jianmu
/
lammps
forked from lijiext/lammps
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

git-svn-id: svn://svn.icms.temple.edu/lammps-ro/trunk@12814 f3b2605a-c512-4ea7-a41b-209d697bcdaa

This commit is contained in:
sjplimp 2014-12-02 02:15:39 +00:00
parent 9f649c3123
commit aeb7fec23a
79 changed files with 403 additions and 624 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

37
src/KOKKOS/comm_kokkos.cpp
View File

@ -166,7 +166,6 @@ void CommKokkos::forward_comm_device(int dummy)
{
int n;
MPI_Request request;
MPI_Status status;
AtomVecKokkos *avec = (AtomVecKokkos *) atom->avec;
double **x = atom->x;
double *buf;
@ -199,7 +198,7 @@ void CommKokkos::forward_comm_device(int dummy)
n,MPI_DOUBLE,sendproc[iswap],0,world);
}
if (size_forward_recv[iswap]) MPI_Wait(&request,&status);
if (size_forward_recv[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
atomKK->modified(ExecutionSpaceFromDevice<DeviceType>::
space,X_MASK);
} else if (ghost_velocity) {
@ -212,7 +211,7 @@ void CommKokkos::forward_comm_device(int dummy)
n = avec->pack_comm_vel(sendnum[iswap],sendlist[iswap],
buf_send,pbc_flag[iswap],pbc[iswap]);
if (n) MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
if (size_forward_recv[iswap]) MPI_Wait(&request,&status);
if (size_forward_recv[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
avec->unpack_comm_vel(recvnum[iswap],firstrecv[iswap],buf_recv);
} else {
if (size_forward_recv[iswap])
@ -224,7 +223,7 @@ void CommKokkos::forward_comm_device(int dummy)
if (n)
MPI_Send(k_buf_send.view<DeviceType>().ptr_on_device(),n,
MPI_DOUBLE,sendproc[iswap],0,world);
if (size_forward_recv[iswap]) MPI_Wait(&request,&status);
if (size_forward_recv[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
avec->unpack_comm_kokkos(recvnum[iswap],firstrecv[iswap],k_buf_recv);
}
@ -258,16 +257,16 @@ void CommKokkos::reverse_comm()
atomKK->sync(Device,ALL_MASK);
}
void CommKokkos::forward_comm_fix(Fix *fix)
void CommKokkos::forward_comm_fix(Fix *fix, int size)
{
k_sendlist.sync<LMPHostType>();
CommBrick::forward_comm_fix(fix);
CommBrick::forward_comm_fix(fix,size);
}
void CommKokkos::reverse_comm_fix(Fix *fix)
void CommKokkos::reverse_comm_fix(Fix *fix, int size)
{
k_sendlist.sync<LMPHostType>();
CommBrick::reverse_comm_fix(fix);
CommBrick::reverse_comm_fix(fix, size);
}
void CommKokkos::forward_comm_compute(Compute *compute)
@ -388,12 +387,11 @@ struct BuildExchangeListFunctor {
template<class DeviceType>
void CommKokkos::exchange_device()
{
int i,m,nsend,nrecv,nrecv1,nrecv2,nlocal;
double lo,hi,value;
int i,nsend,nrecv,nrecv1,nrecv2,nlocal;
double lo,hi;
double **x;
double *sublo,*subhi,*buf;
double *sublo,*subhi;
MPI_Request request;
MPI_Status status;
AtomVecKokkos *avec = (AtomVecKokkos *) atom->avec;
// clear global->local map for owned and ghost atoms
@ -496,7 +494,6 @@ void CommKokkos::exchange_device()
if (procgrid[dim] == 1) {
nrecv = nsend;
buf = buf_send;
if (nrecv) {
atom->nlocal=avec->
unpack_exchange_kokkos(k_buf_send,nrecv,atom->nlocal,dim,lo,hi,
@ -505,11 +502,11 @@ void CommKokkos::exchange_device()
}
} else {
MPI_Sendrecv(&nsend,1,MPI_INT,procneigh[dim][0],0,
&nrecv1,1,MPI_INT,procneigh[dim][1],0,world,&status);
&nrecv1,1,MPI_INT,procneigh[dim][1],0,world,MPI_STATUS_IGNORE);
nrecv = nrecv1;
if (procgrid[dim] > 2) {
MPI_Sendrecv(&nsend,1,MPI_INT,procneigh[dim][1],0,
&nrecv2,1,MPI_INT,procneigh[dim][0],0,world,&status);
&nrecv2,1,MPI_INT,procneigh[dim][0],0,world,MPI_STATUS_IGNORE);
nrecv += nrecv2;
}
if (nrecv > maxrecv) grow_recv_kokkos(nrecv);
@ -519,7 +516,7 @@ void CommKokkos::exchange_device()
world,&request);
MPI_Send(k_buf_send.view<DeviceType>().ptr_on_device(),nsend,
MPI_DOUBLE,procneigh[dim][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
if (procgrid[dim] > 2) {
MPI_Irecv(k_buf_recv.view<DeviceType>().ptr_on_device()+nrecv1,
@ -527,10 +524,9 @@ void CommKokkos::exchange_device()
world,&request);
MPI_Send(k_buf_send.view<DeviceType>().ptr_on_device(),nsend,
MPI_DOUBLE,procneigh[dim][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
buf = buf_recv;
if (nrecv) {
atom->nlocal = avec->
unpack_exchange_kokkos(k_buf_recv,nrecv,atom->nlocal,dim,lo,hi,
@ -643,7 +639,6 @@ void CommKokkos::borders_device() {
double **x;
double *buf,*mlo,*mhi;
MPI_Request request;
MPI_Status status;
AtomVecKokkos *avec = (AtomVecKokkos *) atom->avec;
ExecutionSpace exec_space = ExecutionSpaceFromDevice<DeviceType>::space;
@ -799,14 +794,14 @@ void CommKokkos::borders_device() {
if (sendproc[iswap] != me) {
MPI_Sendrecv(&nsend,1,MPI_INT,sendproc[iswap],0,
&nrecv,1,MPI_INT,recvproc[iswap],0,world,&status);
&nrecv,1,MPI_INT,recvproc[iswap],0,world,MPI_STATUS_IGNORE);
if (nrecv*size_border > maxrecv) grow_recv_kokkos(nrecv*size_border);
if (nrecv) MPI_Irecv(k_buf_recv.view<DeviceType>().ptr_on_device(),
nrecv*size_border,MPI_DOUBLE,
recvproc[iswap],0,world,&request);
if (n) MPI_Send(k_buf_send.view<DeviceType>().ptr_on_device(),n,
MPI_DOUBLE,sendproc[iswap],0,world);
if (nrecv) MPI_Wait(&request,&status);
if (nrecv) MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else {
nrecv = nsend;

4
src/KOKKOS/comm_kokkos.h
View File

@ -39,8 +39,8 @@ class CommKokkos : public CommBrick {
void forward_comm_pair(class Pair *); // forward comm from a Pair
void reverse_comm_pair(class Pair *); // reverse comm from a Pair
void forward_comm_fix(class Fix *); // forward comm from a Fix
void reverse_comm_fix(class Fix *); // reverse comm from a Fix
void forward_comm_fix(class Fix *, int size=0); // forward comm from a Fix
void reverse_comm_fix(class Fix *, int size=0); // reverse comm from a Fix
void forward_comm_compute(class Compute *); // forward from a Compute
void reverse_comm_compute(class Compute *); // reverse from a Compute
void forward_comm_dump(class Dump *); // forward comm from a Dump

1
src/KOKKOS/pair_coul_cut_kokkos.cpp
View File

@ -80,7 +80,6 @@ void PairCoulCutKokkos<DeviceType>::compute(int eflag_in, int vflag_in)
if (neighflag == FULL || neighflag == FULLCLUSTER) no_virial_fdotr_compute = 1;
double ecoul = 0.0;
if (eflag || vflag) ev_setup(eflag,vflag);
else evflag = vflag_fdotr = 0;

2
src/KOKKOS/pair_lj_cut_coul_cut_kokkos.cpp
View File

@ -89,8 +89,6 @@ void PairLJCutCoulCutKokkos<DeviceType>::compute(int eflag_in, int vflag_in)
if (neighflag == FULL || neighflag == FULLCLUSTER) no_virial_fdotr_compute = 1;
double evdwl = 0.0;
double ecoul = 0.0;
if (eflag || vflag) ev_setup(eflag,vflag);
else evflag = vflag_fdotr = 0;

4
src/KOKKOS/pair_lj_cut_coul_long_kokkos.cpp
View File

@ -59,7 +59,7 @@ PairLJCutCoulLongKokkos<DeviceType>::PairLJCutCoulLongKokkos(LAMMPS *lmp):PairLJ
datamask_modify = F_MASK | ENERGY_MASK | VIRIAL_MASK;
cutsq = NULL;
cut_ljsq = NULL;
cut_coulsq = NULL;
cut_coulsq = 0.0;
}
@ -97,8 +97,6 @@ void PairLJCutCoulLongKokkos<DeviceType>::compute(int eflag_in, int vflag_in)
if (neighflag == FULL || neighflag == FULLCLUSTER) no_virial_fdotr_compute = 1;
double evdwl = 0.0;
double ecoul = 0.0;
if (eflag || vflag) ev_setup(eflag,vflag);
else evflag = vflag_fdotr = 0;

1
src/KOKKOS/pair_lj_cut_kokkos.cpp
View File

@ -85,7 +85,6 @@ void PairLJCutKokkos<DeviceType>::compute(int eflag_in, int vflag_in)
if (neighflag == FULL || neighflag == FULLCLUSTER) no_virial_fdotr_compute = 1;
double evdwl = 0.0;
if (eflag || vflag) ev_setup(eflag,vflag);
else evflag = vflag_fdotr = 0;

1
src/KOKKOS/pair_table_kokkos.cpp
View File

@ -98,7 +98,6 @@ void PairTableKokkos<DeviceType>::compute_style(int eflag_in, int vflag_in)
if (neighflag == FULL || neighflag == FULLCLUSTER) no_virial_fdotr_compute = 1;
double evdwl = 0.0;
if (eflag || vflag) ev_setup(eflag,vflag);
else evflag = vflag_fdotr = 0;

28
src/KSPACE/gridcomm.cpp
View File

@ -146,37 +146,37 @@ void GridComm::ghost_notify()
int nplanes = inxlo - outxlo;
if (procxlo != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,procxlo,0,
&ghostxhi,1,MPI_INT,procxhi,0,gridcomm,&status);
&ghostxhi,1,MPI_INT,procxhi,0,gridcomm,MPI_STATUS_IGNORE);
else ghostxhi = nplanes;
nplanes = outxhi - inxhi;
if (procxhi != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,procxhi,0,
&ghostxlo,1,MPI_INT,procxlo,0,gridcomm,&status);
&ghostxlo,1,MPI_INT,procxlo,0,gridcomm,MPI_STATUS_IGNORE);
else ghostxlo = nplanes;
nplanes = inylo - outylo;
if (procylo != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,procylo,0,
&ghostyhi,1,MPI_INT,procyhi,0,gridcomm,&status);
&ghostyhi,1,MPI_INT,procyhi,0,gridcomm,MPI_STATUS_IGNORE);
else ghostyhi = nplanes;
nplanes = outyhi - inyhi;
if (procyhi != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,procyhi,0,
&ghostylo,1,MPI_INT,procylo,0,gridcomm,&status);
&ghostylo,1,MPI_INT,procylo,0,gridcomm,MPI_STATUS_IGNORE);
else ghostylo = nplanes;
nplanes = inzlo - outzlo;
if (proczlo != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,proczlo,0,
&ghostzhi,1,MPI_INT,proczhi,0,gridcomm,&status);
&ghostzhi,1,MPI_INT,proczhi,0,gridcomm,MPI_STATUS_IGNORE);
else ghostzhi = nplanes;
nplanes = outzhi - inzhi;
if (proczhi != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,proczhi,0,
&ghostzlo,1,MPI_INT,proczlo,0,gridcomm,&status);
&ghostzlo,1,MPI_INT,proczlo,0,gridcomm,MPI_STATUS_IGNORE);
else ghostzlo = nplanes;
}
@ -243,7 +243,7 @@ void GridComm::setup()
if (procxlo != me)
MPI_Sendrecv(&sendplanes,1,MPI_INT,procxlo,0,
&recvplanes,1,MPI_INT,procxhi,0,gridcomm,&status);
&recvplanes,1,MPI_INT,procxhi,0,gridcomm,MPI_STATUS_IGNORE);
else recvplanes = sendplanes;
swap[nswap].nunpack =
@ -285,7 +285,7 @@ void GridComm::setup()
if (procxhi != me)
MPI_Sendrecv(&sendplanes,1,MPI_INT,procxhi,0,
&recvplanes,1,MPI_INT,procxlo,0,gridcomm,&status);
&recvplanes,1,MPI_INT,procxlo,0,gridcomm,MPI_STATUS_IGNORE);
else recvplanes = sendplanes;
swap[nswap].nunpack =
@ -327,7 +327,7 @@ void GridComm::setup()
if (procylo != me)
MPI_Sendrecv(&sendplanes,1,MPI_INT,procylo,0,
&recvplanes,1,MPI_INT,procyhi,0,gridcomm,&status);
&recvplanes,1,MPI_INT,procyhi,0,gridcomm,MPI_STATUS_IGNORE);
else recvplanes = sendplanes;
swap[nswap].nunpack =
@ -369,7 +369,7 @@ void GridComm::setup()
if (procyhi != me)
MPI_Sendrecv(&sendplanes,1,MPI_INT,procyhi,0,
&recvplanes,1,MPI_INT,procylo,0,gridcomm,&status);
&recvplanes,1,MPI_INT,procylo,0,gridcomm,MPI_STATUS_IGNORE);
else recvplanes = sendplanes;
swap[nswap].nunpack =
@ -411,7 +411,7 @@ void GridComm::setup()
if (proczlo != me)
MPI_Sendrecv(&sendplanes,1,MPI_INT,proczlo,0,
&recvplanes,1,MPI_INT,proczhi,0,gridcomm,&status);
&recvplanes,1,MPI_INT,proczhi,0,gridcomm,MPI_STATUS_IGNORE);
else recvplanes = sendplanes;
swap[nswap].nunpack =
@ -453,7 +453,7 @@ void GridComm::setup()
if (proczhi != me)
MPI_Sendrecv(&sendplanes,1,MPI_INT,proczhi,0,
&recvplanes,1,MPI_INT,proczlo,0,gridcomm,&status);
&recvplanes,1,MPI_INT,proczlo,0,gridcomm,MPI_STATUS_IGNORE);
else recvplanes = sendplanes;
swap[nswap].nunpack =
@ -500,7 +500,7 @@ void GridComm::forward_comm(KSpace *kspace, int which)
swap[m].recvproc,0,gridcomm,&request);
MPI_Send(buf1,nforward*swap[m].npack,MPI_FFT_SCALAR,
swap[m].sendproc,0,gridcomm);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
kspace->unpack_forward(which,buf2,swap[m].nunpack,swap[m].unpacklist);
@ -525,7 +525,7 @@ void GridComm::reverse_comm(KSpace *kspace, int which)
swap[m].sendproc,0,gridcomm,&request);
MPI_Send(buf1,nreverse*swap[m].nunpack,MPI_FFT_SCALAR,
swap[m].recvproc,0,gridcomm);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
kspace->unpack_reverse(which,buf2,swap[m].npack,swap[m].packlist);

