons and atoms. From a theoretical point of view, there is no strict analytical solution to the many-body problem. Therefore, solving the many-body problem has always been a difficult problem in the field of physics. In order to break through...

Examples

Embodiment Construction

[0019] A GPU-accelerated Hartree approximation method of the present invention will be described in detail below in conjunction with the accompanying drawings.

[0020] like figure 1 Shown, a kind of Hartree approximation method that utilizes GPU to accelerate of the present invention, its basic program structure and organizational flow are:

[0021] (1) Select the base vector and construct a tentative single-particle state, generate a Hartree term, and store it in a linear memory as an initial parameter;

[0022] (2) Generate the solution space of the problem;

[0023] (3) Divide the calculation grid, transfer the input parameters to the video memory, and put the parameters that need to be frequently accessed in the calculation into the shared memory of the GPU;

[0024] (4) Call the device-side function to calculate the particle-particle interaction, and store the results corresponding to each thread in the video memory;

[0025] (5) Copy the calculated results from the v...

Images