libplb-dev 1.5~r1+repack1-3 / usr / include / palabos / latticeBoltzmann / momentTemplates.h

/* This file is part of the Palabos library.
 *
 * Copyright (C) 2011-2015 FlowKit Sarl
 * Route d'Oron 2
 * 1010 Lausanne, Switzerland
 * E-mail contact: contact@flowkit.com
 *
 * The most recent release of Palabos can be downloaded at 
 * <http://www.palabos.org/>
 *
 * The library Palabos is free software: you can redistribute it and/or
 * modify it under the terms of the GNU Affero General Public License as
 * published by the Free Software Foundation, either version 3 of the
 * License, or (at your option) any later version.
 *
 * The library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU Affero General Public License for more details.
 *
 * You should have received a copy of the GNU Affero General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
*/

/** \file
 * Helper functions for the computation of velocity moments for the f's.
 * This file is all about efficiency. The generic template code is specialized
 * for commonly used Lattices, so that a maximum performance can be taken out
 * of each case.
 */
#ifndef MOMENT_TEMPLATES_H
#define MOMENT_TEMPLATES_H

#include "core/globalDefs.h"
#include "core/cell.h"
#include "core/util.h"
#include "latticeBoltzmann/geometricOperationTemplates.h"
#include "latticeBoltzmann/roundOffPolicy.h"

namespace plb {

template<typename T, class Descriptor> struct momentTemplatesImpl;

// This structure forwards the calls to the appropriate helper class
template<typename T, template<typename U> class Descriptor>
struct momentTemplates {

static T get_rhoBar(Cell<T,Descriptor> const& cell) {
    return momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::get_rhoBar(cell.getRawPopulations());
}

static void get_j(Cell<T,Descriptor> const& cell, Array<T,Descriptor<T>::d>& j ) {
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::get_j(cell.getRawPopulations(), j);
}

static T get_eBar(Cell<T,Descriptor> const& cell) {
    return momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::get_eBar(cell.getRawPopulations());
}

static void get_rhoBar_j(Cell<T,Descriptor> const& cell, T& rhoBar, Array<T,Descriptor<T>::d>& j ) {
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::get_rhoBar_j(cell.getRawPopulations(), rhoBar, j);
}

static void get_rhoBar_j_thetaBar(Cell<T,Descriptor> const& cell, T& rhoBar, Array<T,Descriptor<T>::d>& j, T &thetaBar ) {
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
    ::get_rhoBar_j_thetaBar(cell.getRawPopulations(), rhoBar, j, thetaBar);
}

static T compute_rho(Cell<T,Descriptor> const& cell) {
    return momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_rho(cell.getRawPopulations());
}

/// Get order-1 moment of f's, divided by rho ("lattice-boltzmann-velocity", or "uLb")
/** In many cases, such as the plain BGK model, this is equal to to physical
 * velocity, but in other cases not. In presence of a body force g for example,
 * the velocity is uLb + g/2.
 **/
static void compute_uLb(Cell<T,Descriptor> const& cell, Array<T,Descriptor<T>::d>& uLb ) {
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_uLb(cell.getRawPopulations(), uLb);
}

static void compute_rho_uLb(Cell<T,Descriptor> const& cell, T& rho, Array<T,Descriptor<T>::d>& uLb ) {
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_rho_uLb(cell.getRawPopulations(), rho, uLb);
}

static T compute_e(Cell<T,Descriptor> const& cell) {
    return momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_e(cell.getRawPopulations());
}

static T compute_rhoThetaBar(Cell<T,Descriptor> const& cell, T rhoBar, T jSqr) {
    return momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_rhoThetaBar(cell.getRawPopulations(), rhoBar, jSqr);
}

static void compute_rho_rhoThetaBar(Cell<T,Descriptor> const& cell, T& rho, T& rhoThetaBar) {
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_rho_rhoThetaBar(cell.getRawPopulations(), rho, rhoThetaBar);
}

static T compute_theta(Cell<T,Descriptor> const& cell, T rhoBar, T jSqr) {
    return momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_theta(cell.getRawPopulations(), rhoBar, jSqr);
}

static T compute_rhoEpsilon(Cell<T,Descriptor> const& cell, T rhoBar, T jSqr) {
    return momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_rhoEpsilon(cell.getRawPopulations(), rhoBar, jSqr);
}

static void compute_PiNeq(Cell<T,Descriptor> const& cell, T rhoBar, Array<T,Descriptor<T>::d> const& j,
                          Array<T,SymmetricTensor<T,Descriptor>::n>& PiNeq )
{
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_PiNeq(cell.getRawPopulations(), rhoBar, j, PiNeq, Descriptor<T>::invRho(rhoBar));
}