1
src/KSPACE/gridcomm.h
View File

@ -50,7 +50,6 @@ class GridComm : protected Pointers {
int nforward,nreverse;
MPI_Comm gridcomm;
MPI_Request request;
MPI_Status status;
// in = inclusive indices of 3d grid chunk that I own
// out = inclusive indices of 3d grid chunk I own plus ghosts I use

14
src/KSPACE/pair_lj_long_tip4p_long.cpp
View File

@ -1580,10 +1580,16 @@ void *PairLJLongTIP4PLong::extract(const char *str, int &dim)
&mix_flag, &cut_lj_global, NULL};
int i;
for (i=0; ids[i]&&strcmp(ids[i], str); ++i);
if (i <= 2) dim = 2;
else dim = 0;
return ptrs[i];
i=0;
while (ids[i] != NULL) {
if (i <=2) dim = 2;
else dim = 0;
if (strcmp(ids[i],str) == 0)
return ptrs[i];
++i;
}
return NULL;
}

3
src/KSPACE/remap.cpp
View File

@ -65,7 +65,6 @@ void remap_3d(FFT_SCALAR *in, FFT_SCALAR *out, FFT_SCALAR *buf,
// use point-to-point communication
if (!plan->usecollective) {
MPI_Status status;
int i,isend,irecv;
FFT_SCALAR *scratch;
@ -105,7 +104,7 @@ void remap_3d(FFT_SCALAR *in, FFT_SCALAR *out, FFT_SCALAR *buf,
// unpack all messages from scratch -> out
for (i = 0; i < plan->nrecv; i++) {
MPI_Waitany(plan->nrecv,plan->request,&irecv,&status);
MPI_Waitany(plan->nrecv,plan->request,&irecv,MPI_STATUS_IGNORE);
plan->unpack(&scratch[plan->recv_bufloc[irecv]],
&out[plan->recv_offset[irecv]],&plan->unpackplan[irecv]);
}

14
src/MANYBODY/pair_comb.cpp
View File

@ -127,7 +127,7 @@ void PairComb::compute(int eflag, int vflag)
int i,j,k,ii,jj,kk,inum,jnum,iparam_i;
int itype,jtype,ktype,iparam_ij,iparam_ijk;
tagint itag,jtag;
double xtmp,ytmp,ztmp,delx,dely,delz,evdwl,ecoul,fpair;
double xtmp,ytmp,ztmp,delx,dely,delz,evdwl,fpair;
double rsq,rsq1,rsq2;
double delr1[3],delr2[3],fi[3],fj[3],fk[3];
double zeta_ij,prefactor;
@ -141,7 +141,7 @@ void PairComb::compute(int eflag, int vflag)
double vionij,fvionij,sr1,sr2,sr3,Eov,Fov;
int sht_jnum, *sht_jlist, nj;
evdwl = ecoul = 0.0;
evdwl = 0.0;
if (eflag || vflag) ev_setup(eflag,vflag);
else evflag = vflag_fdotr = vflag_atom = 0;
@ -1637,7 +1637,7 @@ double PairComb::yasu_char(double *qf_fix, int &igroup)
double xtmp,ytmp,ztmp;
double rsq1,delr1[3];
int *ilist,*jlist,*numneigh,**firstneigh;
double iq,jq,fqi,fqj,fqij,fqjj;
double iq,jq,fqi,fqij,fqjj;
double potal,fac11,fac11e,sr1,sr2,sr3;
int mr1,mr2,mr3,inty,nj;
@ -1672,7 +1672,7 @@ double PairComb::yasu_char(double *qf_fix, int &igroup)
// loop over full neighbor list of my atoms
fqi = fqj = fqij = fqjj = 0.0;
fqi = fqij = fqjj = 0.0;
for (ii = 0; ii < inum; ii ++) {
i = ilist[ii];
@ -1871,9 +1871,9 @@ void PairComb::qfo_short(Param *param, int i, int j, double rsq,
double iq, double jq, double &fqij, double &fqjj)
{
double r,tmp_fc,tmp_exp1,tmp_exp2;
double bigA,Asi,Asj,vrcs;
double Asi,Asj,vrcs;
double romi = param->addrep,rrcs = param->bigr + param->bigd;
double qi,qj,Di,Dj,bigB,Bsi,Bsj;
double qi,qj,Di,Dj,Bsi,Bsj;
double QUchi,QOchi,QUchj,QOchj,YYDiqp,YYDjqp;
double YYAsiqp,YYAsjqp,YYBsiqp,YYBsjqp;
double caj,cbj,bij,cfqr,cfqs;
@ -1883,7 +1883,7 @@ void PairComb::qfo_short(Param *param, int i, int j, double rsq,
double rslp,rslp2,rslp4,arr1,arr2,fc2j,fc3j;
qi = iq; qj = jq; r = sqrt(rsq);
Di = Dj = Asi = Asj = bigA = Bsi = Bsj = bigB = 0.0;
Di = Dj = Asi = Asj = Bsi = Bsj = 0.0;
QUchi = QOchi = QUchj = QOchj = YYDiqp = YYDjqp =0.0;
YYAsiqp = YYAsjqp = YYBsiqp = YYBsjqp = 0.0;
caj = cbj = vrcs = cfqr = cfqs = 0.0;

12
src/MANYBODY/pair_comb3.cpp
View File

@ -955,7 +955,7 @@ void PairComb3::compute(int eflag, int vflag)
int sht_jnum,*sht_jlist,sht_lnum,*sht_llist;
int sht_mnum,*sht_mlist,sht_pnum,*sht_plist;
int *ilist,*jlist,*numneigh,**firstneigh,mr1,mr2,mr3,inty,nj;
double xtmp,ytmp,ztmp,delx,dely,delz,evdwl,ecoul,fpair;
double xtmp,ytmp,ztmp,delx,dely,delz,evdwl,fpair;
double rsq,rsq1,rsq2,rsq3,iq,jq,yaself;
double eng_tmp,vionij,fvionij,sr1,sr2,sr3;
double zeta_ij,prefac_ij1,prefac_ij2,prefac_ij3,prefac_ij4,prefac_ij5;
@ -986,7 +986,7 @@ void PairComb3::compute(int eflag, int vflag)
double delrl[3], delrm[3], delrp[3], ddprx[3], srmu;
double zet_addi,zet_addj;
evdwl = ecoul = eng_tmp = 0.0;
evdwl = eng_tmp = 0.0;
if (eflag || vflag) ev_setup(eflag,vflag);
else evflag = vflag_fdotr = vflag_atom = 0;
@ -1953,7 +1953,7 @@ double PairComb3::self(Param *param, double qi)
void PairComb3::comb_fa(double r, Param *parami, Param *paramj, double iq,
double jq, double &att_eng, double &att_force)
{
double bigB,Bsi,FBsi;
double Bsi;
double qi,qj,Di,Dj;
double AlfDiAlfDj, YYBn, YYBj;
double alfij1= parami->alpha1;
@ -1965,7 +1965,7 @@ void PairComb3::comb_fa(double r, Param *parami, Param *paramj, double iq,
if (r > parami->bigr + parami->bigd) Bsi = 0.0;
qi = iq; qj = jq;
Di = Dj = Bsi = FBsi = bigB = 0.0;
Di = Dj = Bsi = 0.0;
Di = parami->DU + pow(fabs(parami->bD*(parami->QU-qi)),parami->nD);
Dj = paramj->DU + pow(fabs(paramj->bD*(paramj->QU-qj)),paramj->nD);
YYBn = (parami->aB-fabs(pow(parami->bB*(qi-parami->Qo),10)));
@ -3265,7 +3265,7 @@ double PairComb3::combqeq(double *qf_fix, int &igroup)
int *ilist,*jlist,*numneigh,**firstneigh;
int mr1,mr2,mr3,inty,nj;
double xtmp,ytmp,ztmp,rsq1,delrj[3];
double iq,jq,fqi,fqj,fqij,fqji,sr1,sr2,sr3;
double iq,jq,fqi,fqij,fqji,sr1,sr2,sr3;
double potal,fac11,fac11e;
int sht_jnum,*sht_jlist;
tagint itag, jtag;
@ -3300,7 +3300,7 @@ double PairComb3::combqeq(double *qf_fix, int &igroup)
// loop over full neighbor list of my atoms
fqi = fqj = fqij = fqji = 0.0;
fqi = fqij = fqji = 0.0;
for (ii = 0; ii < inum; ii ++) {
i = ilist[ii];

5
src/MISC/fix_thermal_conductivity.cpp
View File

@ -144,7 +144,6 @@ void FixThermalConductivity::end_of_step()
{
int i,j,m,insert;
double coord,ke;
MPI_Status status;
struct {
double value;
int proc;
@ -283,7 +282,7 @@ void FixThermalConductivity::end_of_step()
if (rmass) sbuf[3] = rmass[j];
else sbuf[3] = mass[type[j]];
MPI_Sendrecv(sbuf,4,MPI_DOUBLE,all[1].proc,0,
rbuf,4,MPI_DOUBLE,all[1].proc,0,world,&status);
rbuf,4,MPI_DOUBLE,all[1].proc,0,world,MPI_STATUS_IGNORE);
vcm[0] = (sbuf[3]*sbuf[0] + rbuf[3]*rbuf[0]) / (sbuf[3] + rbuf[3]);
vcm[1] = (sbuf[3]*sbuf[1] + rbuf[3]*rbuf[1]) / (sbuf[3] + rbuf[3]);
vcm[2] = (sbuf[3]*sbuf[2] + rbuf[3]*rbuf[2]) / (sbuf[3] + rbuf[3]);
@ -302,7 +301,7 @@ void FixThermalConductivity::end_of_step()
if (rmass) sbuf[3] = rmass[j];
else sbuf[3] = mass[type[j]];
MPI_Sendrecv(sbuf,4,MPI_DOUBLE,all[0].proc,0,
rbuf,4,MPI_DOUBLE,all[0].proc,0,world,&status);
rbuf,4,MPI_DOUBLE,all[0].proc,0,world,MPI_STATUS_IGNORE);
vcm[0] = (sbuf[3]*sbuf[0] + rbuf[3]*rbuf[0]) / (sbuf[3] + rbuf[3]);
vcm[1] = (sbuf[3]*sbuf[1] + rbuf[3]*rbuf[1]) / (sbuf[3] + rbuf[3]);
vcm[2] = (sbuf[3]*sbuf[2] + rbuf[3]*rbuf[2]) / (sbuf[3] + rbuf[3]);

5
src/MISC/fix_viscosity.cpp
View File

@ -154,7 +154,6 @@ void FixViscosity::end_of_step()
{
int i,m,insert;
double coord,delta;
MPI_Status status;
struct {
double value;
int proc;
@ -276,7 +275,7 @@ void FixViscosity::end_of_step()
if (rmass) sbuf[1] = rmass[ipos];
else sbuf[1] = mass[type[ipos]];
MPI_Sendrecv(sbuf,2,MPI_DOUBLE,all[1].proc,0,
rbuf,2,MPI_DOUBLE,all[1].proc,0,world,&status);
rbuf,2,MPI_DOUBLE,all[1].proc,0,world,MPI_STATUS_IGNORE);
vcm = (sbuf[1]*sbuf[0] + rbuf[1]*rbuf[0]) / (sbuf[1] + rbuf[1]);
v[ipos][vdim] = 2.0 * vcm - sbuf[0];
pswap += sbuf[1] * (vcm - sbuf[0]);
@ -287,7 +286,7 @@ void FixViscosity::end_of_step()
if (rmass) sbuf[1] = rmass[ineg];
else sbuf[1] = mass[type[ineg]];
MPI_Sendrecv(sbuf,2,MPI_DOUBLE,all[0].proc,0,
rbuf,2,MPI_DOUBLE,all[0].proc,0,world,&status);
rbuf,2,MPI_DOUBLE,all[0].proc,0,world,MPI_STATUS_IGNORE);
vcm = (sbuf[1]*sbuf[0] + rbuf[1]*rbuf[0]) / (sbuf[1] + rbuf[1]);
v[ineg][vdim] = 2.0 * vcm - sbuf[0];
pswap -= sbuf[1] * (vcm - sbuf[0]);

