CudaNeighborListBuild.h

#ifndef CUDANEIGHBORLISTBUILD_H
#define CUDANEIGHBORLISTBUILD_H
//
// Neighbor list build class
//
// (c) Antti-Pekka Hynninen 2014
// aphynninen@hotmail.com
//
#include <iostream>
#include <vector>
#include <cuda.h>
#include "CudaNeighborListStruct.h"

template <int tilesize>
struct num_excl {
  static const int val = ((tilesize*tilesize-1)/32 + 1);
};

template <int tilesize>
struct tile_excl_t {
  unsigned int excl[num_excl<tilesize>::val]; // Exclusion mask
};

struct ientry_t {
  int iatomStart;
  int ish;
  int tileStart;
  int tileEnd;
};

#ifdef USE_SPARSE
template <int tilesize>
struct pairs_t {
  int i[tilesize];
};
#endif

template <int tilesize>
class CudaNeighborListBuild {

private:
  // Disable copy constructor
  CudaNeighborListBuild(const CudaNeighborListBuild&);

  // zone range for this sort
  int izoneStart;
  int izoneEnd;

  // Number of i tiles
  int n_ientry;

  // Total number of tiles
  int n_tile;

  // Estimates to number of tiles
  int n_ientry_est;
  int n_tile_est;

  int tile_excl_len;
  tile_excl_t<tilesize> *tile_excl;

  // Raw un-sorted ientry
  int ientry_raw_len;
  ientry_t *ientry_raw;

  // Final sorted ientry
  int ientry_len;
  ientry_t *ientry;

  int tile_indj_len;
  int *tile_indj;

#ifdef USE_SPARSE
  // Sparse:
  int n_ientry_sparse;
  int n_tile_sparse;

  int pairs_len;
  pairs_t<tilesize> *pairs;
  
  int ientry_sparse_len;
  ientry_t *ientry_sparse;

  int tile_indj_sparse_len;
  int *tile_indj_sparse;
#endif

  // Approximate upper bound for number of cells
  //int ncell_max;

  // Maximum value of n_int_zone[]
  int n_int_zone_max;

  // Atom-Atom exclusion heap
  int exclAtomHeapLen;
  int* exclAtomHeap;

  // ------------------------------------
  // Used for sorting ientry:
  // bucketPos[n_jlist_max+1]
  int bucketPosLen;
  int* bucketPos;

  // bucketIndex[n_ientry]
  int bucketIndexLen;
  int* bucketIndex;
  // ------------------------------------

  // Flag for testing neighborlist build
  bool test;

  void calc_tile_ientry_est(const ZoneParam_t* h_ZoneParam, const float rcut);

  template<typename T>
  int calc_gpu_pairlist(const int n_ientry, const ientry_t* ientry, const int* tile_indj,
			const tile_excl_t<tilesize>* tile_excl,
			const float4* xyzq, const double boxx, const double boxy,
			const double boxz, const double rcut);

  template<typename T>
  int calc_cpu_pairlist(const int* zone_patom, const float4* xyzq,
			const int* loc2glo, const int* atom_excl_pos,
			const int* atom_excl, const double boxx,
			const double boxy, const double boxz,
			const double rcut);
  
  //void set_NlistParam(cudaStream_t stream);
  //void get_NlistParam();

  void init();
  void load(const char *filename);

public:
  CudaNeighborListBuild(const int n_int_zone_max,
			const int izoneStart, const int izoneEnd);
  CudaNeighborListBuild(const int n_int_zone_max,
			const int izoneStart, const int izoneEnd,
			const char *filename);
  ~CudaNeighborListBuild();

  void build(const int ncell, const int cellStart, const int maxNumExcl,
	     const ZoneParam_t* h_ZoneParam, const ZoneParam_t* d_ZoneParam,
	     const float boxx, const float boxy, const float boxz,
	     const float rcut,
	     const float4* xyzq, const int* loc2glo, const int* glo2loc,
	     const int* atomExclPos, const int* atomExcl,
	     const int4* cell_xyz_zone, const int* col_ncellz,
	     const int* col_cell, const float* cell_bz,
	     const int* cell_patom, const bb_t* bb,
	     NlistParam_t* h_NlistParam, NlistParam_t* d_NlistParam,
	     cudaStream_t stream);

  void test_build(const int* zone_patom, const int ncol_tot,
		  const int ncell_tot, const ZoneParam_t *h_ZoneParam,
		  const int atomExclPosLen, const int* atomExclPos,
		  const int atomExclLen, const int* atomExcl,
		  const double boxx, const double boxy, const double boxz,
		  const double rcut, const float4 *xyzq,
		  const int* loc2glo, const int* glo2loc,
		  const int ncoord_glo,
		  const int* cell_patom, const int* col_cell,
		  const float* cell_bz, const bb_t* bb);
  //void reset();
  
  void build_excl(const float boxx, const float boxy, const float boxz,
		  const float rcut,
		  const int n_ijlist, const int3 *ijlist,
		  const int *cell_patom,
		  const float4 *xyzq,
		  cudaStream_t stream=0);
  
  void add_tile_top(const int ntile_top, const int *tile_ind_top,
		    const tile_excl_t<tilesize> *tile_excl_top,
		    cudaStream_t stream=0);

  void set_ientry(int n_ientry, ientry_t *h_ientry, cudaStream_t stream);

#ifdef SPARSE
  void split_dense_sparse(int npair_cutoff);
#endif
  void remove_empty_tiles();
  void analyze();

  void set_test(bool test_in) {this->test = test_in;}

  int get_n_ientry() {return n_ientry;}
  int get_n_ientry_est() {return n_ientry_est;}
  ientry_t* get_ientry() {return ientry;}
  int get_n_tile() {return n_tile;}
  int get_n_tile_est() {return n_tile_est;}
  int* get_tile_indj() {return tile_indj;}
  tile_excl_t<tilesize>* get_tile_excl() {return tile_excl;}

  int get_n_ientry() const {return n_ientry;}
  int get_n_ientry_est() const {return n_ientry_est;}
  const ientry_t* get_ientry() const {return ientry;}
  int get_n_tile() const {return n_tile;}
  int get_n_tile_est() const {return n_tile_est;}
  const int* get_tile_indj() const {return tile_indj;}
  const tile_excl_t<tilesize>* get_tile_excl() const {return tile_excl;}

};

#endif // CUDANEIGHBORLISTBUILD_H