static void compute_PiNeq(Cell<T,Descriptor> const& cell, T rhoBar, Array<T,Descriptor<T>::d> const& j,
                          Array<T,SymmetricTensor<T,Descriptor>::n>& PiNeq, T invRho )
{
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_PiNeq(cell.getRawPopulations(), rhoBar, j, PiNeq, invRho);
}

static void compute_thermal_PiNeq(Cell<T,Descriptor> const& cell, T rhoBar, T thetaBar,
                                  Array<T,Descriptor<T>::d> const& j,
                                  Array<T,SymmetricTensor<T,Descriptor>::n>& PiNeq )
{
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_thermal_PiNeq(cell.getRawPopulations(), rhoBar, thetaBar, j, PiNeq);
}

static void compute_rhoBar_j_PiNeq(Cell<T,Descriptor> const& cell, T& rhoBar, Array<T,Descriptor<T>::d>& j,
                                   Array<T,SymmetricTensor<T,Descriptor>::n>& PiNeq )
{
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_rhoBar_j_PiNeq(cell.getRawPopulations(), rhoBar, j, PiNeq);
}

static void compute_rhoBar_j_PiNeq(Cell<T,Descriptor> const& cell, T& rhoBar, Array<T,Descriptor<T>::d>& j,
                                   Array<T,SymmetricTensor<T,Descriptor>::n>& PiNeq, T invRho )
{
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_rhoBar_j_PiNeq(cell.getRawPopulations(), rhoBar, j, PiNeq, invRho);
}

static void compute_rhoBar_thetaBar_j_PiNeq(Cell<T,Descriptor> const& cell, T& rhoBar, T& thetaBar,
                                            Array<T,Descriptor<T>::d> & j,
                                            Array<T,SymmetricTensor<T,Descriptor>::n>& PiNeq )
{
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_rhoBar_thetaBar_j_PiNeq(cell.getRawPopulations(), rhoBar, thetaBar, j, PiNeq);
}
                                                 
static void compute_rhoBar_thetaBar_j_PiNeq_qNeq(Cell<T,Descriptor> const& cell, T& rhoBar, T& thetaBar,
                                                 Array<T,Descriptor<T>::d> & j,
                                                 Array<T,SymmetricTensor<T,Descriptor>::n>& PiNeq,
                                                 Array<T,SymmetricRankThreeTensor<T,Descriptor>::n>& qNeq)
{
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_rhoBar_thetaBar_j_PiNeq_qNeq(cell.getRawPopulations(), rhoBar, thetaBar, j, PiNeq, qNeq);
}


/// Get local, order-2 moment: sum_i (c_i-uLb)(c_i-uLb) f_i = -rho uLb uLb + sum_i c_i c_i f_i
/** The full stress tensor Pi is equal to P + rho u u. The deviatoric stress tensor sigma
 *  is equal to P - c_s^2 rho I (\sa compute_Pi_neq)
 **/
static void compute_P(Cell<T,Descriptor> const& cell, T rhoBar, Array<T,Descriptor<T>::d> const& j,
                      Array<T,SymmetricTensor<T,Descriptor>::n>& P)
{
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_P(cell.getRawPopulations(), rhoBar, j, P);
}

static void modifyJ(T& cell, Array<T,Descriptor<T>::d> const& newJ) {
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::modifyVelocity(cell.getRawPopulations(), newJ);
}

static void compute_Qneq(Cell<T,Descriptor> const& cell, T rhoBar, Array<T,Descriptor<T>::d> const& j,
                         T thetaBar,
                         Array<T,SymmetricRankThreeTensor<T,Descriptor>::n>& qNeq )
{
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_Qneq(cell.getRawPopulations(), rhoBar, j, thetaBar, qNeq);
}

static void compute_heat_flux(Cell<T,Descriptor> const& cell, T rhoBar, Array<T,Descriptor<T>::d> const& j,
                              T thetaBar,
                              Array<T,Descriptor<T>::d>& q )
{
    momentTemplatesImpl<T,typename Descriptor<T>::BaseDescriptor>
        ::compute_heat_flux(cell.getRawPopulations(), rhoBar, j, thetaBar, q);
}