24
src/MPIIO/dump_atom_mpiio.cpp
View File

@ -275,8 +275,6 @@ void DumpAtomMPIIO::write_header(bigint ndump)
void DumpAtomMPIIO::header_binary(bigint ndump)
{
MPI_Status mpiStatus;
if (performEstimate) {
headerBuffer = (char *) malloc((2*sizeof(bigint)) + (9*sizeof(int)) + (6*sizeof(double)));
@ -320,7 +318,7 @@ void DumpAtomMPIIO::header_binary(bigint ndump)
}
else { // write data
if (me == 0)
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_BYTE,&mpiStatus);
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_BYTE,MPI_STATUS_IGNORE);
mpifo += headerSize;
free(headerBuffer);
}
@ -330,8 +328,6 @@ void DumpAtomMPIIO::header_binary(bigint ndump)
void DumpAtomMPIIO::header_binary_triclinic(bigint ndump)
{
MPI_Status mpiStatus;
if (performEstimate) {
headerBuffer = (char *) malloc((2*sizeof(bigint)) + (9*sizeof(int)) + (6*sizeof(double)));
@ -386,7 +382,7 @@ void DumpAtomMPIIO::header_binary_triclinic(bigint ndump)
else { // write data
if (me == 0)
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_BYTE,&mpiStatus);
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_BYTE,MPI_STATUS_IGNORE);
mpifo += headerSize;
free(headerBuffer);
}
@ -396,8 +392,6 @@ void DumpAtomMPIIO::header_binary_triclinic(bigint ndump)
void DumpAtomMPIIO::header_item(bigint ndump)
{
MPI_Status mpiStatus;
if (performEstimate) {
headerBuffer = (char *) malloc(MAX_TEXT_HEADER_SIZE);
@ -416,7 +410,7 @@ void DumpAtomMPIIO::header_item(bigint ndump)
else { // write data
if (me == 0)
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_CHAR,&mpiStatus);
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_CHAR,MPI_STATUS_IGNORE);
mpifo += headerSize;
free(headerBuffer);
}
@ -426,9 +420,6 @@ void DumpAtomMPIIO::header_item(bigint ndump)
void DumpAtomMPIIO::header_item_triclinic(bigint ndump)
{
MPI_Status mpiStatus;
if (performEstimate) {
headerBuffer = (char *) malloc(MAX_TEXT_HEADER_SIZE);
@ -447,7 +438,7 @@ void DumpAtomMPIIO::header_item_triclinic(bigint ndump)
else { // write data
if (me == 0)
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_CHAR,&mpiStatus);
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_CHAR,MPI_STATUS_IGNORE);
mpifo += headerSize;
free(headerBuffer);
}
@ -465,7 +456,6 @@ void DumpAtomMPIIO::write_data(int n, double *mybuf)
void DumpAtomMPIIO::write_binary(int n, double *mybuf)
{
MPI_Status mpiStatus;
n *= size_one;
if (performEstimate) {
@ -483,7 +473,7 @@ void DumpAtomMPIIO::write_binary(int n, double *mybuf)
memory->create(bufWithSize,byteBufSize,"dump:bufWithSize");
memcpy(bufWithSize,(char*)(&n),sizeof(int));
memcpy(&((char*)bufWithSize)[sizeof(int)],mybuf,(n*sizeof(double)));
MPI_File_write_at_all(mpifh,mpifo+offsetFromHeader,bufWithSize,byteBufSize,MPI_BYTE,&mpiStatus);
MPI_File_write_at_all(mpifh,mpifo+offsetFromHeader,bufWithSize,byteBufSize,MPI_BYTE,MPI_STATUS_IGNORE);
memory->destroy(bufWithSize);
if (flush_flag)
@ -495,8 +485,6 @@ void DumpAtomMPIIO::write_binary(int n, double *mybuf)
void DumpAtomMPIIO::write_string(int n, double *mybuf)
{
MPI_Status mpiStatus;
if (performEstimate) {
#if defined(_OPENMP)
@ -516,7 +504,7 @@ void DumpAtomMPIIO::write_string(int n, double *mybuf)
offsetFromHeader = ((incPrefix-bigintNsme)*sizeof(char));
}
else {
MPI_File_write_at_all(mpifh,mpifo+offsetFromHeader,sbuf,nsme,MPI_CHAR,&mpiStatus);
MPI_File_write_at_all(mpifh,mpifo+offsetFromHeader,sbuf,nsme,MPI_CHAR,MPI_STATUS_IGNORE);
if (flush_flag)
MPI_File_sync(mpifh);
}

9
src/MPIIO/dump_cfg_mpiio.cpp
View File

@ -279,8 +279,6 @@ void DumpCFGMPIIO::write_header(bigint n)
// for unwrapped coords, set to UNWRAPEXPAND (10.0)
// so molecules are not split across periodic box boundaries
MPI_Status mpiStatus;
if (performEstimate) {
headerBuffer = (char *) malloc(MAX_TEXT_HEADER_SIZE);
@ -313,7 +311,7 @@ void DumpCFGMPIIO::write_header(bigint n)
else { // write data
if (me == 0)
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_CHAR,&mpiStatus);
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_CHAR,MPI_STATUS_IGNORE);
mpifo += headerSize;
free(headerBuffer);
}
@ -329,7 +327,6 @@ void DumpCFGMPIIO::write_header(bigint n)
int DumpCFGMPIIO::convert_string_omp(int n, double *mybuf)
{
MPI_Status mpiStatus;
char **mpifh_buffer_line_per_thread;
int mpifhStringCount;
int *mpifhStringCountPerThread, *bufOffset, *bufRange, *bufLength;
@ -483,8 +480,6 @@ void DumpCFGMPIIO::write_data(int n, double *mybuf)
void DumpCFGMPIIO::write_string(int n, double *mybuf)
{
MPI_Status mpiStatus;
if (performEstimate) {
#if defined(_OPENMP)
@ -504,7 +499,7 @@ void DumpCFGMPIIO::write_string(int n, double *mybuf)
offsetFromHeader = ((incPrefix-bigintNsme)*sizeof(char));
}
else {
MPI_File_write_at_all(mpifh,mpifo+offsetFromHeader,sbuf,nsme,MPI_CHAR,&mpiStatus);
MPI_File_write_at_all(mpifh,mpifo+offsetFromHeader,sbuf,nsme,MPI_CHAR,MPI_STATUS_IGNORE);
if (flush_flag)
MPI_File_sync(mpifh);
}

2
src/MPIIO/dump_cfg_mpiio.h
View File

@ -1,4 +1,4 @@
/* ----------------------------------------------------------------------
/* -*- c++ -*- ----------------------------------------------------------
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
http://lammps.sandia.gov, Sandia National Laboratories
Steve Plimpton, sjplimp@sandia.gov

23
src/MPIIO/dump_custom_mpiio.cpp
View File

@ -319,8 +319,6 @@ void DumpCustomMPIIO::write_header(bigint ndump)
void DumpCustomMPIIO::header_binary(bigint ndump)
{
MPI_Status mpiStatus;
if (performEstimate) {
headerBuffer = (char *) malloc((2*sizeof(bigint)) + (9*sizeof(int)) + (6*sizeof(double)));
@ -364,7 +362,7 @@ void DumpCustomMPIIO::header_binary(bigint ndump)
}
else { // write data
if (me == 0)
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_BYTE,&mpiStatus);
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_BYTE,MPI_STATUS_IGNORE);
mpifo += headerSize;
free(headerBuffer);
}
@ -374,8 +372,6 @@ void DumpCustomMPIIO::header_binary(bigint ndump)
void DumpCustomMPIIO::header_binary_triclinic(bigint ndump)
{
MPI_Status mpiStatus;
if (performEstimate) {
headerBuffer = (char *) malloc((2*sizeof(bigint)) + (9*sizeof(int)) + (6*sizeof(double)));
@ -430,7 +426,7 @@ void DumpCustomMPIIO::header_binary_triclinic(bigint ndump)
else { // write data
if (me == 0)
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_BYTE,&mpiStatus);
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_BYTE,MPI_STATUS_IGNORE);
mpifo += headerSize;
free(headerBuffer);
}
@ -440,8 +436,6 @@ void DumpCustomMPIIO::header_binary_triclinic(bigint ndump)
void DumpCustomMPIIO::header_item(bigint ndump)
{
MPI_Status mpiStatus;
if (performEstimate) {
headerBuffer = (char *) malloc(MAX_TEXT_HEADER_SIZE);
@ -460,7 +454,7 @@ void DumpCustomMPIIO::header_item(bigint ndump)
else { // write data
if (me == 0)
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_CHAR,&mpiStatus);
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_CHAR,MPI_STATUS_IGNORE);
mpifo += headerSize;
free(headerBuffer);
}
@ -470,8 +464,6 @@ void DumpCustomMPIIO::header_item(bigint ndump)
void DumpCustomMPIIO::header_item_triclinic(bigint ndump)
{
MPI_Status mpiStatus;
if (performEstimate) {
headerBuffer = (char *) malloc(MAX_TEXT_HEADER_SIZE);
@ -490,7 +482,7 @@ void DumpCustomMPIIO::header_item_triclinic(bigint ndump)
else { // write data
if (me == 0)
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_CHAR,&mpiStatus);
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_CHAR,MPI_STATUS_IGNORE);
mpifo += headerSize;
free(headerBuffer);
}
@ -507,7 +499,6 @@ void DumpCustomMPIIO::write_data(int n, double *mybuf)
void DumpCustomMPIIO::write_binary(int n, double *mybuf)
{
MPI_Status mpiStatus;
n *= size_one;
if (performEstimate) {
@ -525,7 +516,7 @@ void DumpCustomMPIIO::write_binary(int n, double *mybuf)
memory->create(bufWithSize,byteBufSize,"dump:bufWithSize");
memcpy(bufWithSize,(char*)(&n),sizeof(int));
memcpy(&((char*)bufWithSize)[sizeof(int)],mybuf,(n*sizeof(double)));
MPI_File_write_at_all(mpifh,mpifo+offsetFromHeader,bufWithSize,byteBufSize,MPI_BYTE,&mpiStatus);
MPI_File_write_at_all(mpifh,mpifo+offsetFromHeader,bufWithSize,byteBufSize,MPI_BYTE,MPI_STATUS_IGNORE);
memory->destroy(bufWithSize);
if (flush_flag)
@ -537,8 +528,6 @@ void DumpCustomMPIIO::write_binary(int n, double *mybuf)
void DumpCustomMPIIO::write_string(int n, double *mybuf)
{
MPI_Status mpiStatus;
if (performEstimate) {
#if defined(_OPENMP)
@ -558,7 +547,7 @@ void DumpCustomMPIIO::write_string(int n, double *mybuf)
offsetFromHeader = ((incPrefix-bigintNsme)*sizeof(char));
}
else {
MPI_File_write_at_all(mpifh,mpifo+offsetFromHeader,sbuf,nsme,MPI_CHAR,&mpiStatus);
MPI_File_write_at_all(mpifh,mpifo+offsetFromHeader,sbuf,nsme,MPI_CHAR,MPI_STATUS_IGNORE);
if (flush_flag)
MPI_File_sync(mpifh);
}

8
src/MPIIO/dump_xyz_mpiio.cpp
View File

@ -249,8 +249,6 @@ void DumpXYZMPIIO::init_style()
void DumpXYZMPIIO::write_header(bigint n)
{
MPI_Status mpiStatus;
if (performEstimate) {
headerBuffer = (char *) malloc(MAX_TEXT_HEADER_SIZE);
@ -262,7 +260,7 @@ void DumpXYZMPIIO::write_header(bigint n)
else { // write data
if (me == 0)
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_CHAR,&mpiStatus);
MPI_File_write_at(mpifh,mpifo,headerBuffer,headerSize,MPI_CHAR,MPI_STATUS_IGNORE);
mpifo += headerSize;
free(headerBuffer);
}
@ -279,8 +277,6 @@ void DumpXYZMPIIO::write_data(int n, double *mybuf)
void DumpXYZMPIIO::write_string(int n, double *mybuf)
{
MPI_Status mpiStatus;
if (performEstimate) {
#if defined(_OPENMP)
@ -300,7 +296,7 @@ void DumpXYZMPIIO::write_string(int n, double *mybuf)
offsetFromHeader = ((incPrefix-bigintNsme)*sizeof(char));
}
else { // write data
MPI_File_write_at_all(mpifh,mpifo+offsetFromHeader,sbuf,nsme,MPI_CHAR,&mpiStatus);
MPI_File_write_at_all(mpifh,mpifo+offsetFromHeader,sbuf,nsme,MPI_CHAR,MPI_STATUS_IGNORE);
if (flush_flag)
MPI_File_sync(mpifh);
}

9
src/MPIIO/restart_mpiio.cpp
View File

@ -81,7 +81,6 @@ void RestartMPIIO::openForWrite(char *filename)
void RestartMPIIO::write(MPI_Offset headerOffset, int send_size, double *buf)
{
MPI_Status mpiStatus;
bigint incPrefix = 0;
bigint bigintSendSize = (bigint) send_size;
MPI_Scan(&bigintSendSize,&incPrefix,1,MPI_LMP_BIGINT,MPI_SUM,world);
@ -103,7 +102,7 @@ void RestartMPIIO::write(MPI_Offset headerOffset, int send_size, double *buf)
err = MPI_File_write_at_all(mpifh,headerOffset +
((incPrefix-bigintSendSize)*sizeof(double)),
buf,send_size,MPI_DOUBLE,&mpiStatus);
buf,send_size,MPI_DOUBLE,MPI_STATUS_IGNORE);
if (err != MPI_SUCCESS) {
char str[MPI_MAX_ERROR_STRING+128];
char mpiErrorString[MPI_MAX_ERROR_STRING];
@ -126,8 +125,6 @@ void RestartMPIIO::write(MPI_Offset headerOffset, int send_size, double *buf)
void RestartMPIIO::read(MPI_Offset chunkOffset, bigint chunkSize, double *buf)
{
MPI_Status mpiStatus;
int intChunkSize;
bigint remainingSize = 0;
if (chunkSize > INT_MAX) {
@ -137,7 +134,7 @@ void RestartMPIIO::read(MPI_Offset chunkOffset, bigint chunkSize, double *buf)
else intChunkSize = (int) chunkSize;
int err = MPI_File_read_at_all(mpifh,chunkOffset,buf,intChunkSize,
MPI_DOUBLE,&mpiStatus);
MPI_DOUBLE,MPI_STATUS_IGNORE);
if (err != MPI_SUCCESS) {
char str[MPI_MAX_ERROR_STRING+128];
char mpiErrorString[MPI_MAX_ERROR_STRING];
@ -161,7 +158,7 @@ void RestartMPIIO::read(MPI_Offset chunkOffset, bigint chunkSize, double *buf)
remainingSize = 0;
}
int err = MPI_File_read_at(mpifh,currentOffset,&buf[bufOffset],
currentChunkSize,MPI_DOUBLE,&mpiStatus);
currentChunkSize,MPI_DOUBLE,MPI_STATUS_IGNORE);
if (err != MPI_SUCCESS) {
char str[MPI_MAX_ERROR_STRING+128];
char mpiErrorString[MPI_MAX_ERROR_STRING];

