Page MenuHomec4science
Files F65307174

fix_omp.cpp
No OneTemporary
Actions

File Metadata

Created
Sun, Jun 2, 19:30

fix_omp.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.
------------------------------------------------------------------------- */
/* ----------------------------------------------------------------------
OpenMP based threading support for LAMMPS
Contributing author: Axel Kohlmeyer (Temple U)
------------------------------------------------------------------------- */
#include "atom.h"
#include "comm.h"
#include "error.h"
#include "force.h"
#include "neighbor.h"
#include "neigh_request.h"
#include "update.h"
#include "integrate.h"
#include "min.h"
#include "fix_omp.h"
#include "thr_data.h"
#include "thr_omp.h"
#include "pair_hybrid.h"
#include "bond_hybrid.h"
#include "angle_hybrid.h"
#include "dihedral_hybrid.h"
#include "improper_hybrid.h"
#include "kspace.h"
#include <string.h>
#include <stdlib.h>
#include <stdio.h>
#include "suffix.h"
#if defined(LMP_USER_CUDA)
#include "cuda_modify_flags.h"
#endif
using namespace LAMMPS_NS;
using namespace FixConst;
#if defined(LMP_USER_CUDA)
using namespace FixConstCuda;
#endif
static int get_tid()
{
int tid = 0;
#if defined(_OPENMP)
tid = omp_get_thread_num();
#endif
return tid;
}
/* ---------------------------------------------------------------------- */
FixOMP::FixOMP(LAMMPS *lmp, int narg, char **arg) : Fix(lmp, narg, arg),
thr(NULL), last_omp_style(NULL), last_pair_hybrid(NULL),
_nthr(-1), _neighbor(true), _newton(false)
{
if ((narg < 4) || (narg > 6)) error->all(FLERR,"Illegal fix OMP command");
if (strcmp(arg[1],"all") != 0) error->all(FLERR,"Illegal fix OMP command");
int nthreads = 1;
if (narg > 3) {
#if defined(_OPENMP)
if (strcmp(arg[3],"*") == 0)
#pragma omp parallel default(none) shared(nthreads)
nthreads = omp_get_num_threads();
else
nthreads = atoi(arg[3]);
#endif
}
if (nthreads < 1)
error->all(FLERR,"Illegal number of threads requested.");
if (nthreads != comm->nthreads) {
#if defined(_OPENMP)
omp_set_num_threads(nthreads);
#endif
comm->nthreads = nthreads;
if (comm->me == 0) {
if (screen)
fprintf(screen," reset %d OpenMP thread(s) per MPI task\n", nthreads);
if (logfile)
fprintf(logfile," reset %d OpenMP thread(s) per MPI task\n", nthreads);
}
}
if (narg > 4) {
if (strcmp(arg[4],"force/neigh") == 0)
_neighbor = true;
else if (strcmp(arg[4],"force") == 0)
_neighbor = false;
else
error->all(FLERR,"Illegal fix omp mode requested.");
if (comm->me == 0) {
const char * const mode = _neighbor ? "OpenMP capable" : "serial";
if (screen)
fprintf(screen," using %s neighbor list subroutines\n", mode);
if (logfile)
fprintf(logfile," using %s neighbor list subroutines\n", mode);
}
}
#if 0 /* to be enabled when we can switch between half and full neighbor lists */
if (narg > 5) {
if (strcmp(arg[5],"neigh/half") == 0)
_newton = true;
else if (strcmp(arg[5],"neigh/full") == 0)
_newton = false;
else
error->all(FLERR,"Illegal fix OMP command");
if (comm->me == 0) {
const char * const mode = _newton ? "half" : "full";
if (screen)
fprintf(screen," using /omp styles with %s neighbor list builds\n", mode);
if (logfile)
fprintf(logfile," using /omp styles with %s neighbor list builds\n", mode);
}
}
#endif
// allocate list for per thread accumulator manager class instances
// and then have each thread create an instance of this class to
// encourage the OS to use storage that is "close" to each thread's CPU.
thr = new ThrData *[nthreads];
_nthr = nthreads;
#if defined(_OPENMP)
#pragma omp parallel default(none)
#endif
{
const int tid = get_tid();
thr[tid] = new ThrData(tid);
}
}
/* ---------------------------------------------------------------------- */
FixOMP::~FixOMP()
{
#if defined(_OPENMP)
#pragma omp parallel default(none)
#endif
{
const int tid = get_tid();
delete thr[tid];
}
delete[] thr;
}
/* ---------------------------------------------------------------------- */
int FixOMP::setmask()
{
// compatibility with USER-CUDA