};  // struct momentTemplates


// This structure forwards the calls to the appropriate helper class
template<typename T, class Descriptor>
struct momentTemplatesImpl {

static T get_rhoBar(Array<T,Descriptor::q> const& f) {
    T rhoBar = f[0];
    for (plint iPop=1; iPop < Descriptor::q; ++iPop) {
        rhoBar += f[iPop];
    }
    return rhoBar;
}

static void get_j(Array<T,Descriptor::q> const& f, Array<T,Descriptor::d>& j ) {
    for (int iD=0; iD < Descriptor::d; ++iD) {
        j[iD] = f[0]*Descriptor::c[0][iD];
    }
    for (plint iPop=1; iPop < Descriptor::q; ++iPop) {
        for (int iD=0; iD < Descriptor::d; ++iD) {
            j[iD] += f[iPop]*Descriptor::c[iPop][iD];
        }
    }
}

static T get_eBar(Array<T,Descriptor::q> const& f) {
    T eBar = f[0] * Descriptor::cNormSqr[0];
    for (plint iPop=1; iPop < Descriptor::q; ++iPop) {
        eBar += f[iPop] * Descriptor::cNormSqr[iPop];
    }
    return eBar;
}

static void get_rhoBar_j(Array<T,Descriptor::q> const& f, T& rhoBar, Array<T,Descriptor::d>& j ) {
    rhoBar = get_rhoBar(f);
    get_j(f, j);
}

static void get_rhoBar_j_thetaBar(Array<T,Descriptor::q> const& f, T& rhoBar, Array<T,Descriptor::d>& j, T &thetaBar ) {
    get_rhoBar_j(f,rhoBar,j);
    T jSqr = VectorTemplateImpl<T,Descriptor::d>::normSqr(j);
    T invRho = Descriptor::invRho(rhoBar);
    thetaBar = compute_rhoThetaBar(f, rhoBar, jSqr) * invRho;
}


static T compute_rho(Array<T,Descriptor::q> const& f) {
    return Descriptor::fullRho(get_rhoBar(f));
}

static void compute_uLb(Array<T,Descriptor::q> const& f, Array<T,Descriptor::d>& uLb ) {
    get_j(f, uLb);
    T invRho = Descriptor::invRho(get_rhoBar(f));
    for (int iD=0; iD < Descriptor::d; ++iD) {
        uLb[iD] *= invRho;
    }
}

static void compute_rho_uLb(Array<T,Descriptor::q> const& f, T& rho, Array<T,Descriptor::d>& uLb ) {
    get_j(f, uLb);
    T rhoBar = get_rhoBar(f);
    T invRho = Descriptor::invRho(rhoBar);
    rho = Descriptor::fullRho(rhoBar);
    for (int iD=0; iD < Descriptor::d; ++iD) {
        uLb[iD] *= invRho;
    }
}

static T compute_e(Array<T,Descriptor::q> const& f) {
    return get_eBar(f) + Descriptor::SkordosFactor() * Descriptor::d * Descriptor::cs2;
}

static T compute_rhoThetaBar(Array<T,Descriptor::q> const& f, T rhoBar, T jSqr) {
    T invRho = Descriptor::invRho(rhoBar);
    return Descriptor::invCs2 * Descriptor::invD * (get_eBar(f) - invRho*jSqr) - rhoBar;
}

static void compute_rho_rhoThetaBar(Array<T,Descriptor::q> const& f, T& rho, T& rhoThetaBar) {
    T rhoBar, j[Descriptor::d];
    get_rhoBar_j(f, rhoBar, j);
    T jSqr = VectorTemplateImpl<T,Descriptor::d>::normSqr(j);
    rho = Descriptor::fullRho(rhoBar);
    rhoThetaBar = compute_rhoThetaBar(f, rhoBar, jSqr);
}

static T compute_theta(Array<T,Descriptor::q> const& f, T rhoBar, T jSqr) {
    T invRho = Descriptor::invRho(rhoBar);
    T e = compute_e(f);
    return invRho * Descriptor::invD * Descriptor::invCs2 * (e - invRho*jSqr);
}

static T compute_rhoEpsilon(Array<T,Descriptor::q> const& f, T rhoBar, T jSqr) {
    T invRho = Descriptor::invRho(rhoBar);
    T e = compute_e(f);
    return (e - invRho*jSqr) / (T)2;
}

static void compute_PiNeq(Array<T,Descriptor::q> const& f, T rhoBar, Array<T,Descriptor::d> const& j,
                          Array<T,SymmetricTensorImpl<T,Descriptor::d>::n>& PiNeq )
{
   T invRho = Descriptor::invRho(rhoBar);
   compute_PiNeq(f, rhoBar, j, PiNeq, invRho);
}

