Compressible fluid D3Q19 model calculation method and device based on multi-core architecture, medium and product

Abstract

The invention provides a compressible fluid D3Q19 model calculation method and device based on a multi-core architecture, a medium and a product, an SOC chip architecture used for calculation is divided into a three-dimensional multi-logic operation core structure, each logic operation core carries out calculation of a D3Q19 lattice Boltzmann model according to corresponding processing data, and after each logic operation core completes calculation, a D3Q19 lattice Boltzmann model is obtained. Migration and collision operations of data between logic operation cores are carried out; obtaining result data according to the migration and collision operations; and iteratively executing until the result data of the required moment is obtained. The problem of simulation calculation of the D3Q19 lattice Boltzmann model on the flow of the compressible fluid is solved, the calculation process is accelerated, the data transmission efficiency is improved, and the overall operation performance is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of compressible fluid numerical simulation, and in particular provides a compressible fluid D3Q19 model calculation method and device, medium, and product based on a multi-core architecture. Background Art

[0002] A fluid with compressibility is a compressible fluid. Compressible fluids are widely used in engineering fields, such as jet engines, chemical industry, aerospace, etc. The study of the flow characteristics of compressible fluids is very important for the further application of compressible fluids. For example, in the compressible flow of gas, the density of the gas flow will change significantly. Although all gas flows are compressible, when the Mach number (the ratio of flow velocity to the speed of sound) is less than 0.3, the density change caused by the velocity is usually less than 5%, and the gas flow at this time will be regarded as an incompressible flow. One of the important means to study the characteristics of compressible flow is numerical simulation. Numerical simulation refers to the use of appropriate calculation methods and programs to use computers to numerically solve the fluid mechanics equations to correctly reflect the evolution of the fluid.

[0003] The central object of the D3Q19 lattice Boltzmann model is the particle distribution function f(x, t, e), which corresponds to the probability of finding a particle at a given position x with a given speed e. In the prior art, the D3Q19 lattice Boltzmann model can be used for numerical simulation of fluid mechanics, but the traditional method can only simulate incompressible flow problems and is difficult to solve the simulation problem of the flow characteristics of compressible fluids. In addition, the computational task of numerical simulation using the D3Q19 lattice Boltzmann model is complex, the amount of data is huge, the calculation process is slow, the data transmission efficiency is low, and the overall computing performance needs to be optimized. Summary of the invention

[0004] In view of this, the present invention aims to propose a calculation method and device, medium, and product of the D3Q19 lattice Boltzmann model of compressible fluid based on a multi-core architecture, which can solve the simulation calculation problem of the D3Q19 lattice Boltzmann model for compressible fluid flow, speed up the calculation process, improve data transmission efficiency, and optimize the overall computing performance.

[0005] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0006] A method for calculating a compressible fluid D3Q19 model based on a multi-core architecture, comprising:

[0007] S1. Divide the SOC chip architecture used for calculation into a three-dimensional multi-logic operation core structure, and divide the corresponding logic processing tasks for each logic operation core; divide the compressible fluid data in the D3Q19 lattice Boltzmann model equally according to the three-dimensional multi-logic operation core structure, and use the divided data as the processing data of the logic processing tasks of the logic operation cores at the corresponding positions;

[0008] S2. Each logic operation core performs calculation of the D3Q19 lattice Boltzmann model according to the corresponding processing data, wherein the calculation includes LBM calculation and FVM calculation and the merging of the calculation results of the LBM calculation and the FVM calculation, wherein the LBM calculation is used to solve the linear equation and momentum equation of the D3Q19 lattice Boltzmann model, and the FVM calculation is used to solve the entropy equation of the D3Q19 lattice Boltzmann model;

[0009] S3. After each logic operation core completes the calculation, it sends out the data that needs to be sent to other logic operation cores. After the transmission is completed, the migration and collision operations of the data between the logic operation cores are performed; if it is a multi-SOC chip architecture, the migration and collision operations of the data between the SOCs are performed again; the result data is obtained according to the migration and collision operations;

[0010] S4. Substitute the result data as the processing data for the next moment into step S2 for iterative execution until the result data at the required moment is obtained.

[0011] Furthermore, in step S1, the method for dividing the three-dimensional multi-logic operation core structure includes:

[0012] If it is a single SOC chip architecture, the physical computing core of the single SOC is divided into logical computing cores of an N×N×N three-dimensional architecture and numbered respectively;

[0013] If it is a multi-SOC chip architecture, the physical operation cores of multiple SOCs are first divided into a three-dimensional logical structure, and the corresponding logical processing tasks are divided for the physical operation core of each SOC. Then, the physical operation core of each SOC is divided into logical operation cores of an N×N×N three-dimensional architecture and numbered respectively.

[0014] Furthermore, step S1 also includes dividing the memory space into Ping / Pang areas to achieve parallel data transmission and calculation. When the data in the Ping area is being calculated, the data in the Pang area is being communicated and transmitted. When the data in the Ping area is being communicated and transmitted, the data in the Pang area is being calculated.

[0015] Furthermore, in step S3, when the data in the Ping area is undergoing collision migration, the data in the Pang area is undergoing communication and transmission; when the data in the Pang area is undergoing collision migration, the data in the Ping area is undergoing communication and transmission.

[0016] Furthermore, the method for calculating in step S2 includes:

[0017] S201, inputting the density, velocity, pressure, dimensionless temperature, and entropy of the compressible fluid at the initial moment in the processing data;

[0018] S202, calculating the mixed recursive regularized equilibrium distribution function of 19 discrete velocity directions of the D3Q19 model and the correction term of the D3Q19 model;

[0019] S203, calculating the mixed recursive regularized second-order Hermite polynomial coefficients of the D3Q19 model, and then calculating the mixed recursive regularized third-order Hermite polynomial coefficients according to the mixed recursive regularized second-order Hermite polynomial coefficients;

[0020] S204, calculating a regularization term based on the mixed recursive regularized second-order Hermite polynomial coefficients and the mixed recursive regularized third-order Hermite polynomial coefficients, and calculating a post-collision distribution function of 19 discrete velocity directions of the D3Q19 model at the next moment according to the regularization term;

[0021] S205, calculating the density and velocity at the next moment according to the distribution function of the 19 discrete velocity directions of the D3Q19 model after the collision, and completing the LBM calculation;

[0022] S206, using the density and pressure of the compressible fluid at the current moment, calculating the entropy at the current moment;

[0023] S207, calculating the entropy and dimensionless temperature at the next moment according to the entropy at the current moment, and completing the FVM calculation;

[0024] S208. The LBM calculation and FVM calculation results are combined to obtain the density, velocity, pressure, dimensionless temperature, and entropy at the next moment, and the calculation is completed.

