CudaPMEDirectForce.h

#ifndef CUDAPMEDIRECTFORCE_H
#define CUDAPMEDIRECTFORCE_H
#include <cuda.h>
#include <string>
#include "Bonded_struct.h"
#include "CudaNeighborListBuild.h"
#include "CudaDirectForceTypes.h"
#include "CudaEnergyVirial.h"

// If this variable is set, we'll use texture objects.
// Unset for now because of the problem with texture objects on GTX 750
//#define USE_TEXTURE_OBJECTS

//
// Calculates direct non-bonded interactions on GPU
//
// (c) Antti-Pekka Hynninen, 2014, aphynninen@hotmail.com
//
// AT = accumulation type
// CT = calculation type
//

//
// Purely abstract base class. Used to declare the methods that are expected in all derived classes
//
template <typename AT, typename CT>
class CudaPMEDirectForceBase {
public:
  virtual ~CudaPMEDirectForceBase(){}
  virtual void setup(double boxx, double boxy, double boxz, double kappa,
		     double roff, double ron, double e14fac,
		     int vdw_model, int elec_model)=0;
  virtual void get_setup(float& boxx, float& boxy, float& boxz, float& kappa,
			 float& roff, float& ron, float& e14fac,
			 int& vdw_model, int& elec_model)=0;

  virtual void get_box_size(CT &boxx, CT &boxy, CT &boxz)=0;
  virtual void set_box_size(const CT boxx, const CT boxy, const CT boxz)=0;

  virtual void set_calc_vdw(const bool calc_vdw)=0;
  virtual void set_calc_elec(const bool calc_elec)=0;
  virtual bool get_calc_vdw()=0;
  virtual bool get_calc_elec()=0;

  virtual void set_vdwparam(const int nvdwparam, const CT *h_vdwparam)=0;
  virtual void set_vdwparam(const int nvdwparam, const char *filename)=0;
  virtual void set_vdwparam14(const int nvdwparam, const CT *h_vdwparam)=0;
  virtual void set_vdwparam14(const int nvdwparam, const char *filename)=0;

  virtual void set_vdwtype(const int ncoord, const int *h_vdwtype)=0;
  virtual void set_vdwtype(const int ncoord, const char *filename)=0;
  virtual void set_vdwtype(const int ncoord, const int *glo_vdwtype,
			   const int *loc2glo, cudaStream_t stream=0)=0;
  
  virtual void set_14_list(int nin14list, int nex14list,
			   xx14list_t* h_in14list, xx14list_t* h_ex14list,
			   cudaStream_t stream=0)=0;

  virtual void set_14_list(const float4 *xyzq,
			   const float boxx, const float boxy, const float boxz,
			   const int *glo2loc_ind,
			   const int nin14_tbl, const int *in14_tbl, const xx14_t *in14,
			   const int nex14_tbl, const int *ex14_tbl, const xx14_t *ex14,
			   cudaStream_t stream=0)=0;

  virtual void calc_14_force(const float4 *xyzq,
			     const bool calc_energy, const bool calc_virial,
			     const int stride, AT *force, cudaStream_t stream=0)=0;
  
  virtual void calc_force(const float4 *xyzq,
			  const CudaNeighborListBuild<32>& nlist,
			  const bool calc_energy,
			  const bool calc_virial,
			  const int stride, AT *force, cudaStream_t stream=0)=0;

};

//
// Actual class
//
template <typename AT, typename CT>
class CudaPMEDirectForce : public CudaPMEDirectForceBase<AT, CT> {

protected:

  // Energy & Virial
  CudaEnergyVirial &energyVirial;

  // Energy term names
  std::string strVdw;
  std::string strElec;
  std::string strExcl;
  
  // VdW parameters
  int nvdwparam;
  int vdwparam_len;
  CT *vdwparam;
  const bool use_tex_vdwparam;
#ifdef USE_TEXTURE_OBJECTS
  bool vdwParamTexObjActive;
  cudaTextureObject_t vdwParamTexObj;
#endif

