262 lines
9.0 KiB
C++
262 lines
9.0 KiB
C++
// clang-format off
|
|
/* ----------------------------------------------------------------------
|
|
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
|
|
https://www.lammps.org/, Sandia National Laboratories
|
|
Steve Plimpton, sjplimp@sandia.gov
|
|
|
|
Copyright (2003) Sandia Corporation. Under the terms of Contract
|
|
DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains
|
|
certain rights in this software. This software is distributed under
|
|
the GNU General Public License.
|
|
|
|
See the README file in the top-level LAMMPS directory.
|
|
------------------------------------------------------------------------- */
|
|
|
|
/* ----------------------------------------------------------------------
|
|
Contributing author: Mike Brown (ORNL)
|
|
------------------------------------------------------------------------- */
|
|
|
|
#include "pair_sw_gpu.h"
|
|
|
|
#include "atom.h"
|
|
#include "comm.h"
|
|
#include "domain.h"
|
|
#include "error.h"
|
|
#include "force.h"
|
|
#include "gpu_extra.h"
|
|
#include "memory.h"
|
|
#include "neigh_list.h"
|
|
#include "neigh_request.h"
|
|
#include "neighbor.h"
|
|
#include "suffix.h"
|
|
|
|
using namespace LAMMPS_NS;
|
|
|
|
// External functions from cuda library for atom decomposition
|
|
|
|
int sw_gpu_init(const int ntypes, const int inum, const int nall,
|
|
const int max_nbors, const double cell_size, int &gpu_mode,
|
|
FILE *screen, double **ncutsq, double **ncut, double **sigma,
|
|
double **powerp, double **powerq, double **sigma_gamma,
|
|
double **c1, double **c2, double **c3,double **c4,
|
|
double **c5, double **c6, double ***lambda_epsilon,
|
|
double ***costheta, const int *map, int ***e2param);
|
|
void sw_gpu_clear();
|
|
int ** sw_gpu_compute_n(const int ago, const int inum, 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);
|
|
void sw_gpu_compute(const int ago, const int nloc, const int nall,
|
|
const int ln, 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 sw_gpu_bytes();
|
|
|
|
#define MAXLINE 1024
|
|
#define DELTA 4
|
|
|
|
/* ---------------------------------------------------------------------- */
|
|
|
|
PairSWGPU::PairSWGPU(LAMMPS *lmp) : PairSW(lmp), gpu_mode(GPU_FORCE)
|
|
{
|
|
cpu_time = 0.0;
|
|
reinitflag = 0;
|
|
suffix_flag |= Suffix::GPU;
|
|
GPU_EXTRA::gpu_ready(lmp->modify, lmp->error);
|
|
|
|
cutghost = nullptr;
|
|
ghostneigh = 1;
|
|
}
|
|
|
|
/* ----------------------------------------------------------------------
|
|
check if allocated, since class can be destructed when incomplete
|
|
------------------------------------------------------------------------- */
|
|
|
|
PairSWGPU::~PairSWGPU()
|
|
{
|
|
sw_gpu_clear();
|
|
if (allocated)
|
|
memory->destroy(cutghost);
|
|
}
|
|
|
|
/* ---------------------------------------------------------------------- */
|
|
|
|
void PairSWGPU::compute(int eflag, int vflag)
|
|
{
|
|
ev_init(eflag,vflag);
|
|
|
|
int nall = atom->nlocal + atom->nghost;
|
|
int inum, host_start;
|
|
|
|
bool success = true;
|
|
int *ilist, *numneigh, **firstneigh;
|
|
if (gpu_mode != GPU_FORCE) {
|
|
double sublo[3],subhi[3];
|
|
if (domain->triclinic == 0) {
|
|
sublo[0] = domain->sublo[0];
|
|
sublo[1] = domain->sublo[1];
|
|
sublo[2] = domain->sublo[2];
|
|
subhi[0] = domain->subhi[0];
|
|
subhi[1] = domain->subhi[1];
|
|
subhi[2] = domain->subhi[2];
|
|
} else {
|
|
domain->bbox(domain->sublo_lamda,domain->subhi_lamda,sublo,subhi);
|
|
}
|
|
inum = atom->nlocal;
|
|
firstneigh = sw_gpu_compute_n(neighbor->ago, inum, nall,
|
|
atom->x, atom->type, sublo,
|
|
subhi, atom->tag, atom->nspecial,
|
|
atom->special, eflag, vflag, eflag_atom,
|
|
vflag_atom, host_start,
|
|
&ilist, &numneigh, cpu_time, success);
|
|
} else {
|
|
inum = list->inum;
|
|
ilist = list->ilist;
|
|
numneigh = list->numneigh;
|
|
firstneigh = list->firstneigh;
|
|
|
|
sw_gpu_compute(neighbor->ago, inum, nall, inum+list->gnum,
|
|
atom->x, atom->type, ilist, numneigh, firstneigh, eflag,
|
|
vflag, eflag_atom, vflag_atom, host_start, cpu_time,
|
|
success);
|
|
}
|
|
if (!success)
|
|
error->one(FLERR,"Insufficient memory on accelerator");
|
|
}
|
|
|
|
/* ---------------------------------------------------------------------- */
|
|
|
|
void PairSWGPU::allocate()
|
|
{
|
|
PairSW::allocate();
|