[0025] Furthermore, in step S203, if the current moment is the initial moment, the mixed recursive regularized second-order Hermite polynomial coefficients are obtained according to the finite difference second-order Hermite polynomial coefficients; otherwise, the mixed recursive regularized second-order Hermite polynomial coefficients are obtained according to the recursive regularized second-order Hermite polynomial coefficients and the finite difference second-order Hermite polynomial coefficients.

[0026] On the other hand, the present invention also proposes a compressible fluid D3Q19 model calculation device based on a multi-core architecture, comprising:

[0027] Partitioning module: divides the SOC chip architecture used for calculation into a three-dimensional multi-logic operation core structure, and divides the corresponding logic processing tasks for each logic operation core; divides the compressible fluid data in the D3Q19 lattice Boltzmann model equally according to the three-dimensional multi-logic operation core structure, and the divided data are used as processing data for the logic processing tasks of the logic operation cores at the corresponding positions;

[0028] Computation module: Each logic operation core performs calculation of the D3Q19 lattice Boltzmann model according to the corresponding processing data, and the calculation includes LBM calculation and FVM calculation and the combination of the calculation results of the LBM calculation and the FVM calculation. The LBM calculation is used to solve the linear equation and momentum equation of the D3Q19 lattice Boltzmann model, and the FVM calculation is used to solve the entropy equation of the D3Q19 lattice Boltzmann model.

[0029] Migration and collision module: After each logic operation core completes the calculation, it sends out the data that needs to be sent to other logic operation cores. After the transmission is completed, the migration and collision operations of the data between the logic operation cores are performed; if it is a multi-SOC chip architecture, the migration and collision operations of the data between the SOCs are performed again; the result data is obtained according to the migration and collision operations;

[0030] Iterative execution module: the result data is used as the processing data for the next moment, and is substituted into step S2 for iterative execution until the result data at the required moment is obtained.

[0031] Furthermore, the division module includes a single SOC unit and a multi-SOC unit:

[0032] Single SOC unit: The physical computing core of a single SOC is divided into logical computing cores of an N×N×N three-dimensional architecture and numbered separately;

[0033] Multi-SOC unit: First, divide the physical operation cores of multiple SOCs into a three-dimensional logical structure, divide the corresponding logical processing tasks for the physical operation core of each SOC, and then divide the physical operation core of each SOC into logical operation cores of N×N×N three-dimensional architecture, and number them respectively.

[0034] Furthermore, the division module also includes a regional control unit, which divides the memory space into Ping / Pang areas to achieve parallel data transmission and calculation. When the Ping area data is calculated, the Pang area data is communicated and transmitted; when the Ping area data is communicated and transmitted, the Pang area data is calculated.

[0035] Furthermore, the area control unit also includes that when the Ping area data performs collision migration, the Pang area data performs communication and transmission; when the Pang area data performs collision migration, the Ping area data performs communication and transmission.

[0036] Furthermore, the calculation module includes:

[0037] Input unit: input the density, velocity, pressure, dimensionless temperature and entropy of the compressible fluid at the initial moment in the processing data;

[0038] The first calculation unit: calculates the mixed recursive regularized equilibrium distribution function of 19 discrete velocity directions of the D3Q19 model and the correction term of the D3Q19 model;

[0039] The second calculation unit: calculates the mixed recursive regularized second-order Hermite polynomial coefficients of the D3Q19 model, and then calculates the mixed recursive regularized third-order Hermite polynomial coefficients according to the mixed recursive regularized second-order Hermite polynomial coefficients;

[0040] The third calculation unit: calculates the regularization term based on the mixed recursive regularization second-order Hermitian polynomial coefficient and the mixed recursive regularization third-order Hermitian polynomial coefficient, and calculates the distribution function of the D3Q19 model after the collision in 19 discrete velocity directions at the next moment according to the regularization term;

[0041] The fourth calculation unit: calculates the density and velocity at the next moment according to the distribution function of the 19 discrete velocity directions of the D3Q19 model after the collision, and completes the LBM calculation;

[0042] The fifth calculation unit: using the density and pressure of the compressible fluid at the current moment, calculates the entropy at the current moment;

[0043] The sixth calculation unit: calculates the entropy and dimensionless temperature at the next moment according to the entropy at the current moment, and completes the FVM calculation;

[0044] Merging unit: The LBM calculation and FVM calculation results are merged to obtain the density, velocity, pressure, dimensionless temperature, and entropy at the next moment to complete the calculation.

[0045] Furthermore, in the second calculation unit, if the current moment is the initial moment, the mixed recursive regularized second-order Hermite polynomial coefficients are obtained according to the finite difference second-order Hermite polynomial coefficients; otherwise, the mixed recursive regularized second-order Hermite polynomial coefficients are obtained according to the recursive regularized second-order Hermite polynomial coefficients and the finite difference second-order Hermite polynomial coefficients.

[0046] On the other hand, the present invention also provides a storage medium containing computer executable instructions, which, when executed by a computer processor, are used to execute the compressible fluid D3Q19 model calculation method based on a multi-core architecture as described above.

[0047] On the other hand, the present invention also provides a computer program product, which stores at least one program code, and the at least one program code is used to be executed by a processor to implement the above-mentioned compressible fluid D3Q19 model calculation method based on a multi-core architecture.