static void compute_PiNeq(Array<T,Descriptor::q> const& f, T rhoBar, Array<T,Descriptor::d> const& j,
                          Array<T,SymmetricTensorImpl<T,Descriptor::d>::n>& PiNeq, T invRho )
{
    int iPi = 0;
    for (int iAlpha=0; iAlpha < Descriptor::d; ++iAlpha) {
        int iDiagonal = iPi;
        for (int iBeta=iAlpha; iBeta < Descriptor::d; ++iBeta) {
            PiNeq[iPi] = Descriptor::c[0][iAlpha]*
                         Descriptor::c[0][iBeta] * f[0];
            for (plint iPop=1; iPop < Descriptor::q; ++iPop) {
                PiNeq[iPi] += Descriptor::c[iPop][iAlpha]*
                              Descriptor::c[iPop][iBeta] * f[iPop];
            }
            // Stripe off relative velocity
            PiNeq[iPi] -= invRho*j[iAlpha]*j[iBeta];
            ++iPi;
        }
        // Stripe off diagonal term
        PiNeq[iDiagonal] -= Descriptor::cs2 * rhoBar;
    }
}

static void compute_PiNeq(Array<T,Descriptor::q> const& fNeq, Array<T,SymmetricTensorImpl<T,Descriptor::d>::n>& PiNeq)
{
    int iPi = 0;
    for (int iAlpha=0; iAlpha < Descriptor::d; ++iAlpha) {
        for (int iBeta=iAlpha; iBeta < Descriptor::d; ++iBeta) {
            PiNeq[iPi] = Descriptor::c[0][iAlpha]*
                         Descriptor::c[0][iBeta] * fNeq[0];
            for (plint iPop=1; iPop < Descriptor::q; ++iPop) {
                PiNeq[iPi] += Descriptor::c[iPop][iAlpha]*
                              Descriptor::c[iPop][iBeta] * fNeq[iPop];
            }
            ++iPi;
        }
    }
}

static void compute_thermal_PiNeq(Array<T,Descriptor::q> const& f, T rhoBar, T thetaBar,
                                  Array<T,Descriptor::d> const& j,
                                  Array<T,SymmetricTensorImpl<T,Descriptor::d>::n>& PiNeq )
{
    // rhoTheta_bar == rho*theta - 1 
    T rhoTheta_bar = rhoBar * thetaBar + rhoBar + Descriptor::SkordosFactor() * thetaBar;
    
    T invRho = Descriptor::invRho(rhoBar);
    int iPi = 0;
    for (int iAlpha=0; iAlpha < Descriptor::d; ++iAlpha) {
        int iDiagonal = iPi;
        for (int iBeta=iAlpha; iBeta < Descriptor::d; ++iBeta) {
            PiNeq[iPi] = Descriptor::c[0][iAlpha]*
                         Descriptor::c[0][iBeta] * f[0];
            for (plint iPop=1; iPop < Descriptor::q; ++iPop) {
                PiNeq[iPi] += Descriptor::c[iPop][iAlpha]*
                              Descriptor::c[iPop][iBeta] * f[iPop];
            }
            // Stripe off relative velocity
            PiNeq[iPi] -= invRho*j[iAlpha]*j[iBeta];
            ++iPi;
        }
        // Stripe off diagonal term
        PiNeq[iDiagonal] -= Descriptor::cs2 * rhoTheta_bar;
    }
}

static void compute_rhoBar_j_PiNeq(Array<T,Descriptor::q> const& f, T& rhoBar, Array<T,Descriptor::d>& j,
                                   Array<T,SymmetricTensorImpl<T,Descriptor::d>::n>& PiNeq )
{
    get_rhoBar_j(f, rhoBar, j);
    compute_PiNeq(f, rhoBar, j, PiNeq);
}

static void compute_rhoBar_j_PiNeq(Array<T,Descriptor::q> const& f, T& rhoBar, Array<T,Descriptor::d>& j,
                                   Array<T,SymmetricTensorImpl<T,Descriptor::d>::n>& PiNeq, T invRho )
{
    get_rhoBar_j(f, rhoBar, j);
    compute_PiNeq(f, rhoBar, j, PiNeq, invRho);
}

static void compute_rhoBar_thetaBar_j_PiNeq(Array<T,Descriptor::q> const& f, T& rhoBar, T& thetaBar,
                                            Array<T,Descriptor::d> & j,
                                            Array<T,SymmetricTensorImpl<T,Descriptor::d>::n>& PiNeq )
{
    get_rhoBar_j_thetaBar(f,rhoBar,j, thetaBar);
    compute_thermal_PiNeq(f, rhoBar, thetaBar, j, PiNeq);
}

