Différences

Ci-dessous, les différences entre deux révisions de la page.

--- sunfluidh:homedata_poissonsolver_namelist [2017/09/25 10:51] – yann
+++ sunfluidh:homedata_poissonsolver_namelist [2021/04/12 16:30] (Version actuelle) – yann
@@ Ligne 1: / Ligne 1: @@
 ===== Namelist "HomeData_PoissonSolver" =====
 <WRAP info>
-This namelist contains all parameters on the iterative solvers and the nV-cycle multigrid method directly implemented in the code for solving the Poisson's equation ("homemade" development).
+This namelist contains all parameters on the iterative solvers and the nV-cycle multigrid method directly implemented in the code for solving the Poisson's equation (in-house development). The iterative solvers are based on the Successive Over Relaxation methods (SOR).
 </WRAP>
 <note important>
-   * Thse data are taken into account only when the variable "Poisson_NumericalMethod" in the Namelist "Numerical_Methods" is set to the required data (see ...).
+   * These data are only associated with the SOR-type solvers directly implemented in the code ("homemade" development)
+   * These data are taken into account only when the variable "Poisson_NumericalMethod" in the Namelist "Numerical_Methods" is correctly set (see [[New_Numerical_Methods_Setup_Namelist |Numerical_Methods (new version)]]).
      * Poisson_NumericalMethod= Home-Multigrid-ConstantMatrixCoef
      * Poisson_NumericalMethod= Home-Multigrid-VariableMatrixCoef
    * __** Not for the release SUNFLUIDH_EDU**__
 </note>
@@ Ligne 15: / Ligne 15: @@
 ===== Full data set of the namelist =====
-   &Numerical_Methods  NS_NumericalMethod= "BDF2-SchemeO2",
+   &HomeData_PoissonSolver  SolverName="SOR-Redblack-MPI",
-                       MomentumConvection_Scheme="Centered-O2-Conservative" ,
+                            Number_max_Grid= 5,
-                       TemperatureAdvection_Scheme="Centered-O2-Conservative" ,
+                            Number_max_Cycle= 10,
-                       SpeciesAdvection_Scheme="Centered-O2-Convective" ,
+                            Number_Iteration= 0,
-                       Poisson_NumericalMethod="Hypre-VariableMatrixCoef"  /
+                            Number_Iteration_FineToCoarseGrid= 1,
+                            Number_Iteration_CoarseToFineGrid= 5,
+                            Number_Iteration_CoarsestGrid= 10 ,
+                            Relaxation_Coefficient= 1.7 ,
+                            Convergence_Criterion= 1.D-15,
+                            Prolongation_Type="Uniform",
+                            Coarsening_Id(1:5)= 7,7,7,7,7 ,
+                            Off_Set_Poisson_Source_term= .false. /
 <note important>
    * Following the problem treated, some numerical methods are better adapted than others, this point is precised for each available option.
-   * The partial diagonalization method (Poisson_NumericalMethod = "Home-PartialDiagonalization") used for solving the Poisson equation is only permitted for separable problems.
-   * The HYPRE library solvers for solving the Poisson's equation are only available if the HYPRE library has been installed.
 </note>
 ===== Definition of the data set =====
+==== SolverName ====
+   * Type: Character string
+   * Selection of the iterative solver coupled with the multigrid method.
+       * __Jacobi__           : Successive Over Relaxed solver based on the Jacobi's method
+       * __SOR__              : Successive Over Relaxed solver based on the Gauss-Seidel method
+       * __SOR-Redblack__     : Successive Over Relaxed solver based on a "Red-Black" algorithm (useful with OpenMP-parallel computations or MPI domain decompositions because it can ensure the replicability of the solution whatever the number of MPI Processus and OpenMP threads).
+       * __SOR-Redblack-MPI__ : Successive Over Relaxed solver based on a "Red-Black" algorithm optimized for the domain decomposition approach (MPI-parallel computations). ** DO NOT USE IT IN Hydrid parallelisation (MPI+ OpenMP)**.
+       * __SOR-Redblack-MPIOMP__ : Successive Over Relaxed solver based on a "Red-Black" algorithm optimized for the domain decomposition approach (MPI-parallel computations) and the multi-threading parallelism (OpenMP).
+       * __SOR-RRB__ : Successive Over Relaxed solver based on a "Red-Black" algorithm with optimization based on data reordering. This method can be used for MPI-parallel computations but it is not yet ready for OpenMP and hydrid parallel ones.