[0048] Compared with the prior art, the compressible fluid D3Q19 model calculation method, device, medium, and product based on multi-core architecture described in the present invention have the following advantages:

[0049] 1. The present invention uses a hybrid recursive regularization method based on lattice Boltzmann, which can solve the problem of numerical simulation calculation of compressible fluid passing through the D3Q19 field;

[0050] 2. The present invention combines the characteristics of hardware chips and proposes a method of dividing multiple logic operation cores for parallel computing, which can accelerate the calculation process of numerical simulation calculation of compressible fluid passing through the D3Q19 field;

[0051] 3. The present invention combines the characteristics of hardware chips to divide the memory space into Ping / Pang areas, realizes the parallelization of data transmission and calculation, and improves data transmission efficiency;

[0052] 4. The present invention optimizes the calculation method steps of the D3Q19 lattice Boltzmann model and improves the calculation performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0054] Figure 1 This is a schematic diagram of the method flow of Example 1 of the present invention;

[0055] Figure 2 This is a schematic diagram of the D3Q19 lattice Boltzmann model described in Example 1 of the present invention;

[0056] Figure 3 This is a schematic diagram of the structure of a single SOC multi-logic operation core according to Embodiment 1 of the present invention;

[0057] Figure 4This is a schematic diagram of the structure of the multi-SOC multi-logic operation core according to Embodiment 1 of the present invention;

[0058] Figure 5 This is a schematic diagram of the calculation process of the hybrid recursive regularized D3Q19 lattice Boltzmann model described in Example 1 of the present invention;

[0059] Figure 6 This is a schematic diagram of a single SOC multi-logic operation core processing flow of Embodiment 1 of the present invention;

[0060] Figure 7 Schematic diagram of the multi-SOC multi-logic operation core processing flow of embodiment 1 of the present invention. DETAILED DESCRIPTION

[0061] The specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. It should also be noted that, for the convenience of description, only the parts related to the present invention, but not all structures, are shown in the drawings.

[0062] The design concept of the present invention is to use a multi-core architecture to accelerate the calculation of the compressible fluid D3Q19 lattice Boltzmann model based on hybrid recursive regularization.

[0063] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments.

[0064] Embodiment 1:

[0065] In Example 1, a D3Q19 model calculation method for compressible fluid based on a multi-core architecture is proposed, such as Figure 1 As shown, the following steps are included:

[0066] S1. Divide the SOC chip architecture used for calculation into a three-dimensional multi-logic operation core structure, and divide the corresponding logic processing tasks for each logic operation core; divide the compressible fluid data in the D3Q19 lattice Boltzmann model equally according to the three-dimensional multi-logic operation core structure, and use the divided data as the processing data of the logic processing tasks of the logic operation cores at the corresponding positions;

[0067] S2. Each logic operation core performs calculation of the D3Q19 lattice Boltzmann model according to the corresponding processing data, wherein the calculation includes LBM calculation and FVM calculation and the merging of the calculation results of the LBM calculation and the FVM calculation, wherein the LBM calculation is used to solve the linear equation and momentum equation of the D3Q19 lattice Boltzmann model, and the FVM calculation is used to solve the entropy equation of the D3Q19 lattice Boltzmann model;

[0068] S3. After each logic operation core completes the calculation, it sends out the data that needs to be sent to other logic operation cores. After the transmission is completed, the migration and collision operations of the data between the logic operation cores are performed; if it is a multi-SOC chip architecture, the migration and collision operations of the data between the SOCs are performed again; the result data is obtained according to the migration and collision operations;

[0069] S4. Substitute the result data as the processing data for the next moment into step S2 for iterative execution until the result data at the required moment is obtained.

[0070] Each step is described in detail below.

[0071] 1. Step S1:

[0072] The schematic diagram of the discrete velocity e of the D3Q19 model is as follows Figure 2 As shown, Figure 2 The magnitudes of the discrete velocities in the 19 directions are:

[0073]

[0074] In this embodiment, multi-core partitioning is performed based on the chip architecture of Silang Technology.

[0075] 1. Single SOC architecture:

[0076] Due to the characteristics of D3Q19, D3 is three-dimensional and Q19 is a 19-state distribution function. Therefore, the physical computing cores can be logically divided into logical three-dimensional position relationships. According to the characteristics of D3Q19, the computing data volume and data transmission volume of the computing cores are balanced, so as to achieve the purpose of load balancing through multiple cores, thereby achieving faster overall computing speed.

[0077] Based on this, the physical computing core in a single SOC is divided into multiple logical operation cores and numbered, and divided into a three-dimensional structure of N×N×N. Figure 3 As shown in the figure, it can be divided into a 3×3×3 three-dimensional logic operation core structure, and as Figure 3 Each logic operation core number is shown.

[0078] The compressible fluid data in the D3Q19 three-dimensional model is then divided into 3×3×3 equal parts. Each part of the data corresponds to a logic operation core, which serves as the processing data of the corresponding logic operation core and performs numerical simulation operations, thereby achieving more balanced calculation and transmission.

[0079] 2. Multi-SOC architecture:

[0080] In multiple SOCs, it is necessary to first divide multiple physical computing SOCs into a logical three-dimensional structure, number each SOC and divide the corresponding logical processing and tasks. Each SOC also divides the physical computing core into multiple three-dimensional logical operation cores and numbers them to meet balanced computing and transmission, achieve a balance between computing and transmission, make full use of processor resources, and achieve the purpose of acceleration.

[0081] In this embodiment, Figure 4 As shown, multiple SOCs are first divided into a 3×3×3 logical three-dimensional structure, and each SOC is divided into a 3×3×3 three-dimensional logical operation core structure, which are numbered separately. The compressible fluid data in the D3Q19 three-dimensional model is divided and processed for each logical operation core according to the corresponding division method.

[0082] 3. Regardless of single SOC or multi-SOC, after completing the division of the logic operation core, the memory space is divided into Ping / Pang areas to achieve parallel transmission and calculation.

[0083] 2. Step S2:

[0084] The solution of step S2 is mainly the calculation process of each logic operation core for the corresponding processing data after the logic operation core of step S1 is divided.

[0085] The present invention calculates the compressible fluid D3Q19 lattice Boltzmann model based on mixed recursive regularization. The calculation principle is relatively complicated, and the calculation principle is first explained below.