2
src/QEQ/fix_qeq.h
View File

@ -1,4 +1,4 @@
/* ----------------------------------------------------------------------
/* -*- c++ -*- ----------------------------------------------------------
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
http://lammps.sandia.gov, Sandia National Laboratories
Steve Plimpton, sjplimp@sandia.gov

6
src/QEQ/fix_qeq_slater.cpp
View File

@ -222,12 +222,12 @@ double FixQEqSlater::calculate_H(double zei, double zej, double zj,
double erfcr = erfc(alpha*r);
double qqrd2e = force->qqrd2e;
double etmp, etmp1, etmp2, etmp3, etmp4;
double etmp1, etmp2;
double e1, e2, e3, e4;
double ci_jfi, ci_fifj;
e1 = e2 = e3 = e4 = 0.0;
etmp = etmp1 = etmp2 = etmp3 = etmp4 = 0.0;
etmp1 = etmp2 = 0.0;
ci_jfi = -zei*exp2zir - rinv*exp2zir;
@ -394,11 +394,9 @@ void FixQEqSlater::sparse_matvec( sparse_matrix *A, double *x, double *b )
{
int i, j, itr_j;
int nn, NN;
int *ilist;
nn = atom->nlocal;
NN = atom->nlocal + atom->nghost;
ilist = list->ilist;
double r = cutoff;
double woself = 0.50*erfc(alpha*r)/r + alpha/MY_PIS;

5
src/REAX/fix_reax_bonds.cpp
View File

@ -112,7 +112,6 @@ void FixReaxBonds::OutputReaxBonds(bigint ntimestep, FILE *fp)
double cutof3;
double *buf;
MPI_Request irequest;
MPI_Status istatus;
MPI_Comm_size(world,&nprocs);
@ -198,7 +197,7 @@ void FixReaxBonds::OutputReaxBonds(bigint ntimestep, FILE *fp)
} else {
MPI_Irecv(&buf[0],nbuf,MPI_DOUBLE,inode,0,world,&irequest);
MPI_Send(&itmp,0,MPI_INT,inode,0,world);
MPI_Wait(&irequest,&istatus);
MPI_Wait(&irequest,MPI_STATUS_IGNORE);
nlocal_tmp = nint(buf[j++]);
}
@ -237,7 +236,7 @@ void FixReaxBonds::OutputReaxBonds(bigint ntimestep, FILE *fp)
}
} else {
MPI_Recv(&itmp,0,MPI_INT,0,0,world,&istatus);
MPI_Recv(&itmp,0,MPI_INT,0,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(&buf[0],nbuf_local,MPI_DOUBLE,0,0,world);
}

9
src/REPLICA/fix_neb.cpp
View File

@ -133,7 +133,6 @@ void FixNEB::min_post_force(int vflag)
double vprev,vnext,vmax,vmin;
double delx,dely,delz;
double delta1[3],delta2[3];
MPI_Status status;
MPI_Request request;
// veng = PE of this replica
@ -142,11 +141,11 @@ void FixNEB::min_post_force(int vflag)
veng = pe->compute_scalar();
if (ireplica < nreplica-1) MPI_Send(&veng,1,MPI_DOUBLE,procnext,0,uworld);
if (ireplica > 0) MPI_Recv(&vprev,1,MPI_DOUBLE,procprev,0,uworld,&status);
if (ireplica > 0) MPI_Recv(&vprev,1,MPI_DOUBLE,procprev,0,uworld,MPI_STATUS_IGNORE);
if (ireplica > 0) MPI_Send(&veng,1,MPI_DOUBLE,procprev,0,uworld);
if (ireplica < nreplica-1)
MPI_Recv(&vnext,1,MPI_DOUBLE,procnext,0,uworld,&status);
MPI_Recv(&vnext,1,MPI_DOUBLE,procnext,0,uworld,MPI_STATUS_IGNORE);
// xprev,xnext = atom coords of adjacent replicas
// assume order of atoms in all replicas is the same
@ -161,13 +160,13 @@ void FixNEB::min_post_force(int vflag)
MPI_Irecv(xprev[0],3*nlocal,MPI_DOUBLE,procprev,0,uworld,&request);
if (ireplica < nreplica-1)
MPI_Send(x[0],3*nlocal,MPI_DOUBLE,procnext,0,uworld);
if (ireplica > 0) MPI_Wait(&request,&status);
if (ireplica > 0) MPI_Wait(&request,MPI_STATUS_IGNORE);
if (ireplica < nreplica-1)
MPI_Irecv(xnext[0],3*nlocal,MPI_DOUBLE,procnext,0,uworld,&request);
if (ireplica > 0)
MPI_Send(x[0],3*nlocal,MPI_DOUBLE,procprev,0,uworld);
if (ireplica < nreplica-1) MPI_Wait(&request,&status);
if (ireplica < nreplica-1) MPI_Wait(&request,MPI_STATUS_IGNORE);
// trigger potential energy computation on next timestep

5
src/REPLICA/temper.cpp
View File

@ -181,7 +181,6 @@ void Temper::command(int narg, char **arg)
int i,which,partner,swap,partner_set_temp,partner_world;
double pe,pe_partner,boltz_factor,new_temp;
MPI_Status status;
if (me_universe == 0 && universe->uscreen)
fprintf(universe->uscreen,"Setting up tempering ...\n");
@ -254,7 +253,7 @@ void Temper::command(int narg, char **arg)
if (me_universe > partner)
MPI_Send(&pe,1,MPI_DOUBLE,partner,0,universe->uworld);
else
MPI_Recv(&pe_partner,1,MPI_DOUBLE,partner,0,universe->uworld,&status);
MPI_Recv(&pe_partner,1,MPI_DOUBLE,partner,0,universe->uworld,MPI_STATUS_IGNORE);
if (me_universe < partner) {
boltz_factor = (pe - pe_partner) *
@ -267,7 +266,7 @@ void Temper::command(int narg, char **arg)
if (me_universe < partner)
MPI_Send(&swap,1,MPI_INT,partner,0,universe->uworld);
else
MPI_Recv(&swap,1,MPI_INT,partner,0,universe->uworld,&status);
MPI_Recv(&swap,1,MPI_INT,partner,0,universe->uworld,MPI_STATUS_IGNORE);
#ifdef TEMPER_DEBUG
if (me_universe < partner)

8
src/REPLICA/verlet_split.cpp
View File

@ -498,8 +498,6 @@ void VerletSplit::rk_setup()
void VerletSplit::r2k_comm()
{
MPI_Status status;
int n = 0;
if (master) n = atom->nlocal;
MPI_Gatherv(atom->x[0],n*3,MPI_DOUBLE,atom->x[0],xsize,xdisp,
@ -513,7 +511,7 @@ void VerletSplit::r2k_comm()
MPI_Send(flags,2,MPI_INT,0,0,block);
} else if (!master) {
int flags[2];
MPI_Recv(flags,2,MPI_DOUBLE,1,0,block,&status);
MPI_Recv(flags,2,MPI_DOUBLE,1,0,block,MPI_STATUS_IGNORE);
eflag = flags[0]; vflag = flags[1];
}
@ -524,8 +522,8 @@ void VerletSplit::r2k_comm()
MPI_Send(domain->boxlo,3,MPI_DOUBLE,0,0,block);
MPI_Send(domain->boxhi,3,MPI_DOUBLE,0,0,block);
} else if (!master) {
MPI_Recv(domain->boxlo,3,MPI_DOUBLE,1,0,block,&status);
MPI_Recv(domain->boxhi,3,MPI_DOUBLE,1,0,block,&status);
MPI_Recv(domain->boxlo,3,MPI_DOUBLE,1,0,block,MPI_STATUS_IGNORE);
MPI_Recv(domain->boxhi,3,MPI_DOUBLE,1,0,block,MPI_STATUS_IGNORE);
domain->set_global_box();
domain->set_local_box();
force->kspace->setup();

6
src/RIGID/fix_rigid_small.cpp
View File

@ -2484,10 +2484,10 @@ void FixRigidSmall::write_restart_file(char *file)
// all other procs wait for ping, send their chunk to proc 0
int tmp,recvrow;
MPI_Status status;
MPI_Request request;
if (me == 0) {
MPI_Status status;
MPI_Request request;
for (int iproc = 0; iproc < nprocs; iproc++) {
if (iproc) {
MPI_Irecv(&buf[0][0],maxrow*ncol,MPI_DOUBLE,iproc,0,world,&request);
@ -2506,7 +2506,7 @@ void FixRigidSmall::write_restart_file(char *file)
}
} else {
MPI_Recv(&tmp,0,MPI_INT,0,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,0,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(&buf[0][0],sendrow*ncol,MPI_DOUBLE,0,0,world);
}

13
src/SNAP/pair_snap.cpp
View File

@ -741,7 +741,6 @@ void PairSNAP::load_balance()
int* procgrid = comm->procgrid;
int nlocal_up,nlocal_down;
MPI_Status status;
MPI_Request request;
double sub_mid[3];
@ -802,9 +801,9 @@ void PairSNAP::load_balance()
// two stage process, first upstream movement, then downstream
MPI_Sendrecv(&nlocal,1,MPI_INT,sendproc,0,
&nlocal_up,1,MPI_INT,recvproc,0,world,&status);
&nlocal_up,1,MPI_INT,recvproc,0,world,MPI_STATUS_IGNORE);
MPI_Sendrecv(&nlocal,1,MPI_INT,recvproc,0,
&nlocal_down,1,MPI_INT,sendproc,0,world,&status);
&nlocal_down,1,MPI_INT,sendproc,0,world,MPI_STATUS_IGNORE);
nsend = 0;
// send upstream
@ -835,7 +834,7 @@ void PairSNAP::load_balance()
if (nlocal < nlocal_down-1) {
nlocal++;
int get_tag = -1;
MPI_Recv(&get_tag,1,MPI_INT,sendproc,0,world,&status);
MPI_Recv(&get_tag,1,MPI_INT,sendproc,0,world,MPI_STATUS_IGNORE);
// if get_tag -1 the other process didnt have local atoms to send
@ -907,9 +906,9 @@ void PairSNAP::load_balance()
// second pass through the grid
MPI_Sendrecv(&nlocal,1,MPI_INT,sendproc,0,
&nlocal_up,1,MPI_INT,recvproc,0,world,&status);
&nlocal_up,1,MPI_INT,recvproc,0,world,MPI_STATUS_IGNORE);
MPI_Sendrecv(&nlocal,1,MPI_INT,recvproc,0,
&nlocal_down,1,MPI_INT,sendproc,0,world,&status);
&nlocal_down,1,MPI_INT,sendproc,0,world,MPI_STATUS_IGNORE);
// send downstream
@ -938,7 +937,7 @@ void PairSNAP::load_balance()
nlocal++;
int get_tag = -1;
MPI_Recv(&get_tag,1,MPI_INT,recvproc,0,world,&status);
MPI_Recv(&get_tag,1,MPI_INT,recvproc,0,world,MPI_STATUS_IGNORE);
if (get_tag >= 0) {
if (ghostinum >= ghostilist_max) {

10
src/SRD/fix_srd.cpp
View File

@ -1061,7 +1061,6 @@ void FixSRD::vbin_comm(int ishift)
{
BinComm *bcomm1,*bcomm2;
MPI_Request request1,request2;
MPI_Status status;
// send/recv bins in both directions in each dimension
// don't send if nsend = 0
@ -1109,11 +1108,11 @@ void FixSRD::vbin_comm(int ishift)
bcomm2->sendproc,0,world);
}
if (bcomm1->nrecv) {
MPI_Wait(&request1,&status);
MPI_Wait(&request1,MPI_STATUS_IGNORE);
vbin_unpack(rbuf1,vbin,bcomm1->nrecv,bcomm1->recvlist);
}
if (bcomm2->nrecv) {
MPI_Wait(&request2,&status);
MPI_Wait(&request2,MPI_STATUS_IGNORE);
vbin_unpack(rbuf2,vbin,bcomm2->nrecv,bcomm2->recvlist);
}
}
@ -1164,7 +1163,6 @@ void FixSRD::xbin_comm(int ishift, int nval)
{
BinComm *bcomm1,*bcomm2;
MPI_Request request1,request2;
MPI_Status status;
// send/recv bins in both directions in each dimension
// don't send if nsend = 0
@ -1212,11 +1210,11 @@ void FixSRD::xbin_comm(int ishift, int nval)
bcomm2->sendproc,0,world);
}
if (bcomm1->nrecv) {
MPI_Wait(&request1,&status);
MPI_Wait(&request1,MPI_STATUS_IGNORE);
xbin_unpack(rbuf1,vbin,bcomm1->nrecv,bcomm1->recvlist,nval);
}
if (bcomm2->nrecv) {
MPI_Wait(&request2,&status);
MPI_Wait(&request2,MPI_STATUS_IGNORE);
xbin_unpack(rbuf2,vbin,bcomm2->nrecv,bcomm2->recvlist,nval);
}
}

