diff --git a/src/USER-MISC/fix_ipi.cpp b/src/USER-MISC/fix_ipi.cpp
index f9b0b519a..20b3c90f8 100644
--- a/src/USER-MISC/fix_ipi.cpp
+++ b/src/USER-MISC/fix_ipi.cpp
@@ -1,441 +1,481 @@
/* ----------------------------------------------------------------------
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: Michele Ceriotti (EPFL), Axel Kohlmeyer (Temple U)
------------------------------------------------------------------------- */
#include <mpi.h>
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include "fix_ipi.h"
#include "atom.h"
#include "force.h"
#include "update.h"
#include "respa.h"
#include "error.h"
#include "kspace.h"
#include "modify.h"
#include "compute.h"
#include "comm.h"
#include "neighbor.h"
+#include "irregular.h"
#include "domain.h"
#include "compute_pressure.h"
#include <errno.h>
using namespace LAMMPS_NS;
using namespace FixConst;
/******************************************************************************************
* A fix to interface LAMMPS with i-PI - A Python interface for path integral molecular dynamics
* Michele Ceriotti, EPFL (2014)
* Please cite:
* Ceriotti, M., More, J., & Manolopoulos, D. E. (2014).
* i-PI: A Python interface for ab initio path integral molecular dynamics simulations.
* Computer Physics Communications, 185, 1019–1026. doi:10.1016/j.cpc.2013.10.027
* And see [http://github.com/i-pi/i-pi] to download a version of i-PI
******************************************************************************************/
// socket interface
#ifndef _WIN32
#include <stdlib.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <sys/un.h>
#include <netdb.h>
#endif
#define MSGLEN 12
/* Utility functions to simplify the interface with POSIX sockets */
static void open_socket(int &sockfd, int inet, int port, char* host,
Error *error)
/* Opens a socket.
Args:
sockfd: The id of the socket that will be created.
inet: An integer that determines whether the socket will be an inet or unix
domain socket. Gives unix if 0, inet otherwise.
port: The port number for the socket to be created. Low numbers are often
reserved for important channels, so use of numbers of 4 or more digits is
recommended.
host: The name of the host server.
error: pointer to a LAMMPS Error object
*/
{
int ai_err;
#ifdef _WIN32
error->one(FLERR,"i-PI socket implementation requires UNIX environment");
#else
if (inet>0) { // creates an internet socket
// fetches information on the host
struct addrinfo hints, *res;
char service[256];
memset(&hints, 0, sizeof(hints));
hints.ai_socktype = SOCK_STREAM;
hints.ai_family = AF_UNSPEC;
hints.ai_flags = AI_PASSIVE;
sprintf(service,"%d",port); // convert the port number to a string
ai_err = getaddrinfo(host, service, &hints, &res);
if (ai_err!=0)
error->one(FLERR,"Error fetching host data. Wrong host name?");
// creates socket
sockfd = socket(res->ai_family, res->ai_socktype, res->ai_protocol);
if (sockfd < 0)
error->one(FLERR,"Error opening socket");
// makes connection
if (connect(sockfd, res->ai_addr, res->ai_addrlen) < 0)
error->one(FLERR,"Error opening INET socket: wrong port or server unreachable");
freeaddrinfo(res);
} else { // creates a unix socket
struct sockaddr_un serv_addr;
// fills up details of the socket addres
memset(&serv_addr, 0, sizeof(serv_addr));
serv_addr.sun_family = AF_UNIX;
strcpy(serv_addr.sun_path, "/tmp/ipi_");
strcpy(serv_addr.sun_path+9, host);
// creates the socket
sockfd = socket(AF_UNIX, SOCK_STREAM, 0);
// connects
if (connect(sockfd, (struct sockaddr *) &serv_addr, sizeof(serv_addr)) < 0)
error->one(FLERR,"Error opening UNIX socket: server may not be running "
"or the path to the socket unavailable");
}
#endif
}
static void writebuffer(int sockfd, const char *data, int len, Error* error)
/* Writes to a socket.
Args:
sockfd: The id of the socket that will be written to.
data: The data to be written to the socket.
len: The length of the data in bytes.
*/
{
int n;
n = write(sockfd,data,len);
if (n < 0)
error->one(FLERR,"Error writing to socket: broken connection");
}
static void readbuffer(int sockfd, char *data, int len, Error* error)
/* Reads from a socket.
