CudaTetrahedronFEMForceField.h

Go to the documentation of this file.

/******************************************************************************
*       SOFA, Simulation Open-Framework Architecture, version 1.0 beta 3      *
*                (c) 2006-2008 MGH, INRIA, USTL, UJF, CNRS                    *
*                                                                             *
* This library is free software; you can redistribute it and/or modify it     *
* under the terms of the GNU Lesser General Public License as published by    *
* the Free Software Foundation; either version 2.1 of the License, or (at     *
* your option) any later version.                                             *
*                                                                             *
* This 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 Lesser General Public License *
* for more details.                                                           *
*                                                                             *
* You should have received a copy of the GNU Lesser General Public License    *
* along with this library; if not, write to the Free Software Foundation,     *
* Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301 USA.          *
*******************************************************************************
*                               SOFA :: Modules                               *
*                                                                             *
* Authors: The SOFA Team and external contributors (see Authors.txt)          *
*                                                                             *
* Contact information: contact@sofa-framework.org                             *
******************************************************************************/
#ifndef SOFA_GPU_CUDA_CUDATETRAHEDRONFEMFORCEFIELD_H
#define SOFA_GPU_CUDA_CUDATETRAHEDRONFEMFORCEFIELD_H

#include "CudaTypes.h"
#include <sofa/component/forcefield/TetrahedronFEMForceField.h>


namespace sofa
{

namespace gpu
{

namespace cuda
{

template<class DataTypes>
class CudaKernelsTetrahedronFEMForceField;

} // namespace cuda

} // namespace gpu


namespace component
{

namespace forcefield
{

using namespace sofa::defaulttype;

template <class TCoord, class TDeriv, class TReal>
class TetrahedronFEMForceFieldInternalData< gpu::cuda::CudaVectorTypes<TCoord,TDeriv,TReal> >
{
public:
    typedef gpu::cuda::CudaVectorTypes<TCoord,TDeriv,TReal> DataTypes;
    typedef TetrahedronFEMForceField<DataTypes> Main;
    typedef TetrahedronFEMForceFieldInternalData<DataTypes> Data;
    typedef typename DataTypes::VecCoord VecCoord;
    typedef typename DataTypes::VecDeriv VecDeriv;
    typedef typename DataTypes::Coord Coord;
    typedef typename DataTypes::Deriv Deriv;
    typedef typename DataTypes::Real Real;

    typedef typename Main::Element Element;
    typedef typename Main::VecElement VecElement;
    typedef typename Main::Index Index;
    typedef Mat<6, 6, Real> MaterialStiffness;
    typedef Mat<12, 6, Real> StrainDisplacement;

    typedef gpu::cuda::CudaKernelsTetrahedronFEMForceField<DataTypes> Kernels;

    struct GPUElement
    {
        //Vec<4,int> tetra;
        int ia[BSIZE],ib[BSIZE],ic[BSIZE],id[BSIZE];

        //Mat<6,6,Real> K;
        Real gamma_bx2[BSIZE], mu2_bx2[BSIZE];

        //Vec3f initpos[4];
        Real bx[BSIZE],cx[BSIZE];
        Real cy[BSIZE],dx[BSIZE],dy[BSIZE],dz[BSIZE];

        //Mat<12,6,Real> J;
        Real Jbx_bx[BSIZE],Jby_bx[BSIZE],Jbz_bx[BSIZE];
    };

    gpu::cuda::CudaVector<GPUElement> elems;

    struct GPUElementState
    {
        Mat<3,3,Real> Rt;
        Vec<6,Real> S;
        Real dummy;
    };

    struct GPUElementForce
    {
        Vec<4,Real> fA,fB,fC,fD;
    };

    gpu::cuda::CudaVector<GPUElementState> state;
    gpu::cuda::CudaVector<GPUElementForce> eforce;
    int nbElement; 
    int vertex0; 
    int nbVertex; 
    int nbElementPerVertex; 