[0086] 1. Macro control equation:

[0087] The macro equations corresponding to the recursive regularized D3Q19 lattice Boltzmann model for compressible fluid mixing are as follows:

[0088]

[0089]

[0090]

[0091] Equation (1.1) is the mass conservation equation, (1.2) is the momentum conservation equation, and (1.3) is the entropy equation, where ρ is the gas density, p is the pressure, u is the velocity, s is the entropy, T is the temperature, and λ is the thermal conductivity; is the viscous dissipation term; t represents time, and x represents the spatial coordinate;

[0092] α and β represent the Einstein summation subscripts. If Equation 1.2 is written in scalar form, when α represents the x direction, β is x, y, and z respectively; when α represents the y direction, β is x, y, and z respectively; when α represents the z direction, β is x, y, and z respectively.

[0093] Viscous stress Π αβ The expression is:

[0094]

[0095] In formula (1.4), μ is the fluid dynamic viscosity; It is called the modified viscosity coefficient. δ αβ is the symbol for the Kronecker product; γ is the representation of the Einstein summation convention, that is

[0096] The closed thermodynamic equation is:

[0097]

[0098] In formula (1.5), c v is the constant volume specific heat capacity; γ' is the specific heat ratio, and in the case of an ideal gas, it is the constant pressure specific heat capacity c p and constant volume specific heat c v The ratio of R is the ideal gas constant; θ is the dimensionless temperature, T0 is the reference temperature.

[0099] 2. Hybrid recursive regularized D3Q19 lattice Boltzmann model:

[0100] The present invention uses a hybrid recursive regularized D3Q19 model to handle the problem of smooth compressible flow from subsonic to sonic speed. In the hybrid recursive regularized D3Q19 lattice Boltzmann model, the mass conservation equation (1.1) and the momentum conservation equation (1.2) are calculated and solved using the lattice Boltzmann method LBM, and the entropy equation (1.3) is calculated and solved using the finite volume method FVM. The hybrid recursive regularized D3Q19 lattice Boltzmann algorithm for compressible flow can be expressed as:

[0101]

[0102] f i (x α +c iα ,t+δ t ) = f′ i (x α ,t) (1.7)

[0103] Formula (1.6) is the collision process, and formula (1.7) is the migration process. The collision process refers to the process in which particles interact with each other at the grid nodes in each time step; the migration process refers to the process in which particles after collision move from one grid node to an adjacent grid node according to their velocity direction in each time step.

[0104] f i represents the distribution function, f′ i Represents the distribution function after the collision. The distribution function after the collision is the distribution of particles in each discrete velocity direction after the collision process. It is a description of the instantaneous state of particles after they collide with each other at the grid nodes in each time step. It is a key intermediate quantity in the LBM calculation process and is used for the subsequent migration process and the calculation of macroscopic physical quantities.

[0105] is the equilibrium distribution function, is the non-equilibrium distribution function;

[0106] is the dimensionless relaxation time, which is determined by the dynamic viscosity μ, pressure p and time step δ t express, represents the mixed recursive regularization term, where the non-equilibrium distribution function Ψ i Indicates a correction item.

[0107] Where i represents the discrete velocity direction, and the D3Q19 model has 19 discrete velocity directions. t Indicates the time step. c iα Here, i represents the discrete velocity direction (19 in total), and α is the coordinate direction represented by the Einstein summation subscript notation. 1x It is the discrete velocity magnitude of the first discrete velocity direction in the x-coordinate direction.

[0108] The macroscopic density and velocity are expressed by equations (1.8) and (1.9):

[0109] ρ=∑ i f i (1.8)

[0110]

[0111] The equilibrium distribution function of the D3Q19 model is shown by formula (1.10):

[0112]

[0113] The second-order Hermite polynomial Third-order Hermite polynomial [ci δ] αβγ =c iα δ βγ +c iβ δ αγ +c iγ δ αβ , where δ αβ is the symbol of the Kronecker product, when α=β, δ αβ =1, otherwise δ αβ = 0. The second-order coefficient of the Hermite polynomial The third-order coefficients of the Hermite polynomial where [uδ] αβγ =u α δ βγ +u β δ αγ +u γ δ αβ .

[0114] x, y, z are the specific expanded representations of α, β, γ, which are actual coordinate representations. Because formula (1.10) is more important, it is expanded in this way.

[0115] Dimensionless temperature satisfy Given the ideal gas state equation, c s is the lattice sound speed, In actual calculations, it is generally used Calculate the dimensionless temperature, where T is the true temperature, T0 is the reference temperature, and the time step and grid spacing formulas are

[0116] δ t is the time step, δ x is the spatial step length, and the two are user-set parameters. c is the grid sound velocity, which is the spatial step length divided by the spatial step length.

[0117] The thermodynamic closed equation under the grid unit is:

[0118]

[0119] By Mayer's relation Get the dimensionless specific heat capacity at constant volume and specific heat capacity at constant pressure in grid units:

[0120]

[0121] In the mixed recursive model for compressible flow, the regularization term of the D3Q19 model is:

[0122]

[0123]

[0124] In formula (1.13), are the coefficients of the mixed recursive regularized second-order Hermite polynomial, are the coefficients of the recursively regularized second-order Hermite polynomial, are the coefficients of the finite difference second-order Hermite polynomial, The function of these four formulas is to find the coefficients of the mixed recursive regularized second-order Hermite polynomial. and mixed recursive regularized third-order Hermite polynomial coefficients Then bring it into equation (1.13) and solve the regularization term Then the regularization term Collision equation (1.6).

[0125]

[0126]

[0127]

[0128]

[0129] σ is the weight in formula (1.14), which is given by the user and can be a number between 0 and 1.

[0130] There are three methods for calculating the numerical gradient operator, namely the second-order isotropic central difference format, the classical central difference format and the upwind format, as shown in equations (1.18) to (1.20) respectively:

[0131]

[0132]

[0133]

[0134] The method of finding the gradient itself has no practical significance. In fact, it is to use the three formulas (1.18), (1.19) and (1.20) to find the partial derivative of the velocity u in (1.16) with respect to the spatial coordinates. i is the discrete velocity, ω i is the weight coefficient. Indicates the grid position. t represents the time step, t represents the time, and Δx represents the grid spacing.

