Page Menu
Home
c4science
Search
Configure Global Search
Log In
Files
F69120863
pair_buck_coul_long_gpu.cpp
No One
Temporary
Actions
Download File
Edit File
Delete File
View Transforms
Subscribe
Mute Notifications
Award Token
Subscribers
None
File Metadata
Details
File Info
Storage
Attached
Created
Sun, Jun 30, 08:28
Size
9 KB
Mime Type
text/x-c
Expires
Tue, Jul 2, 08:28 (2 d)
Engine
blob
Format
Raw Data
Handle
18567389
Attached To
rLAMMPS lammps
pair_buck_coul_long_gpu.cpp
View Options
/* ----------------------------------------------------------------------
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
http://lammps.sandia.gov, 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: Trung Dac Nguyen (ORNL)
------------------------------------------------------------------------- */
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include "pair_buck_coul_long_gpu.h"
#include "atom.h"
#include "atom_vec.h"
#include "comm.h"
#include "force.h"
#include "neighbor.h"
#include "neigh_list.h"
#include "integrate.h"
#include "memory.h"
#include "error.h"
#include "neigh_request.h"
#include "universe.h"
#include "update.h"
#include "domain.h"
#include <string.h>
#include "kspace.h"
#include "gpu_extra.h"
#define EWALD_F 1.12837917
#define EWALD_P 0.3275911
#define A1 0.254829592
#define A2 -0.284496736
#define A3 1.421413741
#define A4 -1.453152027
#define A5 1.061405429
using namespace LAMMPS_NS;
// External functions from cuda library for atom decomposition
int buckcl_gpu_init(const int ntypes, double **cutsq, double **host_rhoinv,
double **host_buck1, double **host_buck2, double **host_a,
double **host_c, double **offset, 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 g_ewald);
void buckcl_gpu_clear();
int** buckcl_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);
void buckcl_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);
double buckcl_gpu_bytes();
/* ---------------------------------------------------------------------- */
PairBuckCoulLongGPU::PairBuckCoulLongGPU(LAMMPS *lmp) :
PairBuckCoulLong(lmp), gpu_mode(GPU_FORCE)
{
respa_enable = 0;
reinitflag = 0;
cpu_time = 0.0;
GPU_EXTRA::gpu_ready(lmp->modify, lmp->error);
}
/* ----------------------------------------------------------------------
free all arrays
------------------------------------------------------------------------- */
PairBuckCoulLongGPU::~PairBuckCoulLongGPU()
{
buckcl_gpu_clear();
}
/* ---------------------------------------------------------------------- */
void PairBuckCoulLongGPU::compute(int eflag, int vflag)
{
if (eflag || vflag) ev_setup(eflag,vflag);
else evflag = vflag_fdotr = 0;
int nall = atom->nlocal + atom->nghost;
int inum, host_start;
bool success = true;
int *ilist, *numneigh, **firstneigh;
if (gpu_mode != GPU_FORCE) {
inum = atom->nlocal;
firstneigh = buckcl_gpu_compute_n(neighbor->ago, inum, nall, atom->x,
atom->type, domain->sublo, domain->subhi,
atom->tag, atom->nspecial, atom->special,
eflag, vflag, eflag_atom, vflag_atom,
host_start, &ilist, &numneigh, cpu_time,
success, atom->q, domain->boxlo,
domain->prd);
} else {
inum = list->inum;
ilist = list->ilist;
numneigh = list->numneigh;
firstneigh = list->firstneigh;
buckcl_gpu_compute(neighbor->ago, inum, nall, atom->x, atom->type,
ilist, numneigh, firstneigh, eflag, vflag, eflag_atom,
vflag_atom, host_start, cpu_time, success, atom->q,
atom->nlocal, domain->boxlo, domain->prd);
}
if (!success)
error->one(FLERR,"Insufficient memory on accelerator");
if (host_start<inum) {
cpu_time = MPI_Wtime();
cpu_compute(host_start, inum, eflag, vflag, ilist, numneigh, firstneigh);
cpu_time = MPI_Wtime() - cpu_time;
}
}
/* ----------------------------------------------------------------------
init specific to this pair style