Args:
sockfd: The id of the socket that will be read from.
data: The storage array for data read from the socket.
len: The length of the data in bytes.
*/
{
int n, nr;
n = nr = read(sockfd,data,len);
while (nr>0 && n<len ) {
nr=read(sockfd,&data[n],len-n);
n+=nr;
}
if (n == 0)
error->one(FLERR,"Error reading from socket: broken connection");
}
/* ---------------------------------------------------------------------- */
FixIPI::FixIPI(LAMMPS *lmp, int narg, char **arg) :
- Fix(lmp, narg, arg)
+ Fix(lmp, narg, arg), irregular(NULL)
{
/* format for fix:
* fix num group_id ipi host port [unix]
*/
if (strcmp(style,"ipi") != 0 && narg < 5)
error->all(FLERR,"Illegal fix ipi command");
if (atom->tag_enable == 0)
error->all(FLERR,"Cannot use fix ipi without atom IDs");
if (atom->tag_consecutive() == 0)
error->all(FLERR,"Fix ipi requires consecutive atom IDs");
if (strcmp(arg[1],"all"))
error->warning(FLERR,"Fix ipi always uses group all");
host = strdup(arg[3]);
port = force->inumeric(FLERR,arg[4]);
- inet = ((narg > 5) && (strcmp(arg[5],"unix") ==0) ) ? 0 : 1;
+ inet = ((narg > 5) && (strcmp(arg[5],"unix") == 0) ) ? 0 : 1;
master = (comm->me==0) ? 1 : 0;
+ // check if forces should be reinitialized and set flag
+ reset_flag = ((narg > 6) || (strcmp(arg[5],"reset") == 0) ) ? 1 : 0;
+
hasdata = bsize = 0;
// creates a temperature compute for all atoms
char** newarg = new char*[3];
newarg[0] = (char *) "IPI_TEMP";
newarg[1] = (char *) "all";
newarg[2] = (char *) "temp";
modify->add_compute(3,newarg);
delete [] newarg;
// creates a pressure compute to extract the virial
newarg = new char*[5];
newarg[0] = (char *) "IPI_PRESS";
newarg[1] = (char *) "all";
newarg[2] = (char *) "pressure";
newarg[3] = (char *) "IPI_TEMP";
newarg[4] = (char *) "virial";
modify->add_compute(5,newarg);
delete [] newarg;
+
+ // create instance of Irregular class
+ irregular = new Irregular(lmp);
}
/* ---------------------------------------------------------------------- */
FixIPI::~FixIPI()
{
if (bsize) delete[] buffer;
free(host);
modify->delete_compute("IPI_TEMP");
modify->delete_compute("IPI_PRESS");
+ delete irregular;
}
/* ---------------------------------------------------------------------- */
int FixIPI::setmask()
{
int mask = 0;
mask |= INITIAL_INTEGRATE;
mask |= FINAL_INTEGRATE;
return mask;
}
/* ---------------------------------------------------------------------- */
void FixIPI::init()
{
//only opens socket on master process
if (master) open_socket(ipisock, inet, port, host, error);
else ipisock=0;
//! should check for success in socket opening -- but the current open_socket routine dies brutally if unsuccessful
// asks for evaluation of PE at first step
modify->compute[modify->find_compute("thermo_pe")]->invoked_scalar = -1;
modify->addstep_compute_all(update->ntimestep + 1);
kspace_flag = (force->kspace) ? 1 : 0;
// makes sure that neighbor lists are re-built at each step (cannot make assumptions when cycling over beads!)
neighbor->delay = 0;
neighbor->every = 1;
}
void FixIPI::initial_integrate(int vflag)
{
/* This is called at the beginning of the integration loop,
* and will be used to read positions from the socket. Then,
* everything should be updated, since there is no guarantee
* that successive snapshots will be close together (think
* of parallel tempering for instance) */
char header[MSGLEN+1];
if (hasdata)
error->all(FLERR, "i-PI got out of sync in initial_integrate and will die!");
double cellh[9], cellih[9];
int nat;
if (master) { // only read positions on master
// wait until something happens
while (true) {
// while i-PI just asks for status, signal we are ready and wait
readbuffer(ipisock, header, MSGLEN, error); header[MSGLEN]=0;
if (strcmp(header,"STATUS ") == 0 )
writebuffer(ipisock,"READY ",MSGLEN, error);
else break;
}
if (strcmp(header,"EXIT ") == 0 )
error->one(FLERR, "Got EXIT message from i-PI. Now leaving!");
// when i-PI signals it has positions to evaluate new forces,
// read positions and cell data
if (strcmp(header,"POSDATA ") == 0 ) {
readbuffer(ipisock, (char*) cellh, 9*8, error);
readbuffer(ipisock, (char*) cellih, 9*8, error);
readbuffer(ipisock, (char*) &nat, 4, error);
// allocate buffer, but only do this once.
