lammps/lib/gpu/lal_charmm_ext.cpp

/***************************************************************************
                             charmm_long_ext.cpp
                             -------------------
                            W. Michael Brown (ORNL)

  Functions for LAMMPS access to charmm/coul/long acceleration routines.

 __________________________________________________________________________
    This file is part of the LAMMPS Accelerator Library (LAMMPS_AL)
 __________________________________________________________________________

    begin                :
    email                : brownw@ornl.gov
 ***************************************************************************/

#include <iostream>
#include <cassert>
#include <cmath>

#include "lal_charmm.h"

using namespace std;
using namespace LAMMPS_AL;

static CHARMM<PRECISION,ACC_PRECISION> CRMMF;

// ---------------------------------------------------------------------------
// Allocate memory on host and device and copy constants to device
// ---------------------------------------------------------------------------
int crm_gpu_init(const int ntypes, double cut_bothsq, double **host_lj1,
                   double **host_lj2, double **host_lj3, double **host_lj4,
                   double *special_lj, const int inum,
                   const int nall, const int max_nbors, const int maxspecial,
                   const double cell_size, int &gpu_mode, FILE *screen,
                   double host_cut_ljsq, double host_cut_coulsq,
                   double *host_special_coul, const double qqrd2e,
                   const double cut_lj_innersq, const double cut_coul_innersq,
                   const double denom_lj, const double denom_coul,
                   double **epsilon, double **sigma,
                   const bool mix_arithmetic) {
  CRMMF.clear();
  gpu_mode=CRMMF.device->gpu_mode();
  double gpu_split=CRMMF.device->particle_split();
  int first_gpu=CRMMF.device->first_device();
  int last_gpu=CRMMF.device->last_device();
  int world_me=CRMMF.device->world_me();
  int gpu_rank=CRMMF.device->gpu_rank();
  int procs_per_gpu=CRMMF.device->procs_per_gpu();

  CRMMF.device->init_message(screen,"lj/charmm/coul/charmm",first_gpu,
                              last_gpu);

  bool message=false;
  if (CRMMF.device->replica_me()==0 && screen)
    message=true;

  if (message) {
    fprintf(screen,"Initializing Device and compiling on process 0...");
    fflush(screen);
  }

  int init_ok=0;
  if (world_me==0)
    CRMMF.init(ntypes, cut_bothsq, host_lj1, host_lj2, host_lj3, host_lj4,
                special_lj, inum, nall, max_nbors, maxspecial, cell_size,
                gpu_split, screen, host_cut_ljsq, host_cut_coulsq,
                host_special_coul, qqrd2e, cut_lj_innersq, cut_coul_innersq,
                denom_lj, denom_coul, epsilon, sigma, mix_arithmetic);

  CRMMF.device->world_barrier();
  if (message)
    fprintf(screen,"Done.\n");

  for (int i=0; i<procs_per_gpu; i++) {
    if (message) {
      if (last_gpu-first_gpu==0)
        fprintf(screen,"Initializing Device %d on core %d...",first_gpu,i);
      else
        fprintf(screen,"Initializing Devices %d-%d on core %d...",first_gpu,
                last_gpu,i);
      fflush(screen);
    }
    if (gpu_rank==i && world_me!=0)
      init_ok=CRMMF.init(ntypes, cut_bothsq, host_lj1, host_lj2, host_lj3,
                          host_lj4, special_lj, inum, nall, max_nbors,
                          maxspecial, cell_size, gpu_split, screen,
                          host_cut_ljsq, host_cut_coulsq, host_special_coul,
                          qqrd2e, cut_lj_innersq, cut_coul_innersq, denom_lj,
                          denom_coul, epsilon, sigma, mix_arithmetic);

    CRMMF.device->gpu_barrier();
    if (message)
      fprintf(screen,"Done.\n");
  }
  if (message)
    fprintf(screen,"\n");

  if (init_ok==0)
    CRMMF.estimate_gpu_overhead();

  return init_ok;
}

void crm_gpu_clear() {
  CRMMF.clear();
}

int** crm_gpu_compute_n(const int ago, const int inum_full, const int nall,
                          double **host_x, int *host_type, double *sublo,
                          double *subhi, tagint *tag, int **nspecial,
                          tagint **special, const bool eflag, const bool vflag,
                          const bool eatom, const bool vatom, int &host_start,
                          int **ilist, int **jnum, const double cpu_time,
                          bool &success, double *host_q, double *boxlo,
                          double *prd) {
  return CRMMF.compute(ago, inum_full, nall, host_x, host_type, sublo,
                        subhi, tag, nspecial, special, eflag, vflag, eatom,
                        vatom, host_start, ilist, jnum, cpu_time, success,
                        host_q, boxlo, prd);
}

void crm_gpu_compute(const int ago, const int inum_full, const int nall,
                       double **host_x, int *host_type, int *ilist, int *numj,
                       int **firstneigh, const bool eflag, const bool vflag,
                       const bool eatom, const bool vatom, int &host_start,
                       const double cpu_time, bool &success, double *host_q,
                       const int nlocal, double *boxlo, double *prd) {
  CRMMF.compute(ago,inum_full,nall,host_x,host_type,ilist,numj,firstneigh,
                 eflag,vflag,eatom,vatom,host_start,cpu_time,success,host_q,
                 nlocal,boxlo,prd);
}

double crm_gpu_bytes() {
  return CRMMF.host_memory_usage();
}