[0135] Equations (1.18), (1.19), and (1.20) can all be used to calculate the partial derivatives of the velocity u in equation (1.16) with respect to the spatial coordinates. When performing specific calculations, choose one of them depending on the situation. When there is a shock wave problem, use equation (1.20).

[0136] Correction term Ψ of the D3Q19 model i It is an item in the collision equation (1.6). Its function is to improve numerical accuracy and enhance numerical stability. The expression is shown in equation (1.21):

[0137]

[0138] These six variables are second-order Hermite polynomials.

[0139] 3. Finite volume method to solve the entropy equation:

[0140] The entropy equation (1.3) is solved separately using the finite volume method FVM. The time integral of the entropy equation with viscous dissipation term can be solved using the explicit Euler method or the Runge-Kutta method. The first-order explicit Euler method is used as an example. The finite volume method is used to discretize equation (1.3) and we get:

[0141] s n+1 =s n +δ t [RHS C (s n )+RHS D (s n )+Φ n ](1.22)

[0142] RHS in formula (1.22) C represents the convection term, RHS D represents the diffusion term, Φ n Represents the source term related to the heat generated by the viscous dissipation term. The convection flux uses the MUSCL format, and the diffusion term and the viscous dissipation term use the classical second-order central difference format. The MUSCL format is described as follows:

[0143]

[0144] by For example,

[0145]

[0146]

[0147]

[0148] in When κ = -1, it is a second-order upwind format, and when κ = 1, it is a central difference format. When it is a third-order headwind format, it is generally taken in the calculation For the Van Albada limiter, ε is a small value, usually ε=10 -6 .

[0149] 4. Shock wave capture technology:

[0150] For discontinuous compressible flows, the hybrid recursive regularized lattice Boltzmann D3Q19 model uses the shock sensor and the associated artificial viscosity used in the JST (Jameson-Schmidt Turkel) scheme to handle shock waves, where the effective relaxation time is Where μ is the fluid dynamic viscosity, The dimensionless relaxation time is finally obtained Used in studies with shock waves.

[0151] 5. The calculation process of the logic operation core:

[0152] Based on the above calculation principle, the calculation process of each logic operation core is as follows: Figure 5 As shown, including:

[0153] (1) Input the corresponding processing data, including the density, velocity, pressure, dimensionless temperature, and entropy of the compressible fluid at the current moment; (the density, velocity, pressure, dimensionless temperature, and entropy of the compressible fluid at the initial moment are input in the initial calculation);

[0154] (2) Calculate the mixed recursive regularized equilibrium distribution function of the 19 discrete velocity directions of the D3Q19 model according to equation (1.10); and the correction term of the D3Q19 model according to equation (1.21);

[0155] (3) Determine whether the current moment is the initial moment;

[0156] (4) If it is the initial moment, the mixed recursive regularized second-order Hermitian polynomial coefficients are obtained according to the finite difference second-order Hermitian polynomial coefficients; Then according to formula (1.17), the mixed recursive regularized second-order Hermitte polynomial coefficients are used to calculate the mixed recursive regularized third-order Hermitte polynomial coefficients;

[0157] (5) If it is not the initial moment, then according to equations (1.14)-(1.16), the mixed recursive regularized second-order Hermitte polynomial coefficients of the D3Q19 model are calculated according to the recursive regularized second-order Hermitte polynomial coefficients and the finite difference second-order Hermitte polynomial coefficients, and then according to equation (1.17), the mixed recursive regularized second-order Hermitte polynomial coefficients are used to calculate the mixed recursive regularized third-order Hermitte polynomial coefficients;

[0158] (6) Based on the mixed recursive regularized second-order Hermitian polynomial coefficients and the mixed recursive regularized third-order Hermitian polynomial coefficients, the regularization term is calculated according to formula (1.13), and based on the regularization term, the post-collision distribution function of the 19 discrete velocity directions of the D3Q19 model at the next moment is calculated according to formulas (1.6) and (1.7);

[0159] (7) According to the distribution function of the 19 discrete velocity directions of the D3Q19 model after the collision at the next moment, and according to equations (1.8) and (1.9), the density and velocity at the next moment are calculated to complete the LBM calculation;

[0160] (8) Using the density and pressure of the compressible fluid at the current moment, calculate the entropy at the current moment according to formula (1.11);

[0161] (9) Based on the entropy at the current moment and the density at the next moment obtained in step (7), use formula (1.22) to calculate the entropy at the next moment, and use formula (1.11) to calculate the dimensionless temperature at the next moment to complete the FVM calculation.

[0162] After the above calculations are completed, according to Figure 5 , the LBM calculation and FVM calculation results are combined, and after time step iteration, the required compressible fluid data at time n+1 is obtained as output data. However, time step iteration also requires migration collision operations between logical operation cores and migration collision operations between SOC data in a multi-SOC architecture, which is specifically the content of step S3 below.

[0163] 6. Ping / Pang area parallel:

[0164] Since the memory space is divided into Ping / Pang areas, the above-mentioned calculation process and operations such as data transmission and inter-core communication can be carried out in parallel. For example, when the Ping area data is calculated, the Pang area data can be transmitted. When the Ping area data is communicating between cores, the Pang area can be calculated. When the Ping area data is colliding and migrating, the Pang area data can communicate between cores. At the same time, when the Pang area data is colliding and migrating, the next group of Ping area data can be obtained for transmission, and the next group of calculation and transmission can be started. By using the idea of ​​parallel operation of multiple logic operation cores and multiple groups of Ping / Pang areas, the computing units of the processor can be fully utilized to achieve the purpose of accelerated processing.

[0165] 3. Step S3 and Step S4:

[0166] Based on the calculation process of step S2, steps S3 and S4 propose an overall processing flow of multiple logic operation cores, which is divided into two processes: single SOC and multi-SOC.