------------------------------------------------------------------------- */
void PairBuckCoulLongGPU::init_style()
{
if (!atom->q_flag)
error->all(FLERR,
"Pair style buck/coul/long/gpu requires atom attribute q");
if (force->newton_pair)
error->all(FLERR,
"Cannot use newton pair with buck/coul/long/gpu pair style");
// Repeat cutsq calculation because done after call to init_style
double maxcut = -1.0;
double cut;
for (int i = 1; i <= atom->ntypes; i++) {
for (int j = i; j <= atom->ntypes; j++) {
if (setflag[i][j] != 0 || (setflag[i][i] != 0 && setflag[j][j] != 0)) {
cut = init_one(i,j);
cut *= cut;
if (cut > maxcut)
maxcut = cut;
cutsq[i][j] = cutsq[j][i] = cut;
} else
cutsq[i][j] = cutsq[j][i] = 0.0;
}
}
double cell_size = sqrt(maxcut) + neighbor->skin;
cut_coulsq = cut_coul * cut_coul;
// insure use of KSpace long-range solver, set g_ewald
if (force->kspace == NULL)
error->all(FLERR,"Pair style requires a KSpace style");
g_ewald = force->kspace->g_ewald;
int maxspecial=0;
if (atom->molecular)
maxspecial=atom->maxspecial;
int success = buckcl_gpu_init(atom->ntypes+1, cutsq, rhoinv, buck1, buck2,
a, c, offset, force->special_lj, atom->nlocal,
atom->nlocal+atom->nghost, 300, maxspecial,
cell_size, gpu_mode, screen, cut_ljsq,
cut_coulsq, force->special_coul, force->qqrd2e,
g_ewald);
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;
}
}
/* ---------------------------------------------------------------------- */
double PairBuckCoulLongGPU::memory_usage()
{
double bytes = Pair::memory_usage();
return bytes + buckcl_gpu_bytes();
}
/* ---------------------------------------------------------------------- */
void PairBuckCoulLongGPU::cpu_compute(int start, int inum, int eflag,
int vflag, int *ilist, int *numneigh,
int **firstneigh)
{
int i,j,ii,jj,jnum,itype,jtype;
double qtmp,xtmp,ytmp,ztmp,delx,dely,delz,evdwl,ecoul,fpair;
double r,rexp,r2inv,r6inv,forcecoul,forcebuck,factor_coul,factor_lj;
double grij,expm2,prefactor,t,erfc;
int *jlist;
double rsq;
evdwl = ecoul = 0.0;
double **x = atom->x;
double **f = atom->f;
double *q = atom->q;
int *type = atom->type;
double *special_coul = force->special_coul;
double *special_lj = force->special_lj;
double qqrd2e = force->qqrd2e;
// loop over neighbors of my atoms
for (ii = start; ii < inum; ii++) {
i = ilist[ii];
qtmp = q[i];
xtmp = x[i][0];
ytmp = x[i][1];
ztmp = x[i][2];
itype = type[i];
jlist = firstneigh[i];
jnum = numneigh[i];
for (jj = 0; jj < jnum; jj++) {
j = jlist[jj];
factor_lj = special_lj[sbmask(j)];
factor_coul = special_coul[sbmask(j)];
j &= NEIGHMASK;
delx = xtmp - x[j][0];
dely = ytmp - x[j][1];
delz = ztmp - x[j][2];
rsq = delx*delx + dely*dely + delz*delz;
jtype = type[j];
if (rsq < cutsq[itype][jtype]) {
r2inv = 1.0/rsq;
r = sqrt(rsq);
if (rsq < cut_coulsq) {
grij = g_ewald * r;
expm2 = exp(-grij*grij);
t = 1.0 / (1.0 + EWALD_P*grij);
erfc = t * (A1+t*(A2+t*(A3+t*(A4+t*A5)))) * expm2;
prefactor = qqrd2e * qtmp*q[j]/r;
forcecoul = prefactor * (erfc + EWALD_F*grij*expm2);
if (factor_coul < 1.0) forcecoul -= (1.0-factor_coul)*prefactor;
} else forcecoul = 0.0;
if (rsq < cut_ljsq[itype][jtype]) {
r6inv = r2inv*r2inv*r2inv;
rexp = exp(-r*rhoinv[itype][jtype]);
forcebuck = buck1[itype][jtype]*r*rexp - buck2[itype][jtype]*r6inv;
} else forcebuck = 0.0;
fpair = (forcecoul + factor_lj*forcebuck) * r2inv;
f[i][0] += delx*fpair;
f[i][1] += dely*fpair;
f[i][2] += delz*fpair;
if (eflag) {
if (rsq < cut_coulsq) {
ecoul = prefactor*erfc;
if (factor_coul < 1.0) ecoul -= (1.0-factor_coul)*prefactor;
} else ecoul = 0.0;
if (rsq < cut_ljsq[itype][jtype]) {
evdwl = a[itype][jtype]*rexp - c[itype][jtype]*r6inv -
offset[itype][jtype];
evdwl *= factor_lj;
} else evdwl = 0.0;
}
if (evflag) ev_tally_full(i,evdwl,ecoul,fpair,delx,dely,delz);
}
}
}
}
Event Timeline
Log In to Comment