8
src/USER-CUDA/atom_vec_angle_cuda.cpp
View File

@ -397,17 +397,9 @@ int AtomVecAngleCuda::unpack_exchange(double *buf)
if(cuda->oncpu)
return AtomVecAngle::unpack_exchange(buf);
double *sublo,*subhi;
int dim=cuda->shared_data.exchange_dim;
if(domain->box_change)
Cuda_AtomVecAngleCuda_Init(&cuda->shared_data);
if (domain->triclinic == 0) {
sublo = domain->sublo;
subhi = domain->subhi;
} else {
sublo = domain->sublo_lamda;
subhi = domain->subhi_lamda;
}
int mfirst=0;
for(int pi=0;pi<(comm->procgrid[dim]>2?2:1);pi++)

9
src/USER-CUDA/atom_vec_atomic_cuda.cpp
View File

@ -352,18 +352,9 @@ int AtomVecAtomicCuda::unpack_exchange(double *buf)
if(cuda->oncpu)
return AtomVecAtomic::unpack_exchange(buf);
double *sublo,*subhi;
int dim=cuda->shared_data.exchange_dim;
if(domain->box_change)
Cuda_AtomVecAtomicCuda_Init(&cuda->shared_data);
if (domain->triclinic == 0) {
sublo = domain->sublo;
subhi = domain->subhi;
} else {
sublo = domain->sublo_lamda;
subhi = domain->subhi_lamda;
}
int mfirst=0;
for(int pi=0;pi<(comm->procgrid[dim]>2?2:1);pi++)

8
src/USER-CUDA/atom_vec_charge_cuda.cpp
View File

@ -355,18 +355,10 @@ int AtomVecChargeCuda::unpack_exchange(double *buf)
{
if(cuda->oncpu)
return AtomVecCharge::unpack_exchange(buf);
double *sublo,*subhi;
int dim=cuda->shared_data.exchange_dim;
if(domain->box_change)
Cuda_AtomVecChargeCuda_Init(&cuda->shared_data);
if (domain->triclinic == 0) {
sublo = domain->sublo;
subhi = domain->subhi;
} else {
sublo = domain->sublo_lamda;
subhi = domain->subhi_lamda;
}
int mfirst=0;
for(int pi=0;pi<(comm->procgrid[dim]>2?2:1);pi++)

8
src/USER-CUDA/atom_vec_full_cuda.cpp
View File

@ -420,17 +420,9 @@ int AtomVecFullCuda::unpack_exchange(double *buf)
if(cuda->oncpu)
return AtomVecFull::unpack_exchange(buf);
double *sublo,*subhi;
int dim=cuda->shared_data.exchange_dim;
if(domain->box_change)
Cuda_AtomVecFullCuda_Init(&cuda->shared_data);
if (domain->triclinic == 0) {
sublo = domain->sublo;
subhi = domain->subhi;
} else {
sublo = domain->sublo_lamda;
subhi = domain->subhi_lamda;
}
int mfirst=0;
for(int pi=0;pi<(comm->procgrid[dim]>2?2:1);pi++)

136
src/USER-CUDA/comm_cuda.cpp
View File

@ -102,8 +102,6 @@ CommCuda::~CommCuda()
void CommCuda::init()
{
int factor = 1;
if(cuda->shared_data.overlap_comm) factor=maxswap;
if(not buf_send)
grow_send(maxsend,0);
if(not buf_recv)
@ -176,16 +174,11 @@ void CommCuda::forward_comm(int mode)
void CommCuda::forward_comm_cuda()
{
static int count=0;
static double kerneltime=0.0;
static double copytime=0.0;
my_times time1,time2,time3;
int n;
MPI_Request request;
MPI_Status status;
AtomVec *avec = atom->avec;
double **x = atom->x;
cuda->shared_data.domain.xy=domain->xy;
cuda->shared_data.domain.xz=domain->xz;
@ -231,7 +224,7 @@ my_gettime(CLOCK_REALTIME,&time2);
//printf("RecvSize: %i SendSize: %i\n",size_forward_recv_now,n);
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
my_gettime(CLOCK_REALTIME,&time3);
cuda->shared_data.cuda_timings.comm_forward_mpi_upper+=
@ -255,7 +248,7 @@ cuda->shared_data.cuda_timings.comm_forward_mpi_lower+=
buf_send,pbc_flag[iswap],pbc[iswap]);
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
avec->unpack_comm_vel(recvnum[iswap],firstrecv[iswap],buf_recv);
}
else
@ -271,7 +264,7 @@ cuda->shared_data.cuda_timings.comm_forward_mpi_lower+=
buf_send,pbc_flag[iswap],pbc[iswap]);
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
avec->unpack_comm(recvnum[iswap],firstrecv[iswap],buf_recv);
}
@ -308,16 +301,11 @@ cuda->shared_data.cuda_timings.comm_forward_mpi_lower+=
void CommCuda::forward_comm_pack_cuda()
{
static int count=0;
static double kerneltime=0.0;
static double copytime=0.0;
my_times time1,time2,time3;
my_times time1,time2;
int n; // initialize comm buffers & exchange memory
MPI_Request request;
MPI_Status status;
AtomVec *avec = atom->avec;
double **x = atom->x;
cuda->shared_data.domain.xy=domain->xy;
cuda->shared_data.domain.xz=domain->xz;
@ -375,7 +363,7 @@ my_gettime(CLOCK_REALTIME,&time2);
cuda->shared_data.comm.buf_send[iswap],pbc_flag[iswap],pbc[iswap]);
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
avec->unpack_comm(recvnum[iswap],firstrecv[iswap],buf_recv);
}
@ -411,15 +399,10 @@ my_gettime(CLOCK_REALTIME,&time2);
void CommCuda::forward_comm_transfer_cuda()
{
static int count=0;
static double kerneltime=0.0;
static double copytime=0.0;
my_times time1,time2,time3;
my_times time1,time2,time3;
int n;
MPI_Request request;
MPI_Status status;
AtomVec *avec = atom->avec;
double **x = atom->x;
cuda->shared_data.domain.xy=domain->xy;
cuda->shared_data.domain.xz=domain->xz;
cuda->shared_data.domain.yz=domain->yz;
@ -464,7 +447,7 @@ my_gettime(CLOCK_REALTIME,&time2);
cuda->shared_data.cuda_timings.comm_forward_download+=
time2.tv_sec-time1.tv_sec+1.0*(time2.tv_nsec-time1.tv_nsec)/1000000000;
MPI_Send(buf_send,cuda->shared_data.comm.send_size[iswap],MPI_DOUBLE,sendproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
//printf("D: %i \n",cuda->shared_data.comm.send_size[iswap]/1024*4);
CudaWrapper_UploadCudaDataAsync((void*) buf_recv,cuda->shared_data.comm.buf_recv_dev[iswap], size_forward_recv_now*sizeof(double),2);
my_gettime(CLOCK_REALTIME,&time1);
@ -498,7 +481,7 @@ my_gettime(CLOCK_REALTIME,&time1);
my_gettime(CLOCK_REALTIME,&time2);
MPI_Send(cuda->shared_data.comm.buf_send[iswap],cuda->shared_data.comm.send_size[iswap],MPI_DOUBLE,sendproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
my_gettime(CLOCK_REALTIME,&time3);
cuda->shared_data.cuda_timings.comm_forward_mpi_upper+=
@ -520,7 +503,7 @@ cuda->shared_data.cuda_timings.comm_forward_mpi_lower+=
buf_send,pbc_flag[iswap],pbc[iswap]);
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
avec->unpack_comm(recvnum[iswap],firstrecv[iswap],buf_recv);
}
@ -549,15 +532,9 @@ cuda->shared_data.cuda_timings.comm_forward_mpi_lower+=
void CommCuda::forward_comm_unpack_cuda()
{
static int count=0;
static double kerneltime=0.0;
static double copytime=0.0;
my_times time1,time2,time3;
int n;
MPI_Request request;
MPI_Status status;
AtomVec *avec = atom->avec;
double **x = atom->x;
cuda->shared_data.domain.xy=domain->xy;
cuda->shared_data.domain.xz=domain->xz;
@ -599,7 +576,7 @@ void CommCuda::forward_comm_unpack_cuda()
buf_send,pbc_flag[iswap],pbc[iswap]);
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
avec->unpack_comm(recvnum[iswap],firstrecv[iswap],buf_recv);
}
@ -636,7 +613,6 @@ void CommCuda::forward_comm_pair(Pair *pair)
int iswap,n;
double *buf;
MPI_Request request;
MPI_Status status;
int nsize = pair->comm_forward;
@ -658,7 +634,7 @@ void CommCuda::forward_comm_pair(Pair *pair)
MPI_Irecv(buf_recv,nrecv,MPI_DOUBLE,recvproc[iswap],0,
world,&request);
MPI_Send(buf_send,nsend,MPI_DOUBLE,sendproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else buf = buf_send;
@ -677,9 +653,7 @@ void CommCuda::reverse_comm()
{
int n;
MPI_Request request;
MPI_Status status;
AtomVec *avec = atom->avec;
double **f = atom->f;
double *buf;
if(not comm_f_only && not avec->cudable) cuda->downloadAll(); //not yet implemented in CUDA but only needed for non standard atom styles
@ -709,7 +683,7 @@ void CommCuda::reverse_comm()
size_reverse_send_now=(size_reverse_send_now+1)*sizeof(F_CFLOAT)/sizeof(double);
MPI_Send(buf,size_reverse_send_now,MPI_DOUBLE,
recvproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
Cuda_CommCuda_UnpackReverse(&cuda->shared_data,sendnum[iswap],iswap,buf_recv);
} else {
@ -717,7 +691,7 @@ void CommCuda::reverse_comm()
sendproc[iswap],0,world,&request);
n = avec->pack_reverse(recvnum[iswap],firstrecv[iswap],buf_send);
MPI_Send(buf_send,n,MPI_DOUBLE,recvproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
avec->unpack_reverse(sendnum[iswap],sendlist[iswap],buf_recv);
}
@ -761,14 +735,11 @@ void CommCuda::exchange()
void CommCuda::exchange_cuda()
{
int i,m,nsend,nrecv,nrecv1,nrecv2,nlocal;
double lo,hi,value;
double **x;
double *sublo,*subhi,*buf;
int nsend,nrecv,nrecv1,nrecv2,nlocal;
double *buf;
MPI_Request request;
MPI_Status status;
AtomVec *avec = atom->avec;
my_times time1,time2,time3;
my_times time1,time2;
// clear global->local map for owned and ghost atoms
// b/c atoms migrate to new procs in exchange() and
@ -780,23 +751,13 @@ void CommCuda::exchange_cuda()
if (map_style) atom->map_clear();
// subbox bounds for orthogonal or triclinic
if (triclinic == 0) {
sublo = domain->sublo;
subhi = domain->subhi;
} else {
sublo = domain->sublo_lamda;
subhi = domain->subhi_lamda;
}
// loop over dimensions
for (int dim = 0; dim < 3; dim++) {
// fill buffer with atoms leaving my box, using < and >=
// when atom is deleted, fill it in with last atom
cuda->shared_data.exchange_dim=dim;
cuda->shared_data.exchange_dim=dim;
nlocal = atom->nlocal;
avec->maxsend=&maxsend;
@ -819,11 +780,11 @@ void CommCuda::exchange_cuda()
} else {
MPI_Sendrecv(&nsend,1,MPI_INT,procneigh[dim][0],0,
&nrecv1,1,MPI_INT,procneigh[dim][1],0,world,&status);
&nrecv1,1,MPI_INT,procneigh[dim][1],0,world,MPI_STATUS_IGNORE);
nrecv = nrecv1;
if (procgrid[dim] > 2) {
MPI_Sendrecv(&nsend,1,MPI_INT,procneigh[dim][1],0,
&nrecv2,1,MPI_INT,procneigh[dim][0],0,world,&status);
&nrecv2,1,MPI_INT,procneigh[dim][0],0,world,MPI_STATUS_IGNORE);
nrecv += nrecv2;
}
if (nrecv+1 > maxrecv) grow_recv(nrecv+1);
@ -831,13 +792,13 @@ void CommCuda::exchange_cuda()
MPI_Irecv(buf_recv,nrecv1,MPI_DOUBLE,procneigh[dim][1],0,
world,&request);
MPI_Send(buf_send,nsend,MPI_DOUBLE,procneigh[dim][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
if (procgrid[dim] > 2) {
MPI_Irecv(&buf_recv[nrecv1],nrecv2,MPI_DOUBLE,procneigh[dim][0],0,
world,&request);
MPI_Send(buf_send,nsend,MPI_DOUBLE,procneigh[dim][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
if((nrecv1==0)||(nrecv2==0)) buf_recv[nrecv]=0;
}
@ -899,16 +860,12 @@ void CommCuda::borders()
void CommCuda::borders_cuda()
{
int i,n,itype,iswap,dim,ineed,twoneed,smax,rmax;
int nsend,nrecv,nfirst,nlast,ngroup;
double lo,hi;
int *type;
double **x;
double *buf,*mlo,*mhi;
int n,iswap,dim,ineed,twoneed,smax,rmax;
int nsend,nrecv,nfirst,nlast;
double *buf;
MPI_Request request;
MPI_Status status;
AtomVec *avec = atom->avec;
my_times time1,time2,time3;
my_times time1,time2;
// clear old ghosts
@ -931,15 +888,6 @@ void CommCuda::borders_cuda()
// for later swaps in a dim, only check newly arrived ghosts
// store sent atom indices in list for use in future timesteps
x = atom->x;
if (style == SINGLE) {
lo = slablo[iswap];
hi = slabhi[iswap];
} else {
type = atom->type;
mlo = multilo[iswap];
mhi = multihi[iswap];
}
if (ineed % 2 == 0) {
nfirst = nlast;
nlast = atom->nlocal + atom->nghost;
@ -975,13 +923,13 @@ void CommCuda::borders_cuda()
my_gettime(CLOCK_REALTIME,&time1);
if (sendproc[iswap] != me) {
MPI_Sendrecv(&nsend,1,MPI_INT,sendproc[iswap],0,
&nrecv,1,MPI_INT,recvproc[iswap],0,world,&status);
&nrecv,1,MPI_INT,recvproc[iswap],0,world,MPI_STATUS_IGNORE);
if (nrecv*size_border > maxrecv)
grow_recv(nrecv*size_border);
MPI_Irecv(buf_recv,nrecv*size_border,MPI_DOUBLE,
recvproc[iswap],0,world,&request);
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else {
nrecv = nsend;
@ -1034,16 +982,12 @@ cuda->shared_data.cuda_timings.comm_border_mpi+=
void CommCuda::borders_cuda_overlap_forward_comm()
{
int i,n,itype,iswap,dim,ineed,twoneed,smax,rmax;
int nsend,nrecv,nfirst,nlast,ngroup;
double lo,hi;
int *type;
double **x;
double *buf,*mlo,*mhi;
int n,iswap,dim,ineed,twoneed,smax,rmax;
int nsend,nrecv,nfirst,nlast;
double *buf;
MPI_Request request;
MPI_Status status;
AtomVec *avec = atom->avec;
my_times time1,time2,time3;
my_times time1,time2;
// clear old ghosts
@ -1066,15 +1010,6 @@ void CommCuda::borders_cuda_overlap_forward_comm()
// for later swaps in a dim, only check newly arrived ghosts
// store sent atom indices in list for use in future timesteps
x = atom->x;
if (style == SINGLE) {
lo = slablo[iswap];
hi = slabhi[iswap];
} else {
type = atom->type;
mlo = multilo[iswap];
mhi = multihi[iswap];
}
if (ineed % 2 == 0) {
nfirst = nlast;
nlast = atom->nlocal + atom->nghost;
@ -1111,13 +1046,13 @@ void CommCuda::borders_cuda_overlap_forward_comm()
my_gettime(CLOCK_REALTIME,&time1);
if (sendproc[iswap] != me) {
MPI_Sendrecv(&nsend,1,MPI_INT,sendproc[iswap],0,
&nrecv,1,MPI_INT,recvproc[iswap],0,world,&status);
&nrecv,1,MPI_INT,recvproc[iswap],0,world,MPI_STATUS_IGNORE);
if (nrecv*size_border > maxrecv)
grow_recv(nrecv*size_border);
MPI_Irecv(buf_recv,nrecv*size_border,MPI_DOUBLE,
recvproc[iswap],0,world,&request);
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else {
nrecv = nsend;
@ -1171,12 +1106,11 @@ cuda->shared_data.cuda_timings.comm_border_mpi+=
void CommCuda::forward_comm_fix(Fix *fix)
void CommCuda::forward_comm_fix(Fix *fix, int size)
{
int iswap,n;
double *buf;
MPI_Request request;
MPI_Status status;
int nsize = fix->comm_forward;
@ -1206,7 +1140,7 @@ void CommCuda::forward_comm_fix(Fix *fix)
MPI_Irecv(buf_recv,nsize*recvnum[iswap],MPI_DOUBLE,recvproc[iswap],0,
world,&request);
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else buf = buf_send;