|
int n = atom->ntypes;
|
|
|
|
memory->create(cutghost,n+1,n+1,"pair:cutghost");
|
|
}
|
|
|
|
/* ----------------------------------------------------------------------
|
|
init specific to this pair style
|
|
------------------------------------------------------------------------- */
|
|
|
|
void PairSWGPU::init_style()
|
|
{
|
|
double cell_size = cutmax + neighbor->skin;
|
|
|
|
if (atom->tag_enable == 0)
|
|
error->all(FLERR,"Pair style sw/gpu requires atom IDs");
|
|
if (force->newton_pair != 0)
|
|
error->all(FLERR,"Pair style sw/gpu requires newton pair off");
|
|
|
|
double **c1, **c2, **c3, **c4, **c5, **c6;
|
|
double **ncutsq, **ncut, **sigma, **powerp, **powerq, **sigma_gamma;
|
|
double ***lambda_epsilon, ***costheta;
|
|
c1 = c2 = c3 = c4 = c5 = c6 = nullptr;
|
|
ncutsq = ncut = sigma = powerp = powerq = sigma_gamma = nullptr;
|
|
lambda_epsilon = costheta = nullptr;
|
|
|
|
const int tp1 = atom->ntypes + 1;
|
|
|
|
memory->create(ncutsq, tp1, tp1, "pair:ncutsq");
|
|
memory->create(ncut, tp1, tp1, "pair:ncut");
|
|
memory->create(sigma, tp1, tp1, "pair:sigma");
|
|
memory->create(powerp, tp1, tp1, "pair:powerp");
|
|
memory->create(powerq, tp1, tp1, "pair:powerq");
|
|
memory->create(sigma_gamma, tp1, tp1, "pair:sigma_gamma");
|
|
memory->create(c1, tp1, tp1, "pair:c1");
|
|
memory->create(c2, tp1, tp1, "pair:c2");
|
|
memory->create(c3, tp1, tp1, "pair:c3");
|
|
memory->create(c4, tp1, tp1, "pair:c4");
|
|
memory->create(c5, tp1, tp1, "pair:c5");
|
|
memory->create(c6, tp1, tp1, "pair:c6");
|
|
memory->create(lambda_epsilon, tp1, tp1, tp1, "pair:lambda_epsilon");
|
|
memory->create(costheta, tp1, tp1, tp1, "pair:costheta");
|
|
|
|
for (int ii = 1; ii < tp1; ii++) {
|
|
int i = map[ii];
|
|
for (int jj = 1; jj < tp1; jj++) {
|
|
int j = map[jj];
|
|
if (i < 0 || j < 0)
|
|
continue;
|
|
else {
|
|
int ijparam = elem3param[i][j][j];
|
|
ncutsq[ii][jj] = params[ijparam].cutsq;
|
|
ncut[ii][jj] = params[ijparam].cut;
|
|
sigma[ii][jj] = params[ijparam].sigma;
|
|
powerp[ii][jj] = params[ijparam].powerp;
|
|
powerq[ii][jj] = params[ijparam].powerq;
|
|
sigma_gamma[ii][jj] = params[ijparam].sigma_gamma;
|
|
c1[ii][jj] = params[ijparam].c1;
|
|
c2[ii][jj] = params[ijparam].c2;
|
|
c3[ii][jj] = params[ijparam].c3;
|
|
c4[ii][jj] = params[ijparam].c4;
|
|
c5[ii][jj] = params[ijparam].c5;
|
|
c6[ii][jj] = params[ijparam].c6;
|
|
}
|
|
|
|
for (int kk = 1; kk < tp1; kk++) {
|
|
int k = map[kk];
|
|
if (k < 0)
|
|
continue;
|
|
else {
|
|
int ijkparam = elem3param[i][j][k];
|
|
costheta[ii][jj][kk] = params[ijkparam].costheta;
|
|
lambda_epsilon[ii][jj][kk] = params[ijkparam].lambda_epsilon;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
int mnf = 5e-2 * neighbor->oneatom;
|
|
int success = sw_gpu_init(tp1, atom->nlocal, atom->nlocal+atom->nghost, mnf,
|
|
cell_size, gpu_mode, screen, ncutsq, ncut, sigma,
|
|
powerp, powerq, sigma_gamma, c1, c2, c3, c4, c5,
|
|
c6, lambda_epsilon, costheta, map, elem3param);
|
|
|
|
memory->destroy(ncutsq);
|
|
memory->destroy(ncut);
|
|
memory->destroy(sigma);
|
|
memory->destroy(powerp);
|
|
memory->destroy(powerq);
|
|
memory->destroy(sigma_gamma);
|
|
memory->destroy(c1);
|
|
memory->destroy(c2);
|
|
memory->destroy(c3);
|
|
memory->destroy(c4);
|
|
memory->destroy(c5);
|
|
memory->destroy(c6);
|
|
memory->destroy(lambda_epsilon);
|
|
memory->destroy(costheta);
|
|
|
|
GPU_EXTRA::check_flag(success,error,world);
|
|
|
|
if (gpu_mode == GPU_FORCE) {
|
|
int irequest = neighbor->request(this,instance_me);
|
|
neighbor->requests[irequest]->half = 0;
|
|
neighbor->requests[irequest]->full = 1;
|
|
neighbor->requests[irequest]->ghost = 1;
|
|
}
|
|
if (comm->cutghostuser < (2.0*cutmax + neighbor->skin)) {
|
|
comm->cutghostuser=2.0*cutmax + neighbor->skin;
|
|
if (comm->me == 0)
|
|
error->warning(FLERR,"Increasing communication cutoff for GPU style");
|
|
}
|
|
}
|
|
|
|
/* ----------------------------------------------------------------------
|
|
init for one type pair i,j and corresponding j,i
|
|
------------------------------------------------------------------------- */
|
|
|
|
double PairSWGPU::init_one(int i, int j)
|
|
{
|
|
if (setflag[i][j] == 0) error->all(FLERR,"All pair coeffs are not set");
|
|
cutghost[i][j] = cutmax;
|
|
cutghost[j][i] = cutmax;
|
|
|
|
return cutmax;
|
|
}
|
|
|