[0167] 1. Single SOC:

[0168] The overall operation process of a single SOC is as follows Figure 6 As shown, including:

[0169] After the three-dimensional logic operation core architecture is divided, the logic operation core is initialized;

[0170] Parameter initialization, configuration number and other parameters;

[0171] Data initialization: input the processing data of the compressible fluid corresponding to each logic operation core;

[0172] Data is divided into data blocks corresponding to the Ping / Pang area;

[0173] Start time step iteration;

[0174] Transmit Ping zone data;

[0175] The Ping area data is subjected to LBM calculation and FVM calculation as described in step S2, and the Pang area data is transmitted while the calculation is being performed;

[0176] After the LBM and FVM calculations are completed for the Ping area data, inter-core communication is performed, while the LBM and FVM calculations are performed for the Pang area data at the same time;

[0177] After the Ping zone data is communicated between cores, collision migration operation is performed. Meanwhile, after the Pang zone data completes LBM calculation and FVM calculation, inter-core communication is performed.

[0178] After the collision migration operation of the Ping area data, the next data block is iteratively calculated. At the same time, the collision migration operation is performed after the communication between the Pang area data cores.

[0179] After the Pang area data collision migration operation, the next data block is iteratively calculated;

[0180] Iterate the above steps until the iteration ends and output the result data.

[0181] 2. Multiple SOCs:

[0182] The process of multiple SOCs is similar to that of a single SOC, which increases data transmission between SOCs, utilizes the communication bandwidth between SOCs, reduces the amount of calculation in each SOC, or increases the overall computing volume, is able to calculate larger-scale data, utilizes multiple groups of SOCs, accelerates calculations, and improves performance.

[0183] like Figure 7 As shown, including:

[0184] After the three-dimensional logic operation core architecture is divided, it is initialized;

[0185] Parameter initialization, configuration number and other parameters;

[0186] Data initialization: input the processing data of the compressible fluid corresponding to each logic operation core;

[0187] Data is divided into data blocks corresponding to the Ping / Pang area;

[0188] Multi-SOC synchronization;

[0189] Start time step iteration;

[0190] Transmit Ping zone data;

[0191] The Ping area data is subjected to LBM calculation and FVM calculation as described in step S2, and the Pang area data is transmitted while the calculation is being performed;

[0192] After the LBM and FVM calculations are completed for the Ping area data, inter-core communication is performed, while the LBM and FVM calculations are performed for the Pang area data at the same time;

[0193] After the Ping area data is communicated between cores, the inter-core collision migration operation is performed. At the same time, after the Pang area data completes the LBM calculation and FVM calculation, the inter-core communication is performed;

[0194] After the Ping area data inter-core collision migration operation, the SOC collision migration operation is performed. At the same time, after the Pang area data inter-core communication, the inter-core collision migration operation is performed.

[0195] After the collision migration operation between SOCs of Ping area data, the next data block is iteratively calculated. At the same time, after the collision migration operation between cores of Pang area data, the collision migration operation between SOCs is performed.

[0196] After the collision migration operation between SOCs of Pang area data, the next data block is calculated iteratively;

[0197] Iterate the above steps until the iteration ends and output the result data.

[0198] The method proposed in this embodiment 1 divides multiple logical operation cores and divides the memory space into Ping / Pang areas to realize parallel computing. The calculation process is based on the hybrid recursive regularization method of lattice Boltzmann, which can realize the numerical simulation of the compressible fluid D3Q19 model, accelerate the calculation process, improve data transmission capabilities, and improve computing performance.

[0199] Embodiment 2:

[0200] Embodiment 2 of the present invention proposes a compressible fluid D3Q19 model computing device based on a multi-core architecture, comprising:

[0201] Partitioning module: divides the SOC chip architecture used for calculation into a three-dimensional multi-logic operation core structure, and divides the corresponding logic processing tasks for each logic operation core; divides the compressible fluid data in the D3Q19 lattice Boltzmann model equally according to the three-dimensional multi-logic operation core structure, and the divided data are used as processing data for the logic processing tasks of the logic operation cores at the corresponding positions;

[0202] Computation module: Each logic operation core performs calculation of the D3Q19 lattice Boltzmann model according to the corresponding processing data, and the calculation includes LBM calculation and FVM calculation and the combination of the calculation results of the LBM calculation and the FVM calculation. The LBM calculation is used to solve the linear equation and momentum equation of the D3Q19 lattice Boltzmann model, and the FVM calculation is used to solve the entropy equation of the D3Q19 lattice Boltzmann model.

[0203] Migration and collision module: After each logic operation core completes the calculation, it sends out the data that needs to be sent to other logic operation cores. After the transmission is completed, the migration and collision operations of the data between the logic operation cores are performed; if it is a multi-SOC chip architecture, the migration and collision operations of the data between the SOCs are performed again; the result data is obtained according to the migration and collision operations;

[0204] Iterative execution module: the result data is used as the processing data for the next moment, and is substituted into step S2 for iterative execution until the result data at the required moment is obtained.

[0205] Among them, the division module includes single SOC unit and multi-SOC unit:

[0206] Single SOC unit: The physical computing core of a single SOC is divided into logical computing cores of an N×N×N three-dimensional architecture and numbered separately;

[0207] Multi-SOC unit: First, divide the physical operation cores of multiple SOCs into a three-dimensional logical structure, divide the corresponding logical processing tasks for the physical operation core of each SOC, and then divide the physical operation core of each SOC into logical operation cores of N×N×N three-dimensional architecture, and number them respectively.

[0208] The division module also includes an area control unit, which divides the memory space into Ping / Pang areas to achieve parallel data transmission and calculation. When the data in the Ping area is calculated, the data in the Pang area is communicated and transmitted. When the data in the Ping area is communicated and transmitted, the data in the Pang area is calculated.

[0209] The area control unit also includes: when the Ping area data performs collision migration, the Pang area data performs communication and transmission; when the Pang area data performs collision migration, the Ping area data performs communication and transmission.