  // VdW 1-4 parameters
  int nvdwparam14;
  int vdwparam14_len;
  CT *vdwparam14;
  const bool use_tex_vdwparam14;
#ifdef USE_TEXTURE_OBJECTS
  bool vdwParam14TexObjActive;
  cudaTextureObject_t vdwParam14TexObj;
#endif

  // 1-4 interaction and exclusion lists
  int nin14list;
  int in14list_len;
  xx14list_t* in14list;

  int nex14list;
  int ex14list_len;
  xx14list_t* ex14list;

  // VdW types
  int vdwtype_len;
  int *vdwtype;
  
  // Type of VdW and electrostatic models (see above: NONE, VDW_VSH, VDW_VSW ...)
  int vdw_model;
  int elec_model;

  // These flags are true if the vdw/elec terms are calculated
  // true by default
  bool calc_vdw;
  bool calc_elec;

  // Lookup table for Ewald. Used if elec_model == EWALD_LOOKUP
  CT *ewald_force;
  int n_ewald_force;

  // Host version of setup
  DirectSettings_t *h_setup;

  void setup_ewald_force(CT h);
  void set_elec_model(int elec_model, CT h=0.01);
  void update_setup();

  void setup_vdwparam(const int type, const int nvdwparam, const CT *h_vdwparam);
  void load_vdwparam(const char *filename, const int nvdwparam, CT **h_vdwparam);

public:

  CudaPMEDirectForce(CudaEnergyVirial &energyVirial,
		     const char *nameVdw, const char *nameElec, const char *nameExcl);
  ~CudaPMEDirectForce();

  void setup(double boxx, double boxy, double boxz, double kappa,
	     double roff, double ron, double e14fac,
	     int vdw_model, int elec_model);
  
  void get_setup(float& boxx, float& boxy, float& boxz, float& kappa,
		 float& roff, float& ron, float& e14fac,
		 int& vdw_model, int& elec_model);
    
  void clearTextures();
  
  void get_box_size(CT &boxx, CT &boxy, CT &boxz);
  void set_box_size(const CT boxx, const CT boxy, const CT boxz);

  void set_calc_vdw(const bool calc_vdw);
  void set_calc_elec(const bool calc_elec);
  bool get_calc_vdw();
  bool get_calc_elec();

  void set_vdwparam(const int nvdwparam, const CT *h_vdwparam);
  void set_vdwparam(const int nvdwparam, const char *filename);
  void set_vdwparam14(const int nvdwparam, const CT *h_vdwparam);
  void set_vdwparam14(const int nvdwparam, const char *filename);

  void set_vdwtype(const int ncoord, const int *h_vdwtype);
  void set_vdwtype(const int ncoord, const char *filename);
  void set_vdwtype(const int ncoord, const int *glo_vdwtype,
		   const int *loc2glo, cudaStream_t stream=0);

  void set_14_list(int nin14list, int nex14list,
		   xx14list_t* h_in14list, xx14list_t* h_ex14list,
		   cudaStream_t stream=0);

  void set_14_list(const float4 *xyzq,
		   const float boxx, const float boxy, const float boxz,
		   const int *glo2loc_ind,
		   const int nin14_tbl, const int *in14_tbl, const xx14_t *in14,
		   const int nex14_tbl, const int *ex14_tbl, const xx14_t *ex14,
		   cudaStream_t stream=0);

  void calc_14_force(const float4 *xyzq,
		     const bool calc_energy, const bool calc_virial,
		     const int stride, AT *force, cudaStream_t stream=0);

  void calc_force(const float4 *xyzq,
		  const CudaNeighborListBuild<32>& nlist,
		  const bool calc_energy,
		  const bool calc_virial,
		  const int stride, AT *force, cudaStream_t stream=0);

};

#endif // CUDAPMEDIRECTFORCE_H