2
src/USER-CUDA/comm_cuda.h
View File

@ -40,7 +40,7 @@ public:
virtual void borders(); // setup list of atoms to communicate
virtual void borders_cuda(); // setup list of atoms to communicate
virtual void borders_cuda_overlap_forward_comm();
virtual void forward_comm_fix(class Fix *); // forward comm from a Fix
virtual void forward_comm_fix(class Fix *, int size=0); // forward comm from a Fix

53
src/USER-CUDA/cuda.cpp
View File

@ -63,29 +63,55 @@ Cuda::Cuda(LAMMPS* lmp) : Pointers(lmp)
Cuda_Cuda_GetCompileSettings(&shared_data);
if(shared_data.compile_settings.prec_glob != sizeof(CUDA_CFLOAT) / 4) printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n # CUDA WARNING: Global Precision: cuda %i cpp %i\n\n", shared_data.compile_settings.prec_glob, sizeof(CUDA_CFLOAT) / 4);
if (universe->me == 0) {
if(shared_data.compile_settings.prec_x != sizeof(X_CFLOAT) / 4) printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n # CUDA WARNING: X Precision: cuda %i cpp %i\n\n", shared_data.compile_settings.prec_x, sizeof(X_CFLOAT) / 4);
if(shared_data.compile_settings.prec_glob != sizeof(CUDA_CFLOAT) / 4)
printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n"
" # CUDA WARNING: Global Precision: cuda %i cpp %i\n\n",
shared_data.compile_settings.prec_glob, (int) sizeof(CUDA_CFLOAT) / 4);
if(shared_data.compile_settings.prec_v != sizeof(V_CFLOAT) / 4) printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n # CUDA WARNING: V Precision: cuda %i cpp %i\n\n", shared_data.compile_settings.prec_v, sizeof(V_CFLOAT) / 4);
if(shared_data.compile_settings.prec_x != sizeof(X_CFLOAT) / 4)
printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n"
" # CUDA WARNING: X Precision: cuda %i cpp %i\n\n",
shared_data.compile_settings.prec_x, (int) sizeof(X_CFLOAT) / 4);
if(shared_data.compile_settings.prec_f != sizeof(F_CFLOAT) / 4) printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n # CUDA WARNING: F Precision: cuda %i cpp %i\n\n", shared_data.compile_settings.prec_f, sizeof(F_CFLOAT) / 4);
if(shared_data.compile_settings.prec_v != sizeof(V_CFLOAT) / 4)
printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n"
" # CUDA WARNING: V Precision: cuda %i cpp %i\n\n",
shared_data.compile_settings.prec_v, (int) sizeof(V_CFLOAT) / 4);
if(shared_data.compile_settings.prec_pppm != sizeof(PPPM_CFLOAT) / 4) printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n # CUDA WARNING: PPPM Precision: cuda %i cpp %i\n\n", shared_data.compile_settings.prec_pppm, sizeof(PPPM_CFLOAT) / 4);
if(shared_data.compile_settings.prec_f != sizeof(F_CFLOAT) / 4)
printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n"
" # CUDA WARNING: F Precision: cuda %i cpp %i\n\n",
shared_data.compile_settings.prec_f, (int) sizeof(F_CFLOAT) / 4);
if(shared_data.compile_settings.prec_fft != sizeof(FFT_CFLOAT) / 4) printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n # CUDA WARNING: FFT Precision: cuda %i cpp %i\n\n", shared_data.compile_settings.prec_fft, sizeof(FFT_CFLOAT) / 4);
if(shared_data.compile_settings.prec_pppm != sizeof(PPPM_CFLOAT) / 4)
printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n"
" # CUDA WARNING: PPPM Precision: cuda %i cpp %i\n\n",
shared_data.compile_settings.prec_pppm, (int) sizeof(PPPM_CFLOAT) / 4);
if(shared_data.compile_settings.prec_fft != sizeof(FFT_CFLOAT) / 4)
printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n"
" # CUDA WARNING: FFT Precision: cuda %i cpp %i\n\n",
shared_data.compile_settings.prec_fft, (int) sizeof(FFT_CFLOAT) / 4);
#ifdef FFT_CUFFT
if(shared_data.compile_settings.cufft != 1) printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n # CUDA WARNING: cufft: cuda %i cpp %i\n\n", shared_data.compile_settings.cufft, 1);
if(shared_data.compile_settings.cufft != 1)
printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n"
" # CUDA WARNING: cufft: cuda %i cpp %i\n\n",
shared_data.compile_settings.cufft, 1);
#else
if(shared_data.compile_settings.cufft != 0) printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n # CUDA WARNING: cufft: cuda %i cpp %i\n\n", shared_data.compile_settings.cufft, 0);
if(shared_data.compile_settings.cufft != 0)
printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n"
" # CUDA WARNING: cufft: cuda %i cpp %i\n\n",
shared_data.compile_settings.cufft, 0);
#endif
if(shared_data.compile_settings.arch != CUDA_ARCH) printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n # CUDA WARNING: arch: cuda %i cpp %i\n\n", shared_data.compile_settings.cufft, CUDA_ARCH);
if(shared_data.compile_settings.arch != CUDA_ARCH)
printf("\n\n # CUDA WARNING: Compile Settings of cuda and cpp code differ! \n"
" # CUDA WARNING: arch: cuda %i cpp %i\n\n",
shared_data.compile_settings.cufft, CUDA_ARCH);
}
cu_x = 0;
cu_v = 0;
@ -454,7 +480,6 @@ void Cuda::checkResize()
{
MYDBG(printf("# CUDA: Cuda::checkResize ...\n");)
cuda_shared_atom* cu_atom = & shared_data.atom;
cuda_shared_pair* cu_pair = & shared_data.pair;
cu_atom->q_flag = atom->q_flag;
cu_atom->rmass_flag = atom->rmass ? 1 : 0;
cu_atom->nall = atom->nlocal + atom->nghost;

4
src/USER-CUDA/domain_cuda.h
View File

@ -29,8 +29,8 @@ class DomainCuda : public Domain {
void reset_box();
void pbc();
void lamda2x(int);
void x2lamda(int);
virtual void lamda2x(int);
virtual void x2lamda(int);
protected:
class Cuda *cuda;

4
src/USER-CUDA/fix_aveforce_cuda.cpp
View File

@ -170,10 +170,6 @@ void FixAveForceCuda::init()
if (xstyle == EQUAL || ystyle == EQUAL || zstyle == EQUAL) varflag = EQUAL;
else varflag = CONSTANT;
// ncount = total # of atoms in group
int *mask = atom->mask;
int nlocal = atom->nlocal;
}
/* ---------------------------------------------------------------------- */

5
src/USER-CUDA/fix_nh_cuda.cpp
View File

@ -833,8 +833,6 @@ void FixNHCuda::final_integrate()
void FixNHCuda::initial_integrate_respa(int vflag, int ilevel, int iloop)
{
int i;
// set timesteps by level
dtv = step_respa[ilevel];
@ -1430,7 +1428,6 @@ double FixNHCuda::compute_vector(int n)
}
}
int i;
double volume;
double kt = boltz * t_target;
double lkt = tdof * kt;
@ -1653,7 +1650,7 @@ void FixNHCuda::nhc_temp_integrate()
void FixNHCuda::nhc_press_integrate()
{
int ich,i;
double expfac,factor_etap,wmass,kecurrent;
double expfac,factor_etap,kecurrent;
double kt = boltz * t_target;
double lkt_press = kt;

8
src/USER-CUDA/neigh_full_cuda.cpp
View File

@ -63,8 +63,6 @@ void NeighborCuda::full_bin_cuda(NeighList *list)
}
int bin_dim_tmp[3];
int bin_nmax_tmp;
//printf("Hallo\n");
timespec starttime,endtime;
do
{
do
@ -98,8 +96,6 @@ void NeighborCuda::full_bin_cuda(NeighList *list)
// cuda->cu_debugdata->memset_device(0);
int maxneighbors=slist->maxneighbors;
int *ilist = list->ilist;
int *numneigh = list->numneigh;
if((nex_type!=slist->nex_type)||
(nex_group!=slist->nex_group)||
@ -136,12 +132,8 @@ void NeighborCuda::full_bin_cuda(NeighList *list)
//printf("C %i %i %i\n",nex_type,nex_group,nex_mol);
}
int overflow = 0;
int inum = 0;
int npnt = 0;
do
{
npnt=0;
inum=0;
overflow=0;
clist->grow_device();
slist->cutneighsq=cutneighsq;

4
src/USER-CUDA/pair_eam_alloy_cuda.cpp
View File

@ -154,8 +154,8 @@ void PairEAMAlloyCuda::read_file(char *filename)
char **words = new char*[file->nelements+1];
nwords = 0;
char *first = strtok(line," \t\n\r\f");
while (words[nwords++] = strtok(NULL," \t\n\r\f")) continue;
strtok(line," \t\n\r\f");
while ((words[nwords++] = strtok(NULL," \t\n\r\f"))) continue;
file->elements = new char*[file->nelements];
for (int i = 0; i < file->nelements; i++) {

4
src/USER-CUDA/pair_eam_fs_cuda.cpp
View File

@ -154,8 +154,8 @@ void PairEAMFSCuda::read_file(char *filename)
char **words = new char*[file->nelements+1];
nwords = 0;
char *first = strtok(line," \t\n\r\f");
while (words[nwords++] = strtok(NULL," \t\n\r\f")) continue;
strtok(line," \t\n\r\f");
while ((words[nwords++] = strtok(NULL," \t\n\r\f"))) continue;
file->elements = new char*[file->nelements];
for (int i = 0; i < file->nelements; i++) {

6
src/USER-CUDA/pair_tersoff_zbl_cuda.cpp
View File

@ -105,7 +105,7 @@ void PairTersoffZBLCuda::read_file(char *file)
// strip comment, skip line if blank
if (ptr = strchr(line,'#')) *ptr = '\0';
if ((ptr = strchr(line,'#'))) *ptr = '\0';
nwords = atom->count_words(line);
if (nwords == 0) continue;
@ -124,7 +124,7 @@ void PairTersoffZBLCuda::read_file(char *file)
if (eof) break;
MPI_Bcast(&n,1,MPI_INT,0,world);
MPI_Bcast(line,n,MPI_CHAR,0,world);
if (ptr = strchr(line,'#')) *ptr = '\0';
if ((ptr = strchr(line,'#'))) *ptr = '\0';
nwords = atom->count_words(line);
}
@ -135,7 +135,7 @@ void PairTersoffZBLCuda::read_file(char *file)
nwords = 0;
words[nwords++] = strtok(line," \t\n\r\f");
while (words[nwords++] = strtok(NULL," \t\n\r\f")) continue;
while ((words[nwords++] = strtok(NULL," \t\n\r\f"))) continue;
// ielement,jelement,kelement = 1st args
// if all 3 args are in element list, then parse this line