[0210] In addition, the calculation module includes:

[0211] Input unit: input the density, velocity, pressure, dimensionless temperature and entropy of the compressible fluid at the initial moment in the processing data;

[0212] The first calculation unit: calculates the mixed recursive regularized equilibrium distribution function of 19 discrete velocity directions of the D3Q19 model and the correction term of the D3Q19 model;

[0213] The second calculation unit: calculates the mixed recursive regularized second-order Hermite polynomial coefficients of the D3Q19 model, and then calculates the mixed recursive regularized third-order Hermite polynomial coefficients according to the mixed recursive regularized second-order Hermite polynomial coefficients;

[0214] The third calculation unit: calculates the regularization term based on the mixed recursive regularization second-order Hermitian polynomial coefficient and the mixed recursive regularization third-order Hermitian polynomial coefficient, and calculates the distribution function of the D3Q19 model after the collision in 19 discrete velocity directions at the next moment according to the regularization term;

[0215] The fourth calculation unit: calculates the density and velocity at the next moment according to the distribution function of the 19 discrete velocity directions of the D3Q19 model after the collision, and completes the LBM calculation;

[0216] The fifth calculation unit: using the density and pressure of the compressible fluid at the current moment, calculates the entropy at the current moment;

[0217] The sixth calculation unit: calculates the entropy and dimensionless temperature at the next moment according to the entropy at the current moment, and completes the FVM calculation;

[0218] Merging unit: The LBM calculation and FVM calculation results are merged to obtain the density, velocity, pressure, dimensionless temperature, and entropy at the next moment to complete the calculation.

[0219] In the second calculation unit, if the current moment is the initial moment, the mixed recursive regularized second-order Hermite polynomial coefficients are obtained according to the finite difference second-order Hermite polynomial coefficients; otherwise, the mixed recursive regularized second-order Hermite polynomial coefficients are obtained according to the recursive regularized second-order Hermite polynomial coefficients and the finite difference second-order Hermite polynomial coefficients.

[0220] The compressible fluid D3Q19 model calculation device based on multi-core architecture proposed in the embodiment of the present invention can execute the compressible fluid D3Q19 model calculation method based on multi-core architecture provided in Example 1 of the present invention, and has functional modules corresponding to the execution method and the same beneficial effects as the method.

[0221] Embodiment 3:

[0222] Embodiment 4 of the present invention further provides a storage medium comprising computer executable instructions, which, when executed by a computer processor, are used to execute the compressible fluid D3Q19 model calculation method based on a multi-core architecture as provided in Embodiment 1 above.

[0223] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0224] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0225] Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0226] Computer program code for performing the operation of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer, partially on a remote computer, or entirely on a remote computer or device. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0227] Embodiment 4:

[0228] An embodiment of the present invention provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the compressible fluid D3Q19 model calculation method based on a multi-core architecture provided in the above-mentioned embodiment 1.

[0229] Note that the above are only preferred embodiments of the present invention and the technical principles used. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of the present invention, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. A method for calculating the compressible fluid D3Q19 model based on a multi-core architecture, characterized in that: include: S1. Divide the SOC chip architecture used for computing into a three-dimensional multi-logic operation core structure, and divide the corresponding logic processing tasks for each logic operation core; The compressible fluid data in the D3Q19 lattice Boltzmann model is equally divided according to the three-dimensional multi-logic operation core structure, and the divided data are used as the processing data of the logic processing tasks of the corresponding position logic operation core; S2. Each logic operation core performs calculation of the D3Q19 lattice Boltzmann model according to the corresponding processing data, wherein the calculation includes LBM calculation and FVM calculation and the merging of the calculation results of the LBM calculation and the FVM calculation, wherein the LBM calculation is used to solve the linear equation and momentum equation of the D3Q19 lattice Boltzmann model, and the FVM calculation is used to solve the entropy equation of the D3Q19 lattice Boltzmann model; S3, after each logic operation core completes the calculation, it sends the data that needs to be sent to other logic operation cores, and after the transmission is completed, the data migration and collision operations between the logic operation cores are performed; If it is a multi-SOC chip architecture, then the migration and collision operations of data between SOCs are performed; the result data is obtained according to the migration and collision operations; S4. Substitute the result data as the processing data for the next moment into step S2 for iterative execution until the result data at the required moment is obtained.

2. The method for calculating the compressible fluid D3Q19 model based on a multi-core architecture according to claim 1, characterized in that: In step S1, the method for dividing the three-dimensional multi-logic operation core structure includes: If it is a single SOC chip architecture, the physical computing core of the single SOC is divided into logical computing cores of N×N×N three-dimensional architecture and numbered respectively; If it is a multi-SOC chip architecture, the physical operation cores of multiple SOCs are first divided into a three-dimensional logical structure, and the corresponding logical processing tasks are divided for the physical operation core of each SOC. Then, the physical operation core of each SOC is divided into logical operation cores of an N×N×N three-dimensional architecture and numbered respectively.

3. The method for calculating the compressible fluid D3Q19 model based on a multi-core architecture according to claim 1, characterized in that: Step S1 also includes dividing the memory space into Ping / Pang areas to achieve parallel data transmission and calculation. When the data in the Ping area is being calculated, the data in the Pang area is being communicated and transmitted. When the data in the Ping area is being communicated and transmitted, the data in the Pang area is being calculated.

4. The method for calculating the compressible fluid D3Q19 model based on a multi-core architecture according to claim 3 is characterized in that: In step S3, when the data in the Ping area is undergoing collision migration, the data in the Pang area is undergoing communication and transmission; when the data in the Pang area is undergoing collision migration, the data in the Ping area is undergoing communication and transmission.

5. The method for calculating the compressible fluid D3Q19 model based on a multi-core architecture according to claim 1, characterized in that: The method for calculating in step S2 includes: S201, inputting the density, velocity, pressure, dimensionless temperature, and entropy of the compressible fluid at the initial moment in the processing data; S202, calculating the mixed recursive regularized equilibrium distribution function of 19 discrete velocity directions of the D3Q19 model and the correction term of the D3Q19 model; S203, calculating the mixed recursive regularized second-order Hermite polynomial coefficients of the D3Q19 model, and then calculating the mixed recursive regularized third-order Hermite polynomial coefficients according to the mixed recursive regularized second-order Hermite polynomial coefficients; S204, calculating a regularization term based on the mixed recursive regularized second-order Hermite polynomial coefficients and the mixed recursive regularized third-order Hermite polynomial coefficients, and calculating a post-collision distribution function of 19 discrete velocity directions of the D3Q19 model at the next moment according to the regularization term; S205, calculating the density and velocity at the next moment according to the distribution function of the 19 discrete velocity directions of the D3Q19 model after the collision, and completing the LBM calculation; S206, using the density and pressure of the compressible fluid at the current moment, calculating the entropy at the current moment; S207, calculating the entropy and dimensionless temperature at the next moment according to the entropy at the current moment, and completing the FVM calculation; S208. The LBM calculation and FVM calculation results are combined to obtain the density, velocity, pressure, dimensionless temperature, and entropy at the next moment, and the calculation is completed.