------
+==== Number_max_Grid (multigrid method only) ====
-===== Solving the conservation equations =====
+   * Type : Integer value (maximum value 16)
------
+   * Maximum number of grid levels used in the multigrid method.
-====  NS_NumericalMethod  ====
+==== Number_max_Cycle (multigrid method only) ====
-(equivalent to "Numerical_Scheme" in the previous release)
+   * Type : Integer value
-   * Type: character string
+   * Maximum number of V-cycles permitted in the multigrid method.
-   * Selection of the numerical scheme for solving the conservation equations :
+==== Number_Iteration (only if the SOR method is used) ====
-     * __"BDF2-SchemeO2"__ : 2nd order Backward Differentiation Formula for the time discretization. Semi-implicit scheme on the viscous or diffusion terms. 2nd order spatial discretization (centered for the viscous/diffusion terms, different schemes are available for the convection/advection terms - see further). \\ Preferentially used for incompressible or low Mach-number flows without multi-species components (except two phase incompressible flows).
+   * Type : integer value
-     * __"CN-SchemeO2"__ : 2nd order Crank-Nicolson scheme (semi-implicit scheme on the viscous or diffusion terms. 2nd order spatial discretization (centered for the viscous/diffusion terms, different schemes are available for the convection/advection terms - see further).\\ Preferentially used for incompressible or low Mach-number flows without multi-species components (except two phase incompressible flows). __** Not for the release SUNFLUIDH_EDU**__ .
+   * Number of iterations performed by the SOR method in order to approach the solution on each grid level. This value is the same for any grid level of the Multigrid procedure.
-     * __"BDF1-SchemeO2"__ : 1st order Backward Differentiation Formula. Semi-implicit scheme on the viscous or diffusion terms. 2nd order spatial discretization (centered for the viscous/diffusion terms, different schemes are available for the convection/advection terms - see further). __** Not for the release SUNFLUIDH_EDU**__ .
+==== Number_Iteration_FineToCoarseGrid (only if the SOR method is used) ====
-     * __"ExplicitPredCorO2-SchemeO2-Knio"__ : Explicit predictor-corrector scheme : 2nd order Adams Bashforth - Runge-Kutta scheme (Knio et al., JCP 1998). \\ This scheme is mainly used for low Mach Number flows with multi-species components. __** Not for the release SUNFLUIDH_EDU**__ .
+   * Type : integer value
-     * __"ExplicitPredCorO2-SchemeO2-Njam"__ : 2nd order Explicit scheme based on Strang's operator-splitting with a multi-timestep procedure on the viscous/diffusion terms (2nd order Adams Bashforth - Runge-Kutta schemes) The density is explicitly solved from the differential formulation of the equation of state and the Temperature is deduced from the equation of state. The projection method is based on the Chorin's formulation by considering Poisson's Operator with constant coefficients (Njam et al., JCP 1999). \\ This scheme is mainly used for reactive flows under low Mach number hypothesis. __** Not for the release SUNFLUIDH_EDU**__ .
+   * Number of iterations performed by the SOR solver during the restriction step of each V-cycle of the multigrid procedure. For some problems, the minimization of iterations during the restriction step can improve the time-performance of the code.\\ __BEWARE__ : this data is enabled only if the variable Number_Iteration= 0 or is removed from the Namelist.
-     * __"ExplicitPredCorO2-SchemeO2-Rixen"__ : 2nd order Explicit scheme based on Strang's operator-splitting with multistep  a multi-timestep procedure on the viscous/diffusion terms (2nd order Adams Bashforth - Runge-Kutta schemes). The temperature is explicitly solved from the enthalpy equation and the density is deduced from the equation of state. The projection method is based on the Goda's formulation by considering Poisson's Operator with variable coefficients (Rixin Yu et al., JCP 2012). __** Not for the release SUNFLUIDH_EDU**__ .
+==== Number_Iteration_CoarsestGrid (only if the SOR method is used) ====
-     * __"ExplRKO3-CompactSchemeO4"__ : 4th order hermittian scheme in space + 3rd order explicit Runge-Kutta scheme in time. The projection method is based on the Goda's formulation by considering Poisson's Operator with constant coefficients (Knikker, ijnmf 2008, 2009). \\ This numerical scheme is used for incompressible or low Mach number flows with or without heat transfer. __** Not for the release SUNFLUIDH_EDU**__ .