if (bsize==0) {
bsize=3*nat;
buffer = new double[bsize];
} else if (bsize != 3*nat)
error->one(FLERR, "Number of atoms changed along the way.");
// finally read position data into buffer
readbuffer(ipisock, (char*) buffer, 8*bsize, error);
} else
error->one(FLERR, "Wrapper did not send positions, I will now die!");
}
// shares the atomic coordinates with everyone
MPI_Bcast(&nat,1,MPI_INT,0,world);
// must also allocate the buffer on the non-head nodes
if (bsize==0) {
bsize=3*nat;
buffer = new double[bsize];
}
MPI_Bcast(cellh,9,MPI_DOUBLE,0,world);
MPI_Bcast(cellih,9,MPI_DOUBLE,0,world);
MPI_Bcast(buffer,bsize,MPI_DOUBLE,0,world);
//updates atomic coordinates and cell based on the data received
double *boxhi = domain->boxhi;
double *boxlo = domain->boxlo;
double posconv;
posconv=0.52917721*force->angstrom;
boxlo[0] = -0.5*cellh[0]*posconv;
boxlo[1] = -0.5*cellh[4]*posconv;
boxlo[2] = -0.5*cellh[8]*posconv;
boxhi[0] = -boxlo[0];
boxhi[1] = -boxlo[1];
boxhi[2] = -boxlo[2];
domain->xy = cellh[1]*posconv;
domain->xz = cellh[2]*posconv;
domain->yz = cellh[5]*posconv;
- // signal that the box has (or may have) changed
+ // do error checks on simulation box and set small for triclinic boxes
+ domain->set_initial_box();
+ // reset global and local box using the new box dimensions
domain->reset_box();
+ // signal that the box has (or may have) changed
domain->box_change = 1;
- if (kspace_flag) force->kspace->setup();
// picks local atoms from the buffer
double **x = atom->x;
int *mask = atom->mask;
int nlocal = atom->nlocal;
if (igroup == atom->firstgroup) nlocal = atom->nfirst;
for (int i = 0; i < nlocal; i++) {
if (mask[i] & groupbit) {
x[i][0]=buffer[3*(atom->tag[i]-1)+0]*posconv;
x[i][1]=buffer[3*(atom->tag[i]-1)+1]*posconv;
x[i][2]=buffer[3*(atom->tag[i]-1)+2]*posconv;
}
}
+ // insure atoms are in current box & update box via shrink-wrap
+ // has to be be done before invoking Irregular::migrate_atoms()
+ // since it requires atoms be inside simulation box
+
+ if (domain->triclinic) domain->x2lamda(atom->nlocal);
+ domain->pbc();
+ domain->reset_box();
+ if (domain->triclinic) domain->lamda2x(atom->nlocal);
+
+ // move atoms to new processors via irregular()
+ // only needed if migrate_check() says an atom moves to far
+ if (domain->triclinic) domain->x2lamda(atom->nlocal);
+ if (irregular->migrate_check()) irregular->migrate_atoms();
+ if (domain->triclinic) domain->lamda2x(atom->nlocal);
+
+ // check if kspace solver is used
+ if (reset_flag && kspace_flag) {
+ // reset kspace, pair, angles, ... b/c simulation box might have changed.
+ // kspace->setup() is in some cases not enough since, e.g., g_ewald needs
+ // to be reestimated due to changes in box dimensions.
+ force->init();
+ // setup_grid() is necessary for pppm since init() is not calling
+ // setup() nor setup_grid() upon calling init().
+ if (force->kspace->pppmflag) force->kspace->setup_grid();
+ // other kspace styles might need too another setup()?
