ICODE-ADC/lammps
4
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

first working implementation of fix alchemy. still w/o support for pressure and variable cell.

Browse Source
This commit is contained in:
Axel Kohlmeyer
2023-02-24 19:00:53 -05:00
parent 7a4b23938e
commit 6d12f7925b
3 changed files with 45 additions and 40 deletions
Download Patch File Download Diff File Expand all files Collapse all files

75
src/REPLICA/fix_alchemy.cpp
View File

@ -15,8 +15,11 @@
#include "atom.h"
#include "comm.h"
#include "compute.h"
#include "error.h"
#include "memory.h"
#include "modify.h"
#include "respa.h"
#include "universe.h"
#include "update.h"
@ -35,14 +38,17 @@ static double get_lambda(const bigint &step, const bigint &begin, const bigint &
/* ---------------------------------------------------------------------- */
FixAlchemy::FixAlchemy(LAMMPS *lmp, int narg, char **arg) : Fix(lmp, narg, arg), coordbuf(nullptr)
FixAlchemy::FixAlchemy(LAMMPS *lmp, int narg, char **arg) : Fix(lmp, narg, arg)
{
if (narg != 3) error->all(FLERR, "Incorrect number of arguments for fix alchemy");
lambda = epot[0] = epot[1] = 0.0;
no_change_box = 1;
vector_flag = 1;
size_vector = 3;
size_vector = 4;
extvector = 0;
ilevel_respa = 0;
// set up communicators
if (universe->nworlds != 2) error->all(FLERR, "Must use exactly two partitions");
@ -50,7 +56,8 @@ FixAlchemy::FixAlchemy(LAMMPS *lmp, int narg, char **arg) : Fix(lmp, narg, arg),
int key = universe->iworld;
MPI_Comm_split(universe->uworld, color, key, &samerank);
if (narg != 3) error->all(FLERR, "Incorrect number of arguments for fix alchemy");
id_pe = std::string(id) + "_pe";
modify->add_compute(id_pe + " all pe");
}
/* ---------------------------------------------------------------------- */
@ -58,6 +65,7 @@ FixAlchemy::FixAlchemy(LAMMPS *lmp, int narg, char **arg) : Fix(lmp, narg, arg),
FixAlchemy::~FixAlchemy()
{
MPI_Comm_free(&samerank);
modify->delete_compute(id_pe);
}
/* ---------------------------------------------------------------------- */
@ -76,53 +84,50 @@ void FixAlchemy::init() {}
/* ---------------------------------------------------------------------- */
void FixAlchemy::setup(int /*vflag*/)
void FixAlchemy::setup(int vflag)
{
const int nall = atom->nlocal;
const auto tag = atom->tag;
const auto x = atom->x;
memory->destroy(coordbuf);
memory->create(coordbuf, 4 * sizeof(double) * nall, "alchemy:coordbuf");
if (universe->iworld == 0) {
int m = 0;
for (int i = 0; i < nall; ++i) {
coordbuf[m++] = ubuf(tag[i]).d;
coordbuf[m++] = x[i][0];
coordbuf[m++] = x[i][1];
coordbuf[m++] = x[i][2];
}
if (utils::strmatch(update->integrate_style, "^respa")) {
auto respa = dynamic_cast<Respa *>(update->integrate);
respa->copy_flevel_f(ilevel_respa);
post_force_respa(vflag, ilevel_respa, 0);
respa->copy_f_flevel(ilevel_respa);
} else {
post_force(vflag);
}
MPI_Bcast(coordbuf, 4 * nall, MPI_DOUBLE, 0, samerank);
if (universe->iworld == 1) {
int m = 0;
for (int i = 0; i < nall; ++i) {
tagint mytag = (tagint) ubuf(coordbuf[m++]).i;
x[atom->map(mytag)][0] = coordbuf[m++];
x[atom->map(mytag)][1] = coordbuf[m++];
x[atom->map(mytag)][2] = coordbuf[m++];
}
}
lambda = get_lambda(update->ntimestep, update->beginstep, update->endstep, universe->iworld);
}
/* ---------------------------------------------------------------------- */
void FixAlchemy::post_integrate()
{
int nall = 3 * (atom->nlocal + atom->nghost);
// MPI_Bcast(&atom->x[0][0], nall, MPI_DOUBLE, 0, samerank);
const int nall = atom->nlocal + atom->nghost;
MPI_Bcast(&atom->x[0][0], 3 * nall, MPI_DOUBLE, 0, samerank);
}
/* ---------------------------------------------------------------------- */
void FixAlchemy::post_force(int /*vflag*/)
{
const int nall = atom->nlocal;
auto f = atom->f;
lambda = get_lambda(update->ntimestep, update->beginstep, update->endstep, universe->iworld);
double *commbuf = new double[3 * nall];
int m = 0;
for (int i = 0; i < nall; ++i) {
commbuf[m++] = f[i][0] * lambda;
commbuf[m++] = f[i][1] * lambda;
commbuf[m++] = f[i][2] * lambda;
}
MPI_Allreduce(commbuf, &f[0][0], 3 * nall, MPI_DOUBLE, MPI_SUM, samerank);
auto pe = modify->get_compute_by_id(id_pe);
commbuf[0] = commbuf[1] = commbuf[2] = 0.0;
commbuf[universe->iworld] = pe->compute_scalar() / comm->nprocs;
commbuf[2] = lambda * pe->compute_scalar() / comm->nprocs;
MPI_Allreduce(commbuf, epot, 3, MPI_DOUBLE, MPI_SUM, universe->uworld);
delete[] commbuf;
pe->addstep(update->ntimestep + 1);
}
/* ---------------------------------------------------------------------- */