+   * Type : integer value
-     * __"CN-CompactSchemeO4"__ : 4th order hermittian scheme in space + semi-implicit Crank-Nicolson scheme in time. The projection method is based on the Goda's formulation by considering Poisson's Operator with constant coefficients (Knikker, ijnmf 2008, 2009). \\ This numerical scheme is used for incompressible or low Mach number flows with or without heat transfer. __** Not for the release SUNFLUIDH_EDU**__ .
+   * Number of iterations performed by the SOR solver on the coarsest grid of each V-cycle of the multigrid procedure.\\ __BEWARE__ : this data is enabled only if the variable Number_Iteration= 0 or is removed from the Namelist.
-     * __"RK3CN-CompactSchemeO4"__ : 4th order hermittian scheme in space + hybrid RK3/Crank-Nicolson scheme in time. The projection method is based on the Goda's formulation by considering Poisson's Operator with constant coefficients (Knikker, ijnmf 2008, 2009). \\ This numerical scheme is used for incompressible or low Mach number flows with or without heat transfer. __** Not for the release SUNFLUIDH_EDU**__ .
+==== Number_Iteration_CoarseToFineFGrid (only if the SOR method is used) ====
-     * __"CN-SchemeO2-SpecialLowMachFlow"__: Crank-Nicolson semi-implicit scheme (as the option 2) with a predictor-corrector procedure for solving species mass fractions and temperature. This numerical scheme is used for low Mach number flows. __** Not for the release SUNFLUIDH_EDU**__ .
+   * Type : integer value
+   * Number of iterations performed by the SOR solver during the prolongation step of each V-cycle of the multigrid procedure.\\ __BEWARE__ : this data is enabled only if the variable Number_Iteration= 0 or is removed from the Namelist.
-==== MomentumConvection_Scheme ====
+==== Relaxation_Coefficient (only if the SOR method is used) ====
-(equivalent to "Convective_Flux_Discretization_Type" in the previous release)
+   * Type : real value between one and two
+   * over-relaxation coefficient used in the SOR method.
+==== Convergence_Criterion (only if the SOR method is used) ====
+   * Type : real value
+   * The nV-cycle procedure of the multigrid method is stopped if the L2-norm of the Poisson equation's residu is lower than the criterion chosen.
+==== Coarsening_Id(:) ====
+   * Type : array of integer (max. size = 16)
+   * Define how the grid is coarsened during the restriction stage for each grid level "ng". The array size must be equal to the number of grid levels. For the grid level "ng", the coarsening is defined in regard to the previous (finner) grid level "ng-1". The value in Coarsening_Id(ng) "defines the direction(s) in which the coarsening occurs. Values are defined as :
+         * along the I-direction : 1
+         * along the J-direction : 2
+         * along the K-direction : 4
+         * along I and J directions: 3
+         * along I and K directions: 5
+         * along J and K directions: 6
+         * along I,J and K directions: 7
+   * 7 is the default value considered by the code when Coarsening_Id is not explicitely set (coarsening in all directions).
+   * The value of Coarsening_Id(1) is not considered since ng= 1 is related to the main grid, the finnest one, and it does not result from a restriction stage.
+   * For example, in order to define a coarsening process over 5 grid levels by considering the I-direction on the 2nd grid level, the J and K directions on the 3rd level, and all directions for any other levels : Coarsening_Id(1:5)= 7,1,6,7,7 or Coarsening_Id(2:5)= 1,6,7,7, or Coarsening_Id(2:3)= 1,6 are equivalent.
+====  Prolongation_Type ====
    * Type : character string
-   * Selection of the 2nd order spatial discretization for the convection flux in the momentum equations. The options are :
+   * "Uniform " : The prolongation operator is of zero order. By considering two successive grid levels, errors on the coarser grid are directly reported on the finner one.This is the default value when "Prolongation_Type" is not explicitly set.
-     * __"Centered-O2-Conservative"__ : 2nd order centered scheme in the conservative formulation.
+   * "Linear" : The prolongation operator is of first-order. By considering two successive grid levels, errors on the coarsest grid are reported on the finest one by means of a tri-linear interpolation (in 3D). __BEWARE__ when the numerical simulation contains immersed bodies which can induce small inaccuracies.
-     * __"Centered-O2-Convective"__   : 2nd order centered scheme in the advective formulation.
+==== Off_Set_Poisson_Source_term ====
-     * __"Centered-O2-Skewsymmetric"__: 2nd order centered scheme in the skew-symemetric formulation (developer's test).
+   * Type : Boolean value. __** Not for the release SUNFLUIDH_EDU**__.