+ } else if (!reset_flag && kspace_flag) {
+ // original version
+ force->kspace->setup();
+ }
+
// compute PE. makes sure that it will be evaluated at next step
modify->compute[modify->find_compute("thermo_pe")]->invoked_scalar = -1;
modify->addstep_compute_all(update->ntimestep+1);
hasdata=1;
}
void FixIPI::final_integrate()
{
/* This is called after forces and energy have been computed. Now we only need to
* communicate them back to i-PI so that the integration can continue. */
char header[MSGLEN+1];
double vir[9], pot=0.0;
double forceconv, potconv, posconv, pressconv, posconv3;
char retstr[1024];
// conversions from LAMMPS units to atomic units, which are used by i-PI
potconv=3.1668152e-06/force->boltz;
posconv=0.52917721*force->angstrom;
posconv3=posconv*posconv*posconv;
forceconv=potconv*posconv;
pressconv=1/force->nktv2p*potconv*posconv3;
// compute for potential energy
pot=modify->compute[modify->find_compute("thermo_pe")]->compute_scalar();
pot*=potconv;
// probably useless check
if (!hasdata)
error->all(FLERR, "i-PI got out of sync in final_integrate and will die!");
int nat=bsize/3;
double **f= atom->f;
double lbuf[bsize];
// reassembles the force vector from the local arrays
int nlocal = atom->nlocal;
if (igroup == atom->firstgroup) nlocal = atom->nfirst;
for (int i = 0; i < bsize; ++i) lbuf[i]=0.0;
for (int i = 0; i < nlocal; i++) {
lbuf[3*(atom->tag[i]-1)+0]=f[i][0]*forceconv;
lbuf[3*(atom->tag[i]-1)+1]=f[i][1]*forceconv;
lbuf[3*(atom->tag[i]-1)+2]=f[i][2]*forceconv;
}
MPI_Allreduce(lbuf,buffer,bsize,MPI_DOUBLE,MPI_SUM,world);
for (int i = 0; i < 9; ++i) vir[i]=0.0;
int press_id = modify->find_compute("IPI_PRESS");
Compute* comp_p = modify->compute[press_id];
comp_p->compute_vector();
double myvol = domain->xprd*domain->yprd*domain->zprd/posconv3;
vir[0] = comp_p->vector[0]*pressconv*myvol;
vir[4] = comp_p->vector[1]*pressconv*myvol;
vir[8] = comp_p->vector[2]*pressconv*myvol;
vir[1] = comp_p->vector[3]*pressconv*myvol;
vir[2] = comp_p->vector[4]*pressconv*myvol;
vir[5] = comp_p->vector[5]*pressconv*myvol;
retstr[0]=0;
if (master) {
while (true) {
readbuffer(ipisock, header, MSGLEN, error); header[MSGLEN]=0;
if (strcmp(header,"STATUS ") == 0 )
writebuffer(ipisock,"HAVEDATA ",MSGLEN, error);
else break;
}
if (strcmp(header,"EXIT ") == 0 )
error->one(FLERR, "Got EXIT message from i-PI. Now leaving!");
if (strcmp(header,"GETFORCE ") == 0 ) {
writebuffer(ipisock,"FORCEREADY ",MSGLEN, error);
writebuffer(ipisock,(char*) &pot,8, error);
writebuffer(ipisock,(char*) &nat,4, error);
writebuffer(ipisock,(char*) buffer, bsize*8, error);
writebuffer(ipisock,(char*) vir,9*8, error);
nat=strlen(retstr); writebuffer(ipisock,(char*) &nat,4, error);
writebuffer(ipisock,(char*) retstr, nat, error);
}
else
error->one(FLERR, "Wrapper did not ask for forces, I will now die!");
}
hasdata=0;
}
diff --git a/src/USER-MISC/fix_ipi.h b/src/USER-MISC/fix_ipi.h
index 0859a513c..0bb3717de 100644
--- a/src/USER-MISC/fix_ipi.h
+++ b/src/USER-MISC/fix_ipi.h
@@ -1,45 +1,49 @@
/* -*- c++ -*- ----------------------------------------------------------
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.
------------------------------------------------------------------------- */
#ifdef FIX_CLASS
FixStyle(ipi,FixIPI)
#else
#ifndef LMP_FIX_IPI_H
#define LMP_FIX_IPI_H
#include "fix.h"
namespace LAMMPS_NS {
class FixIPI : public Fix {
public:
FixIPI(class LAMMPS *, int, char **);
virtual ~FixIPI();
int setmask();
virtual void init();
virtual void initial_integrate(int);
virtual void final_integrate();
protected:
char *host; int port; int inet, master, hasdata;
int ipisock, me; double *buffer; long bsize;
int kspace_flag;
+ int reset_flag;
+
+ private:
+ class Irregular *irregular;
};
}
#endif
#endif