static void compute_rhoBar_thetaBar_j_PiNeq_qNeq(Array<T,Descriptor::q> const& f, T& rhoBar, T& thetaBar,
                                                 Array<T,Descriptor::d> & j,
                                                 Array<T,SymmetricTensorImpl<T,Descriptor::d>::n>& PiNeq,
                                                 Array<T,SymmetricRankThreeTensorImpl<T,Descriptor::d>::n>& qNeq)
{
    compute_rhoBar_thetaBar_j_PiNeq(f, rhoBar, thetaBar, j, PiNeq);
    compute_Qneq(f, rhoBar, j, thetaBar, qNeq );
}


static void compute_P(Array<T,Descriptor::q> const& f, T rhoBar, Array<T,Descriptor::d> const& j,
                      Array<T,SymmetricTensorImpl<T,Descriptor::d>::n>& P )
{
    T invRho = Descriptor::invRho(rhoBar);
    plint iP = 0;
    for (int iAlpha=0; iAlpha < Descriptor::d; ++iAlpha) {
        for (int iBeta=iAlpha; iBeta < Descriptor::d; ++iBeta) {
            P[iP] = Descriptor::c[0][iAlpha]*
                    Descriptor::c[0][iBeta] * f[0];
            for (plint iPop=1; iPop < Descriptor::q; ++iPop) {
                P[iP] += Descriptor::c[iPop][iAlpha]*
                         Descriptor::c[iPop][iBeta] * f[iPop];
            }
            // Stripe off relative velocity
            P[iP] -= invRho*j[iAlpha]*j[iBeta];
            ++iP;
        }
    }
}

static void modifyJ(T* f, Array<T,Descriptor::d> const& newJ) {
    T rhoBar;
    Array<T,Descriptor::d> oldJ;
    get_rhoBar_j(f, rhoBar, oldJ);
    T invRho = Descriptor::invRho(rhoBar);
    const T oldJSqr = VectorTemplateImpl<T,Descriptor::d>::normSqr(oldJ);
    const T newJSqr = VectorTemplateImpl<T,Descriptor::d>::normSqr(newJ);
    for (plint iPop=0; iPop<Descriptor::q; ++iPop) {
        f[iPop] = f[iPop]
                         - bgk_ma2_equilibrium(iPop, rhoBar, invRho, oldJ, oldJSqr)
                         + bgk_ma2_equilibrium(iPop, rhoBar, invRho, newJ, newJSqr);
    }
}

static void compute_Qneq(Array<T,Descriptor::q> const& f, T rhoBar, Array<T,Descriptor::d> const& j,
                         T thetaBar,
                         Array<T,SymmetricRankThreeTensorImpl<T,Descriptor::d>::n>& qNeq )
{
    typedef Descriptor L;
    T invRho = L::invRho(rhoBar);
    T factor = L::cs2+thetaBar;
    plint iQ = 0;
    for (plint iA = 0; iA < L::d; ++iA) {
        for (plint iB = iA; iB < L::d; ++iB) {
            for (plint iC = iB; iC < L::d; ++iC) {
                qNeq[iQ] = - j[iA]*j[iB]*j[iC]*invRho*invRho;
                for (plint iPop = 0; iPop < L::q; ++iPop) {
                    qNeq[iQ] = L::c[iPop][iA]*L::c[iPop][iB]*L::c[iPop][iC]*f[iPop];
                }
                if (iA == iB && iB == iC) {
                    qNeq[iQ] -= (T)3 * factor * j[iA];
                }
                else if (iA == iB && iB != iC) {
                    qNeq[iQ] -= factor * j[iC];
                }
                else if (iA == iC && iC != iB) {
                    qNeq[iQ] -= factor * j[iB];
                }
                else if (iB == iC && iC != iA) {
                    qNeq[iQ] -= factor * j[iA];
                }
                
                ++iQ;
            }
        }
    }
    
}

static void compute_heat_flux(Array<T,Descriptor::q> const& f, T rhoBar, Array<T,Descriptor::d> const& j,
                              T thetaBar,
                              Array<T,Descriptor::d>& q)
{
    
    typedef Descriptor L;
    Array<T,SymmetricRankThreeTensorImpl<T,Descriptor::d>::n> qNeq;
    compute_Qneq(f, rhoBar, j,thetaBar,qNeq );
    SymmetricRankThreeTensorImpl<T,L::d>::contractLastTwoIndexes(qNeq,q);
}

};  // struct momentTemplatesImpl

}  // namespace plb

#include "latticeBoltzmann/momentTemplates2D.h"
#include "latticeBoltzmann/momentTemplates3D.h"

#endif  // MOMENT_TEMPLATES_H