-     * __"Upwind-O1-Conservative"__   : not commented (developer's test)
+   * The source term is shifted from its averaged value such as its integral value over the domain is zero. This could facilitate the convergence in some particuliar cases. this option must be used with caution. For expert users only.
-     * __"Quick-O2-Conservative"__ : QUICK scheme in the conservative formulation. __** Not for the release SUNFLUIDH_EDU**__ .
-     * __"Upwind-O2-Conservative"__ : 2nd order Upwind scheme in the conservative formulation. __** Not for the release SUNFLUIDH_EDU**__ .
-     * __"Eno-O2-Conservative"__ : 2nd order ENO scheme in the conservative formulation (in progress). __** Not for the release SUNFLUIDH_EDU**__ .
-==== TemperatureAdvection_Scheme ====
-(equivalent to "Temperature_Advective_Flux_Discretization_Type" in the previous release)
-   * Type : character string
-   * Selection of the 2nd order spatial discretization for the advection flux in the temperature/enthalpy equation. The options are the same as previously.
-==== SpeciesAdvection_Scheme ====
-(equivalent to "Species_Advective_Flux_Discretization_Type" in the previous release)
-   * Type : character string.  __** Not for the release SUNFLUIDH_EDU**__ .
-   * Selection of the 2nd order spatial discretization for the advection flux in the species equations. The options are the same as previously.
-==== Explicit_Solving_of_Density ====
-   * Type : integer value. __** Not for the release SUNFLUIDH_EDU**__ .
-   * Selection of numerical schemes based on flux limiters with TVD properties for solving the density. Some of proposed methods are "experimental" and they must be used with caution. The options are :
-     * 0 : The mass equation is not solved. For low Mach number flows, the density is deduced from the equation of state. This supposes that species mass fractions or/and temperature are calculated from their conservation equations if it is required.
-      * 1, 2, 4, 5 : Obsolete methods
-      * 3 : Solving the density from the differential equation of state. This option is automatically selected when the numerical scheme of Njam et al. is used (Numerical_Scheme=5)
-      * 6 : The mass equation is solved with the Lax-Wendroff + Superbee TVD scheme (test).
-      * 7 : The mass equation is solved with the Lax-Wendroff + SuperC TVD scheme (test).
-      * 8 : The mass equation is solved with the Lax-Wendroff + HyperC TVD scheme (test).
-      * 9 : The mass equation is solved with the Lax-Wendroff + Van Leer TVD scheme (test).
-      * 10 : The mass equation is solved with the Lax-Wendroff + a TVD scheme defined by the user (test).
-      * 11 : The mass equation is solved with a WENO5 scheme (test).
------
-===== Solving the Poisson's equation =====
------
-====  Poisson_NumericalMethod ====
-(equivalent to "Numerical_Method_Poisson_Equation " in the previous release)
-   * Type : Character string
-   * Selection of the numerical method for solving the Poisson equation in accordance to the projection method. The solution is the pressure time increment ($\Phi= P^{n+1}-P^{n}$, Goda's method) used to update the velocity field according to the principle of the projection method (When the numerical method of Njam et al. is used to solve the Navier-Stokes equations, the pressure is solved in place of its time increment (Chorin's method). The options are :
-     * __"Home-Multigrid-ConstantMatrixCoef"__ : Successive Over-Relaxed method (SOR) coupled with a multigrid method in order to accelerate the convergence. The matrix elements depend on the cell size only (constant elements).
-     * __"Home-HelmholtzApproximation"__ : Poisson's operator (constant matrix elements) is approximated by a Helmholtz's operator (experimental method proposed by J.L. Guermond) - For incompressible flow only.
-     * __"Home-PartialDiagonalization"__ : Partial diagonalisation of the Laplacian operator (constant elements). BEWARE : the problem must be separable.
-     * __"Home-Multigrid-VariableMatrixCoef"__ : SOR iterative method coupled with a multigrid method in order to accelerate the convergence. The matrix elements depend on the cell size and the density, the source term is defined from the divergence of the velocity.
-     * __"Hypre-ConstantMatrixCoef"__ : Poisson's equation is solved by the HYPRE library tools. The matrix elements are constants. The parameters of the HYPRE library are set in the namelist [[ | ]].
-     * __"Hypre-VariableMatrixCoef"__: Poisson's equation is solved by the HYPRE library tools. The matrix elements are non-constants. The parameters of the HYPRE library are set in the namelist [[ | ]].