    gpu::cuda::CudaVector<int> velems;
    TetrahedronFEMForceFieldInternalData() : nbElement(0), vertex0(0), nbVertex(0), nbElementPerVertex(0) {}
    void init(int nbe, int v0, int nbv, int nbelemperv)
    {
        elems.clear();
        state.clear();
        eforce.clear();
        velems.clear();
        nbElement = nbe;
        elems.resize((nbe+BSIZE-1)/BSIZE);
        state.resize(nbe);
        eforce.resize(nbe);
        vertex0 = v0;
        nbVertex = nbv;
        nbElementPerVertex = nbelemperv;
        int nbloc = (nbVertex+BSIZE-1)/BSIZE;
        velems.resize(nbloc*nbElementPerVertex*BSIZE);
        for (unsigned int i=0;i<velems.size();i++)
            velems[i] = 0;
    }
    int size() const { return nbElement; }
    void setV(int vertex, int num, int index)
    {
        vertex -= vertex0;
        int bloc = vertex/BSIZE;
        int b_x  = vertex%BSIZE;
        velems[ bloc*BSIZE*nbElementPerVertex // start of the bloc
              + num*BSIZE                     // offset to the element
              + b_x                           // offset to the vertex
              ] = index+1;
    }

    void setE(int i, const Element& indices, const Coord& /*a*/, const Coord& b, const Coord& c, const Coord& d, const MaterialStiffness& K, const StrainDisplacement& /*J*/)
    {
        /*std::cout << "CPU Info:\n a = "<<a<<"\n b = "<<b<<"\n c = "<<c<<"\n d = "<<d<<"\n K = "
            <<K[0]<<"\n     "<<K[1]<<"\n     "<<K[2]<<"\n     "
            <<K[3]<<"\n     "<<K[4]<<"\n     "<<K[5]<<"\n J="
            <<J[0]<<"\n     "<<J[1]<<"\n     "<<J[2]<<"\n     "
            <<J[3]<<"\n     "<<J[4]<<"\n     "<<J[5]<<"\n     "
            <<J[6]<<"\n     "<<J[7]<<"\n     "<<J[8]<<"\n     "
            <<J[9]<<"\n     "<<J[10]<<"\n     "<<J[11]<<std::endl;*/
        GPUElement& e = elems[i/BSIZE]; i = i%BSIZE;
        e.ia[i] = indices[0] - vertex0;
        e.ib[i] = indices[1] - vertex0;
        e.ic[i] = indices[2] - vertex0;
        e.id[i] = indices[3] - vertex0;
        e.bx[i] = b[0];
        e.cx[i] = c[0]; e.cy[i] = c[1];
        e.dx[i] = d[0]; e.dy[i] = d[1]; e.dz[i] = d[2];
        Real bx2 = e.bx[i] * e.bx[i];
        e.gamma_bx2[i] = K[0][1] * bx2;
        e.mu2_bx2[i] = 2*K[3][3] * bx2;
        e.Jbx_bx[i] = (e.cy[i] * e.dz[i]) / e.bx[i];
        e.Jby_bx[i] = (-e.cx[i] * e.dz[i]) / e.bx[i];
        e.Jbz_bx[i] = (e.cx[i]*e.dy[i] - e.cy[i]*e.dx[i]) / e.bx[i];
        //e.dummy[i] = 0;
        /*std::cout << "GPU Info:\n b = "<<e.bx<<"\n c = "<<e.cx<<" "<<e.cy<<"\n d = "<<e.dx<<" "<<e.dy<<" "<<e.dz<<"\n K = "
            <<(e.gamma_bx2+e.mu2_bx2)/bx2<<" "<<(e.gamma_bx2)/bx2<<" "<<(e.gamma_bx2)/bx2<<" 0 0 0\n     "
            <<(e.gamma_bx2)/bx2<<" "<<(e.gamma_bx2+e.mu2_bx2)/bx2<<" "<<(e.gamma_bx2)/bx2<<" 0 0 0\n     "
            <<(e.gamma_bx2)/bx2<<" "<<(e.gamma_bx2)/bx2<<" "<<(e.gamma_bx2+e.mu2_bx2)/bx2<<" 0 0 0\n     "