19
src/USER-CUDA/pppm_cuda.cpp
View File

@ -411,48 +411,47 @@ void PPPMCuda::init()
// if no neighbor proc, value is from self since I have ghosts regardless
int nplanes;
MPI_Status status;
nplanes = nxlo_in - nxlo_out;
if (comm->procneigh[0][0] != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,comm->procneigh[0][0],0,
&nxhi_ghost,1,MPI_INT,comm->procneigh[0][1],0,
world,&status);
world,MPI_STATUS_IGNORE);
else nxhi_ghost = nplanes;
nplanes = nxhi_out - nxhi_in;
if (comm->procneigh[0][1] != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,comm->procneigh[0][1],0,
&nxlo_ghost,1,MPI_INT,comm->procneigh[0][0],
0,world,&status);
0,world,MPI_STATUS_IGNORE);
else nxlo_ghost = nplanes;
nplanes = nylo_in - nylo_out;
if (comm->procneigh[1][0] != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,comm->procneigh[1][0],0,
&nyhi_ghost,1,MPI_INT,comm->procneigh[1][1],0,
world,&status);
world,MPI_STATUS_IGNORE);
else nyhi_ghost = nplanes;
nplanes = nyhi_out - nyhi_in;
if (comm->procneigh[1][1] != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,comm->procneigh[1][1],0,
&nylo_ghost,1,MPI_INT,comm->procneigh[1][0],0,
world,&status);
world,MPI_STATUS_IGNORE);
else nylo_ghost = nplanes;
nplanes = nzlo_in - nzlo_out;
if (comm->procneigh[2][0] != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,comm->procneigh[2][0],0,
&nzhi_ghost,1,MPI_INT,comm->procneigh[2][1],0,
world,&status);
world,MPI_STATUS_IGNORE);
else nzhi_ghost = nplanes;
nplanes = nzhi_out - nzhi_in;
if (comm->procneigh[2][1] != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,comm->procneigh[2][1],0,
&nzlo_ghost,1,MPI_INT,comm->procneigh[2][0],0,
world,&status);
world,MPI_STATUS_IGNORE);
else nzlo_ghost = nplanes;
// test that ghost overlap is not bigger than my sub-domain
@ -1291,7 +1290,7 @@ void PPPMCuda::poisson(int eflag, int vflag)
{
#ifndef FFT_CUFFT
PPPMOld::poisson();
PPPMOld::poisson(eflag,vflag);
return;
#endif
#ifdef FFT_CUFFT
@ -1413,10 +1412,6 @@ void PPPMCuda::slabcorr(int eflag)
double dipole_all;
MPI_Allreduce(&dipole,&dipole_all,1,MPI_DOUBLE,MPI_SUM,world);
// compute corrections
double e_slabcorr = 2.0*MY_PI*dipole_all*dipole_all/volume;
//if (eflag) energy += qqrd2e*scale * e_slabcorr;
// need to add a correction to make non-neutral systems and per-atom energy translationally invariant
if (eflag || fabs(qsum) > SMALL)

2
src/USER-CUDA/pppm_cuda.h
View File

@ -98,7 +98,7 @@ class PPPMCuda : public PPPMOld {
virtual void particle_map();
virtual void make_rho();
void poisson(int, int);
virtual void poisson(int, int);
virtual void fieldforce();
virtual void slabcorr(int);
double*** vdx_brick_tmp;

65
src/USER-CUDA/pppm_old.cpp
View File

@ -332,48 +332,47 @@ void PPPMOld::init()
// if no neighbor proc, value is from self since I have ghosts regardless
int nplanes;
MPI_Status status;
nplanes = nxlo_in - nxlo_out;
if (comm->procneigh[0][0] != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,comm->procneigh[0][0],0,
&nxhi_ghost,1,MPI_INT,comm->procneigh[0][1],0,
world,&status);
world,MPI_STATUS_IGNORE);
else nxhi_ghost = nplanes;
nplanes = nxhi_out - nxhi_in;
if (comm->procneigh[0][1] != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,comm->procneigh[0][1],0,
&nxlo_ghost,1,MPI_INT,comm->procneigh[0][0],
0,world,&status);
0,world,MPI_STATUS_IGNORE);
else nxlo_ghost = nplanes;
nplanes = nylo_in - nylo_out;
if (comm->procneigh[1][0] != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,comm->procneigh[1][0],0,
&nyhi_ghost,1,MPI_INT,comm->procneigh[1][1],0,
world,&status);
world,MPI_STATUS_IGNORE);
else nyhi_ghost = nplanes;
nplanes = nyhi_out - nyhi_in;
if (comm->procneigh[1][1] != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,comm->procneigh[1][1],0,
&nylo_ghost,1,MPI_INT,comm->procneigh[1][0],0,
world,&status);
world,MPI_STATUS_IGNORE);
else nylo_ghost = nplanes;
nplanes = nzlo_in - nzlo_out;
if (comm->procneigh[2][0] != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,comm->procneigh[2][0],0,
&nzhi_ghost,1,MPI_INT,comm->procneigh[2][1],0,
world,&status);
world,MPI_STATUS_IGNORE);
else nzhi_ghost = nplanes;
nplanes = nzhi_out - nzhi_in;
if (comm->procneigh[2][1] != me)
MPI_Sendrecv(&nplanes,1,MPI_INT,comm->procneigh[2][1],0,
&nzlo_ghost,1,MPI_INT,comm->procneigh[2][0],0,
world,&status);
world,MPI_STATUS_IGNORE);
else nzlo_ghost = nplanes;
// test that ghost overlap is not bigger than my sub-domain
@ -704,7 +703,7 @@ void PPPMOld::compute(int eflag, int vflag)
// return gradients (electric fields) in 3d brick decomposition
// also performs per-atom calculations via poisson_peratom()
poisson();
poisson(eflag,vflag);
// all procs communicate E-field values
// to fill ghost cells surrounding their 3d bricks
@ -1198,7 +1197,6 @@ void PPPMOld::brick2fft()
{
int i,n,ix,iy,iz;
MPI_Request request;
MPI_Status status;
// pack my ghosts for +x processor
// pass data to self or +x processor
@ -1215,7 +1213,7 @@ void PPPMOld::brick2fft()
else {
MPI_Irecv(buf2,nbuf,MPI_FFT_SCALAR,comm->procneigh[0][0],0,world,&request);
MPI_Send(buf1,n,MPI_FFT_SCALAR,comm->procneigh[0][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1239,7 +1237,7 @@ void PPPMOld::brick2fft()
else {
MPI_Irecv(buf2,nbuf,MPI_FFT_SCALAR,comm->procneigh[0][1],0,world,&request);
MPI_Send(buf1,n,MPI_FFT_SCALAR,comm->procneigh[0][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1263,7 +1261,7 @@ void PPPMOld::brick2fft()
else {
MPI_Irecv(buf2,nbuf,MPI_FFT_SCALAR,comm->procneigh[1][0],0,world,&request);
MPI_Send(buf1,n,MPI_FFT_SCALAR,comm->procneigh[1][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1287,7 +1285,7 @@ void PPPMOld::brick2fft()
else {
MPI_Irecv(buf2,nbuf,MPI_FFT_SCALAR,comm->procneigh[1][1],0,world,&request);
MPI_Send(buf1,n,MPI_FFT_SCALAR,comm->procneigh[1][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1311,7 +1309,7 @@ void PPPMOld::brick2fft()
else {
MPI_Irecv(buf2,nbuf,MPI_FFT_SCALAR,comm->procneigh[2][0],0,world,&request);
MPI_Send(buf1,n,MPI_FFT_SCALAR,comm->procneigh[2][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1335,7 +1333,7 @@ void PPPMOld::brick2fft()
else {
MPI_Irecv(buf2,nbuf,MPI_FFT_SCALAR,comm->procneigh[2][1],0,world,&request);
MPI_Send(buf1,n,MPI_FFT_SCALAR,comm->procneigh[2][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1366,7 +1364,6 @@ void PPPMOld::fillbrick()
{
int i,n,ix,iy,iz;
MPI_Request request;
MPI_Status status;
// pack my real cells for +z processor
// pass data to self or +z processor
@ -1386,7 +1383,7 @@ void PPPMOld::fillbrick()
else {
MPI_Irecv(buf2,nbuf,MPI_FFT_SCALAR,comm->procneigh[2][0],0,world,&request);
MPI_Send(buf1,n,MPI_FFT_SCALAR,comm->procneigh[2][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1416,7 +1413,7 @@ void PPPMOld::fillbrick()
else {
MPI_Irecv(buf2,nbuf,MPI_FFT_SCALAR,comm->procneigh[2][1],0,world,&request);
MPI_Send(buf1,n,MPI_FFT_SCALAR,comm->procneigh[2][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1446,7 +1443,7 @@ void PPPMOld::fillbrick()
else {
MPI_Irecv(buf2,nbuf,MPI_FFT_SCALAR,comm->procneigh[1][0],0,world,&request);
MPI_Send(buf1,n,MPI_FFT_SCALAR,comm->procneigh[1][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1476,7 +1473,7 @@ void PPPMOld::fillbrick()
else {
MPI_Irecv(buf2,nbuf,MPI_FFT_SCALAR,comm->procneigh[1][1],0,world,&request);
MPI_Send(buf1,n,MPI_FFT_SCALAR,comm->procneigh[1][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1506,7 +1503,7 @@ void PPPMOld::fillbrick()
else {
MPI_Irecv(buf2,nbuf,MPI_FFT_SCALAR,comm->procneigh[0][0],0,world,&request);
MPI_Send(buf1,n,MPI_FFT_SCALAR,comm->procneigh[0][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1536,7 +1533,7 @@ void PPPMOld::fillbrick()
else {
MPI_Irecv(buf2,nbuf,MPI_FFT_SCALAR,comm->procneigh[0][1],0,world,&request);
MPI_Send(buf1,n,MPI_FFT_SCALAR,comm->procneigh[0][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1557,7 +1554,6 @@ void PPPMOld::fillbrick_peratom()
{
int i,n,ix,iy,iz;
MPI_Request request;
MPI_Status status;
// pack my real cells for +z processor
// pass data to self or +z processor
@ -1584,7 +1580,7 @@ void PPPMOld::fillbrick_peratom()
MPI_Irecv(buf4,nbuf_peratom,MPI_FFT_SCALAR,
comm->procneigh[2][0],0,world,&request);
MPI_Send(buf3,n,MPI_FFT_SCALAR,comm->procneigh[2][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1627,7 +1623,7 @@ void PPPMOld::fillbrick_peratom()
MPI_Irecv(buf4,nbuf_peratom,MPI_FFT_SCALAR,
comm->procneigh[2][1],0,world,&request);
MPI_Send(buf3,n,MPI_FFT_SCALAR,comm->procneigh[2][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1670,7 +1666,7 @@ void PPPMOld::fillbrick_peratom()
MPI_Irecv(buf4,nbuf_peratom,MPI_FFT_SCALAR,
comm->procneigh[1][0],0,world,&request);
MPI_Send(buf3,n,MPI_FFT_SCALAR,comm->procneigh[1][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1713,7 +1709,7 @@ void PPPMOld::fillbrick_peratom()
MPI_Irecv(buf4,nbuf_peratom,MPI_FFT_SCALAR,
comm->procneigh[1][1],0,world,&request);
MPI_Send(buf3,n,MPI_FFT_SCALAR,comm->procneigh[1][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1756,7 +1752,7 @@ void PPPMOld::fillbrick_peratom()
MPI_Irecv(buf4,nbuf_peratom,MPI_FFT_SCALAR,
comm->procneigh[0][0],0,world,&request);
MPI_Send(buf3,n,MPI_FFT_SCALAR,comm->procneigh[0][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1799,7 +1795,7 @@ void PPPMOld::fillbrick_peratom()
MPI_Irecv(buf4,nbuf_peratom,MPI_FFT_SCALAR,
comm->procneigh[0][1],0,world,&request);
MPI_Send(buf3,n,MPI_FFT_SCALAR,comm->procneigh[0][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
n = 0;
@ -1914,7 +1910,7 @@ void PPPMOld::make_rho()
FFT-based Poisson solver
------------------------------------------------------------------------- */
void PPPMOld::poisson()
void PPPMOld::poisson(int,int)
{
int i,j,k,n;
double eng;
@ -2245,7 +2241,6 @@ void PPPMOld::fieldforce_peratom()
double *q = atom->q;
double **x = atom->x;
double **f = atom->f;
int nlocal = atom->nlocal;
@ -2578,7 +2573,7 @@ void PPPMOld::compute_group_group(int groupbit_A, int groupbit_B, int BA_flag)
error->all(FLERR,"Cannot (yet) use K-space slab "
"correction with compute group/group");
int i,j;
int i;
if (!group_allocate_flag) {
allocate_groups();
@ -2590,11 +2585,6 @@ void PPPMOld::compute_group_group(int groupbit_A, int groupbit_B, int BA_flag)
f2group[1] = 0; //force in y-direction
f2group[2] = 0; //force in z-direction
double *q = atom->q;
int nlocal = atom->nlocal;
int *mask = atom->mask;
// map my particle charge onto my local 3d density grid
make_rho_groups(groupbit_A,groupbit_B,BA_flag);
@ -2758,7 +2748,6 @@ void PPPMOld::make_rho_groups(int groupbit_A, int groupbit_B, int BA_flag)
void PPPMOld::poisson_groups(int BA_flag)
{
int i,j,k,n;
double eng;
// reuse memory (already declared)

2
src/USER-CUDA/pppm_old.h
View File

@ -116,7 +116,7 @@ class PPPMOld : public KSpace {
virtual void brick2fft();
virtual void fillbrick();
virtual void fillbrick_peratom();
virtual void poisson();
virtual void poisson(int,int);
virtual void poisson_peratom();
virtual void fieldforce();
virtual void fieldforce_peratom();