// our fix doesn't need any data transfer.
#if defined(LMP_USER_CUDA)
if (lmp->cuda) {
int mask = 0;
mask |= PRE_FORCE_CUDA;
mask |= PRE_FORCE_RESPA;
mask |= MIN_PRE_FORCE;
return mask;
}
#endif
int mask = 0;
mask |= PRE_FORCE;
mask |= PRE_FORCE_RESPA;
mask |= MIN_PRE_FORCE;
return mask;
}
/* ---------------------------------------------------------------------- */
void FixOMP::init()
{
if (strstr(update->integrate_style,"respa") != NULL)
error->all(FLERR,"Cannot use r-RESPA with /omp styles");
int check_hybrid;
last_pair_hybrid = NULL;
last_omp_style = NULL;
char *last_omp_name = NULL;
char *last_hybrid_name = NULL;
char *last_force_name = NULL;
// determine which is the last force style with OpenMP
// support as this is the one that has to reduce the forces
#define CheckStyleForOMP(name) \
check_hybrid = 0; \
if (force->name) { \
if ( (strcmp(force->name ## _style,"hybrid") == 0) || \
(strcmp(force->name ## _style,"hybrid/overlay") == 0) ) \
check_hybrid=1; \
if (force->name->suffix_flag & Suffix::OMP) { \
last_force_name = #name; \
last_omp_name = force->name ## _style; \
last_omp_style = (void *) force->name; \
} \
}
#define CheckHybridForOMP(name,Class) \
if (check_hybrid) { \
Class ## Hybrid *style = (Class ## Hybrid *) force->name; \
for (int i=0; i < style->nstyles; i++) { \
if (style->styles[i]->suffix_flag & Suffix::OMP) { \
last_force_name = #name; \
last_omp_name = style->keywords[i]; \
last_omp_style = style->styles[i]; \
} \
} \
}
CheckStyleForOMP(pair);
CheckHybridForOMP(pair,Pair);
if (check_hybrid) {
last_pair_hybrid = last_omp_style;
last_hybrid_name = last_omp_name;
}
CheckStyleForOMP(bond);
CheckHybridForOMP(bond,Bond);
CheckStyleForOMP(angle);
CheckHybridForOMP(angle,Angle);
CheckStyleForOMP(dihedral);
CheckHybridForOMP(dihedral,Dihedral);
CheckStyleForOMP(improper);
CheckHybridForOMP(improper,Improper);
CheckStyleForOMP(kspace);
#undef CheckStyleForOMP
#undef CheckHybridForOMP
set_neighbor_omp();
// diagnostic output
if (comm->me == 0) {
if (last_omp_style) {
if (last_pair_hybrid) {
if (screen)
fprintf(screen,"Hybrid pair style last /omp style %s\n", last_hybrid_name);
if (logfile)
fprintf(logfile,"Hybrid pair style last /omp style %s\n", last_hybrid_name);
}
if (screen)
fprintf(screen,"Last active /omp style is %s_style %s\n",
last_force_name, last_omp_name);
if (logfile)
fprintf(logfile,"Last active /omp style is %s_style %s\n",
last_force_name, last_omp_name);
} else {
if (screen)
fprintf(screen,"No /omp style for force computation currently active\n");
if (logfile)
fprintf(screen,"No /omp style for force computation currently active\n");
}
}
}
/* ---------------------------------------------------------------------- */
void FixOMP::set_neighbor_omp()
{
// select or deselect multi-threaded neighbor
// list build depending on setting in package omp.
// NOTE: since we are at the top of the list of
// fixes, we cannot adjust neighbor lists from
// other fixes. those have to be re-implemented
// as /omp fix styles. :-(
const int neigh_omp = _neighbor ? 1 : 0;
const int nrequest = neighbor->nrequest;
for (int i = 0; i < nrequest; ++i)
neighbor->requests[i]->omp = neigh_omp;
}
/* ---------------------------------------------------------------------- */
// adjust size and clear out per thread accumulator arrays
void FixOMP::pre_force(int)
{
const int nall = atom->nlocal + atom->nghost;
double **f = atom->f;
double **torque = atom->torque;
double *erforce = atom->erforce;
double *de = atom->de;
double *drho = atom->drho;
#if defined(_OPENMP)
#pragma omp parallel default(none) shared(f,torque,erforce,de,drho)
#endif
{
const int tid = get_tid();
thr[tid]->check_tid(tid);
thr[tid]->init_force(nall,f,torque,erforce,de,drho);
}
}
/* ---------------------------------------------------------------------- */
double FixOMP::memory_usage()
{
double bytes = comm->nthreads * (sizeof(ThrData *) + sizeof(ThrData));
bytes += comm->nthreads * thr[0]->memory_usage();
return bytes;
}

Event Timeline

c4science · Help