            <<"0 0 0 "<<(e.mu2_bx2/2)/bx2<<" 0 0\n     "
            <<"0 0 0 0 "<<(e.mu2_bx2/2)/bx2<<" 0\n     "
            <<"0 0 0 0 0 "<<(e.mu2_bx2/2)/bx2<<"\n J = "

            <<(-e.Jbx_bx)*e.bx<<" 0 0 "<<(-e.Jby_bx-e.dz)*e.bx<<" 0 "<<(-e.Jbz_bx+e.dy-e.cy)*e.bx<<"\n     "
            <<"0 "<<(-e.Jby_bx-e.dz)*e.bx<<" 0 "<<(-e.Jbx_bx)*e.bx<<" "<<(-e.Jbz_bx+e.dy-e.cy)*e.bx<<" 0\n     "
            <<"0 0 "<<(-e.Jbz_bx+e.dy-e.cy)*e.bx<<" 0 "<<(-e.Jby_bx-e.dz)*e.bx<<" "<<(-e.Jbx_bx)*e.bx<<"\n     "

            <<(e.Jbx_bx)*e.bx<<" 0 0 "<<(e.Jby_bx)*e.bx<<" 0 "<<(e.Jbz_bx)*e.bx<<"\n     "
            <<"0 "<<(e.Jby_bx)*e.bx<<" 0 "<<(e.Jbx_bx)*e.bx<<" "<<(e.Jbz_bx)*e.bx<<" 0\n     "
            <<"0 0 "<<(e.Jbz_bx)*e.bx<<" 0 "<<(e.Jby_bx)*e.bx<<" "<<(e.Jbx_bx)*e.bx<<"\n     "
            
            <<(0)*e.bx<<" 0 0 "<<(e.dz)*e.bx<<" 0 "<<(-e.dy)*e.bx<<"\n     "
            <<"0 "<<(e.dz)*e.bx<<" 0 "<<(0)*e.bx<<" "<<(-e.dy)*e.bx<<" 0\n     "
            <<"0 0 "<<(-e.dy)*e.bx<<" 0 "<<(e.dz)*e.bx<<" "<<(0)*e.bx<<"\n     "

            <<(0)*e.bx<<" 0 0 "<<(0)*e.bx<<" 0 "<<(e.cy)*e.bx<<"\n     "
            <<"0 "<<(0)*e.bx<<" 0 "<<(0)*e.bx<<" "<<(e.cy)*e.bx<<" 0\n     "
            <<"0 0 "<<(e.cy)*e.bx<<" 0 "<<(0)*e.bx<<" "<<(0)*e.bx<<std::endl;*/
    }

    static void reinit(Main* m);
    static void addForce(Main* m, VecDeriv& f, const VecCoord& x, const VecDeriv& /*v*/);
    static void addDForce (Main* m, VecDeriv& df, const VecDeriv& dx, double kFactor, double bFactor);
};

//
// TetrahedronFEMForceField
//

// I know using macros is bad design but this is the only way not to repeat the code for all CUDA types
#define CudaTetrahedronFEMForceField_DeclMethods(T) \
    template<> void TetrahedronFEMForceField< T >::reinit(); \
    template<> void TetrahedronFEMForceField< T >::addForce(VecDeriv& f, const VecCoord& x, const VecDeriv& v); \
    template<> void TetrahedronFEMForceField< T >::addDForce(VecDeriv& df, const VecDeriv& dx, double kFactor, double bFactor);

CudaTetrahedronFEMForceField_DeclMethods(gpu::cuda::CudaVec3fTypes);
CudaTetrahedronFEMForceField_DeclMethods(gpu::cuda::CudaVec3f1Types);


#undef CudaTetrahedronFEMForceField_DeclMethods

} // namespace forcefield

} // namespace component

} // namespace sofa

#endif