6. The method for calculating the compressible fluid D3Q19 model based on a multi-core architecture according to claim 5, characterized in that: In step S203, if the current moment is the initial moment, the mixed recursive regularized second-order Hermite polynomial coefficients are obtained according to the finite difference second-order Hermite polynomial coefficients; otherwise, the mixed recursive regularized second-order Hermite polynomial coefficients are obtained according to the recursive regularized second-order Hermite polynomial coefficients and the finite difference second-order Hermite polynomial coefficients.

7. A compressible fluid D3Q19 model computing device based on a multi-core architecture, characterized in that: include: Partitioning module: divides the SOC chip architecture used for computing into a three-dimensional multi-logic operation core structure, and divides the corresponding logic processing tasks for each logic operation core; The compressible fluid data in the D3Q19 lattice Boltzmann model is equally divided according to the three-dimensional multi-logic operation core structure, and the divided data are used as the processing data of the logic processing tasks of the corresponding position logic operation core; Computation module: Each logic operation core performs calculation of the D3Q19 lattice Boltzmann model according to the corresponding processing data, and the calculation includes LBM calculation and FVM calculation and the combination of the calculation results of the LBM calculation and the FVM calculation. The LBM calculation is used to solve the linear equation and momentum equation of the D3Q19 lattice Boltzmann model, and the FVM calculation is used to solve the entropy equation of the D3Q19 lattice Boltzmann model. Migration and collision module: After each logic operation core completes the calculation, it sends the data that needs to be sent to other logic operation cores. After the transmission is completed, the migration and collision operations of the data between the logic operation cores are performed; If it is a multi-SOC chip architecture, then the migration and collision operations of data between SOCs are performed; the result data is obtained according to the migration and collision operations; Iterative execution module: the result data is used as the processing data for the next moment, and is substituted into step S2 for iterative execution until the result data at the required moment is obtained.

8. The compressible fluid D3Q19 model computing device based on multi-core architecture according to claim 7, characterized in that: The division module includes single SOC unit and multi-SOC unit: Single SOC unit: The physical computing core of a single SOC is divided into logical computing cores of an N×N×N three-dimensional architecture and numbered separately; Multi-SOC unit: First, divide the physical operation cores of multiple SOCs into a three-dimensional logical structure, divide the corresponding logical processing tasks for the physical operation core of each SOC, and then divide the physical operation core of each SOC into logical operation cores of N×N×N three-dimensional architecture, and number them respectively.

9. The compressible fluid D3Q19 model computing device based on multi-core architecture according to claim 7, characterized in that: The division module also includes an area control unit, which divides the memory space into Ping / Pang areas to achieve parallel data transmission and calculation. When the data in the Ping area is calculated, the data in the Pang area is communicated and transmitted. When the data in the Ping area is communicated and transmitted, the data in the Pang area is calculated.

10. The compressible fluid D3Q19 model computing device based on multi-core architecture according to claim 9, characterized in that: The area control unit also includes: when the Ping area data performs collision migration, the Pang area data performs communication and transmission; when the Pang area data performs collision migration, the Ping area data performs communication and transmission.

11. The compressible fluid D3Q19 model computing device based on multi-core architecture according to claim 7, characterized in that: The computing module includes: Input unit: input the density, velocity, pressure, dimensionless temperature and entropy of the compressible fluid at the initial moment in the processing data; The first calculation unit: calculates the mixed recursive regularized equilibrium distribution function of 19 discrete velocity directions of the D3Q19 model and the correction term of the D3Q19 model; The second calculation unit: calculates the mixed recursive regularized second-order Hermite polynomial coefficients of the D3Q19 model, and then calculates the mixed recursive regularized third-order Hermite polynomial coefficients according to the mixed recursive regularized second-order Hermite polynomial coefficients; The third calculation unit: calculates the regularization term based on the mixed recursive regularized second-order Hermite polynomial coefficient and the mixed recursive regularized third-order Hermite polynomial coefficient, and calculates the post-collision distribution function of 19 discrete velocity directions of the D3Q19 model at the next moment according to the regularization term; The fourth calculation unit: calculates the density and velocity at the next moment according to the distribution function of the 19 discrete velocity directions of the D3Q19 model after the collision, and completes the LBM calculation; The fifth calculation unit: using the density and pressure of the compressible fluid at the current moment, calculates the entropy at the current moment; The sixth calculation unit: calculates the entropy and dimensionless temperature at the next moment according to the entropy at the current moment, and completes the FVM calculation; Merging unit: The LBM calculation and FVM calculation results are merged to obtain the density, velocity, pressure, dimensionless temperature, and entropy at the next moment to complete the calculation.

12. The compressible fluid D3Q19 model computing device based on multi-core architecture according to claim 11, characterized in that: In the second calculation unit, if the current moment is the initial moment, the mixed recursive regularized second-order Hermite polynomial coefficients are obtained according to the finite difference second-order Hermite polynomial coefficients; otherwise, the mixed recursive regularized second-order Hermite polynomial coefficients are obtained according to the recursive regularized second-order Hermite polynomial coefficients and the finite difference second-order Hermite polynomial coefficients.

13. A storage medium containing computer executable instructions, characterized in that: The computer executable instructions, when executed by a computer processor, are used to execute the compressible fluid D3Q19 model calculation method based on a multi-core architecture as described in any one of claims 1-6.

14. A computer program product, wherein the computer program product stores at least one program code, characterized in that: The at least one program code is used to be executed by a processor to implement the compressible fluid D3Q19 model calculation method based on a multi-core architecture as described in any one of claims 1-6.

Images