61
src/USER-CUDA/verlet_cuda.cpp
View File

@ -105,8 +105,6 @@ void VerletCuda::setup()
atom->setup();
cuda_shared_atom* cu_atom = & cuda->shared_data.atom;
cuda_shared_domain* cu_domain = & cuda->shared_data.domain;
cuda_shared_pair* cu_pair = & cuda->shared_data.pair;
cu_atom->update_nlocal = 1;
cu_atom->update_nmax = 1;
@ -408,8 +406,6 @@ void VerletCuda::setup_minimal(int flag)
//cuda->allocate();
cuda_shared_atom* cu_atom = & cuda->shared_data.atom;
cuda_shared_domain* cu_domain = & cuda->shared_data.domain;
cuda_shared_pair* cu_pair = & cuda->shared_data.pair;
cu_atom->update_nlocal = 1;
cu_atom->update_nmax = 1;
@ -590,20 +586,9 @@ void VerletCuda::run(int n)
cuda->setTimingsZero();
static double testtime = 0.0;
// my_gettime(CLOCK_REALTIME,&starttime);
// my_gettime(CLOCK_REALTIME,&endtime);
// testtime+=endtime.tv_sec-starttime.tv_sec+1.0*(endtime.tv_nsec-starttime.tv_nsec)/1000000000;
// printf("Time: %lf\n",testtime);*/
cuda_shared_domain* cu_domain = & cuda->shared_data.domain;
int nflag, ntimestep, sortflag;
int n_initial_integrate = modify_cuda->n_initial_integrate;
int n_post_integrate = modify_cuda->n_post_integrate;
int n_final_integrate = modify_cuda->n_final_integrate;
int n_pre_exchange = modify_cuda->n_pre_exchange;
int n_pre_neighbor = modify_cuda->n_pre_neighbor;
int n_pre_force = modify_cuda->n_pre_force;
@ -694,7 +679,7 @@ void VerletCuda::run(int n)
my_gettime(CLOCK_REALTIME, &starttime);
timer->stamp();
comm->forward_comm(1);
timer->stamp(TIME_COMM);
timer->stamp(Timer::COMM);
my_gettime(CLOCK_REALTIME, &endtime);
cuda->shared_data.cuda_timings.comm_forward_total +=
endtime.tv_sec - starttime.tv_sec + 1.0 * (endtime.tv_nsec - starttime.tv_nsec) / 1000000000;
@ -714,7 +699,7 @@ void VerletCuda::run(int n)
//start force calculation asynchronus
cuda->shared_data.comm.comm_phase = 1;
force->pair->compute(eflag, vflag);
timer->stamp(TIME_PAIR);
timer->stamp(Timer::PAIR);
//CudaWrapper_Sync();
//download comm buffers from GPU, perform MPI communication and upload buffers again
@ -723,11 +708,11 @@ void VerletCuda::run(int n)
my_gettime(CLOCK_REALTIME, &endtime);
cuda->shared_data.cuda_timings.comm_forward_total +=
endtime.tv_sec - starttime.tv_sec + 1.0 * (endtime.tv_nsec - starttime.tv_nsec) / 1000000000;
timer->stamp(TIME_COMM);
timer->stamp(Timer::COMM);
//wait for force calculation
CudaWrapper_Sync();
timer->stamp(TIME_PAIR);
timer->stamp(Timer::PAIR);
//unpack communication buffers
my_gettime(CLOCK_REALTIME, &starttime);
@ -736,7 +721,7 @@ void VerletCuda::run(int n)
cuda->shared_data.cuda_timings.comm_forward_total +=
endtime.tv_sec - starttime.tv_sec + 1.0 * (endtime.tv_nsec - starttime.tv_nsec) / 1000000000;
timer->stamp(TIME_COMM);
timer->stamp(Timer::COMM);
MYDBG(printf("# CUDA VerletCuda::iterate: communicate done\n");)
cuda->shared_data.cuda_timings.test1 +=
endtotal.tv_sec - starttotal.tv_sec + 1.0 * (endtotal.tv_nsec - starttotal.tv_nsec) / 1000000000;
@ -747,7 +732,7 @@ void VerletCuda::run(int n)
my_gettime(CLOCK_REALTIME, &endtime);
cuda->shared_data.cuda_timings.comm_forward_total +=
endtime.tv_sec - starttime.tv_sec + 1.0 * (endtime.tv_nsec - starttime.tv_nsec) / 1000000000;
timer->stamp(TIME_COMM);
timer->stamp(Timer::COMM);
MYDBG(printf("# CUDA VerletCuda::iterate: communicate done\n");)
}
} else {
@ -837,7 +822,7 @@ void VerletCuda::run(int n)
cuda->shared_data.buffer_new = 2;
MYDBG(printf("# CUDA VerletCuda::iterate: neighbor build\n");)
timer->stamp(TIME_COMM);
timer->stamp(Timer::COMM);
my_gettime(CLOCK_REALTIME, &endtime);
cuda->shared_data.cuda_timings.test2 +=
endtime.tv_sec - starttime.tv_sec + 1.0 * (endtime.tv_nsec - starttime.tv_nsec) / 1000000000;
@ -845,7 +830,7 @@ void VerletCuda::run(int n)
//rebuild neighbor list
test_atom(testatom, "Pre Neighbor");
neighbor->build(0);
timer->stamp(TIME_NEIGHBOR);
timer->stamp(Timer::NEIGH);
MYDBG(printf("# CUDA VerletCuda::iterate: neighbor done\n");)
//if bonded interactions are used (in this case collect_forces_later is true), transfer data which only changes upon exchange/border routines from GPU to CPU
if(cuda->shared_data.pair.collect_forces_later) {
@ -932,7 +917,7 @@ void VerletCuda::run(int n)
if(not cuda->shared_data.pair.collect_forces_later)
CudaWrapper_Sync();
timer->stamp(TIME_PAIR);
timer->stamp(Timer::PAIR);
}
//calculate bonded interactions
@ -942,11 +927,11 @@ void VerletCuda::run(int n)
if(n_pre_force == 0) Verlet::force_clear();
else cuda->cu_f->downloadAsync(2);
timer->stamp(TIME_PAIR);
timer->stamp(Timer::PAIR);
if(neighbor->lastcall == update->ntimestep) {
neighbor->build_topology();
timer->stamp(TIME_NEIGHBOR);
timer->stamp(Timer::NEIGH);
}
test_atom(testatom, "pre bond force");
@ -959,7 +944,7 @@ void VerletCuda::run(int n)
if(force->improper) force->improper->compute(eflag, vflag);
timer->stamp(TIME_BOND);
timer->stamp(Timer::BOND);
}
//collect forces in case pair force and bonded interactions were overlapped, and either no KSPACE or a GPU KSPACE style is used
@ -984,7 +969,7 @@ void VerletCuda::run(int n)
if(vflag) cuda->cu_virial->download();
timer->stamp(TIME_PAIR);
timer->stamp(Timer::PAIR);
my_gettime(CLOCK_REALTIME, &endtime);
cuda->shared_data.cuda_timings.pair_force_collection +=
@ -1002,7 +987,7 @@ void VerletCuda::run(int n)
if(n_pre_force == 0) Verlet::force_clear();
else cuda->cu_f->downloadAsync(2);
timer->stamp(TIME_PAIR);
timer->stamp(Timer::PAIR);
}
force->kspace->compute(eflag, vflag);
@ -1010,7 +995,7 @@ void VerletCuda::run(int n)
if((not cuda->shared_data.pppm.cudable_force) && (not cuda->shared_data.pair.collect_forces_later))
cuda->uploadAll();
timer->stamp(TIME_KSPACE);
timer->stamp(Timer::KSPACE);
}
//collect forces in case pair forces and kspace was overlaped
@ -1033,7 +1018,7 @@ void VerletCuda::run(int n)
if(vflag) cuda->cu_virial->download();
timer->stamp(TIME_PAIR);
timer->stamp(Timer::PAIR);
my_gettime(CLOCK_REALTIME, &endtime);
cuda->shared_data.cuda_timings.pair_force_collection +=
@ -1043,7 +1028,7 @@ void VerletCuda::run(int n)
//send forces on ghost atoms back to other GPU: THIS SHOULD NEVER HAPPEN
if(force->newton) {
comm->reverse_comm();
timer->stamp(TIME_COMM);
timer->stamp(Timer::COMM);
}
test_atom(testatom, "post force");
@ -1069,7 +1054,7 @@ void VerletCuda::run(int n)
timer->stamp();
output->write(ntimestep);
timer->stamp(TIME_OUTPUT);
timer->stamp(Timer::OUTPUT);
}
@ -1117,8 +1102,7 @@ void VerletCuda::force_clear()
if(cuda->cu_torque) cuda->cu_torque->memset_device(0);
return;
#if 0
//The rest should not be necessary
int i;
@ -1179,9 +1163,10 @@ void VerletCuda::force_clear()
}
}
}
#endif
}
void VerletCuda::test_atom(int aatom, char* string) //printing properties of one atom for test purposes
void VerletCuda::test_atom(int aatom, const char* string) //printing properties of one atom for test purposes
{
if(not dotestatom) return;
@ -1192,7 +1177,9 @@ void VerletCuda::test_atom(int aatom, char* string) //printing properties of on
for(int i = 0; i < atom->nlocal + atom->nghost; i++) {
if((atom->tag[i] == aatom) && (i < atom->nlocal)) {
printf("%i # CUDA %s: %i %i %e %e %e %i ", comm->me, string, update->ntimestep, atom->tag[i], atom->x[i][0], atom->v[i][0], atom->f[i][0], i);
printf("%i # CUDA %s: " BIGINT_FORMAT " %i %e %e %e %i ",
comm->me, string, update->ntimestep, atom->tag[i],
atom->x[i][0], atom->v[i][0], atom->f[i][0], i);
if(atom->molecular && (i < atom->nlocal)) {
printf(" // %i %i %i ", atom->num_bond[i], atom->num_angle[i], atom->num_dihedral[i]);

2
src/USER-CUDA/verlet_cuda.h
View File

@ -43,7 +43,7 @@ class VerletCuda : public Verlet
void setup_minimal(int);
void run(int);
void test_atom(int atom,char* astring); //debugging purpose
void test_atom(int atom,const char* astring); //debugging purpose
int dotestatom; //debugging purpose
protected:

8
src/USER-INTEL/verlet_split_intel.cpp
View File

@ -516,8 +516,6 @@ void VerletSplitIntel::rk_setup()
void VerletSplitIntel::r2k_comm()
{
MPI_Status status;
int n = 0;
if (master) n = atom->nlocal;
MPI_Gatherv(atom->x[0],n*3,MPI_DOUBLE,atom->x[0],xsize,xdisp,
@ -531,7 +529,7 @@ void VerletSplitIntel::r2k_comm()
MPI_Send(flags,2,MPI_INT,0,0,block);
} else if (!master) {
int flags[2];
MPI_Recv(flags,2,MPI_DOUBLE,1,0,block,&status);
MPI_Recv(flags,2,MPI_DOUBLE,1,0,block,MPI_STATUS_IGNORE);
eflag = flags[0]; vflag = flags[1];
}
@ -542,8 +540,8 @@ void VerletSplitIntel::r2k_comm()
MPI_Send(domain->boxlo,3,MPI_DOUBLE,0,0,block);
MPI_Send(domain->boxhi,3,MPI_DOUBLE,0,0,block);
} else if (!master) {
MPI_Recv(domain->boxlo,3,MPI_DOUBLE,1,0,block,&status);
MPI_Recv(domain->boxhi,3,MPI_DOUBLE,1,0,block,&status);
MPI_Recv(domain->boxlo,3,MPI_DOUBLE,1,0,block,MPI_STATUS_IGNORE);
MPI_Recv(domain->boxhi,3,MPI_DOUBLE,1,0,block,MPI_STATUS_IGNORE);
domain->set_global_box();
domain->set_local_box();
force->kspace->setup();

60
src/USER-LB/fix_lb_fluid.cpp
View File

@ -802,7 +802,6 @@ void FixLbFluid::calc_fluidforce(void)
double **x = atom->x;
int i,j,k,m;
MPI_Request requests[20];
MPI_Status statuses[20];
double forceloc[3],force[3];
double torqueloc[3],torque[3];
@ -901,7 +900,7 @@ void FixLbFluid::calc_fluidforce(void)
MPI_Irecv(&Fftempx[2][0][0][0],1,passxtemp,comm->procneigh[0][0],30,world,&requests[7]);
MPI_Irecv(&Fftempx[3][0][0][0],1,passxtemp,comm->procneigh[0][0],40,world,&requests[8]);
MPI_Irecv(&Fftempx[4][0][0][0],1,passxtemp,comm->procneigh[0][0],50,world,&requests[9]);
MPI_Waitall(10,requests,statuses);
MPI_Waitall(10,requests,MPI_STATUS_IGNORE);
for(j=0; j<subNby+3; j++){
for(k=0; k<subNbz+3; k++){
@ -927,7 +926,7 @@ void FixLbFluid::calc_fluidforce(void)
MPI_Irecv(&Fftempy[0][2][0][0],1,passytemp,comm->procneigh[1][0],30,world,&requests[7]);
MPI_Irecv(&Fftempy[0][3][0][0],1,passytemp,comm->procneigh[1][0],40,world,&requests[8]);
MPI_Irecv(&Fftempy[0][4][0][0],1,passytemp,comm->procneigh[1][0],50,world,&requests[9]);
MPI_Waitall(10,requests,statuses);
MPI_Waitall(10,requests,MPI_STATUS_IGNORE);
for(i=0; i<subNbx+3; i++){
for(k=0; k<subNbz+3; k++){
@ -953,7 +952,7 @@ void FixLbFluid::calc_fluidforce(void)
MPI_Irecv(&Fftempz[0][0][2][0],1,passztemp,comm->procneigh[2][0],30,world,&requests[7]);
MPI_Irecv(&Fftempz[0][0][3][0],1,passztemp,comm->procneigh[2][0],40,world,&requests[8]);
MPI_Irecv(&Fftempz[0][0][4][0],1,passztemp,comm->procneigh[2][0],50,world,&requests[9]);
MPI_Waitall(10,requests,statuses);
MPI_Waitall(10,requests,MPI_STATUS_IGNORE);
for(i=0; i<subNbx+3; i++){
for(j=0; j<subNby+3; j++){
@ -1837,7 +1836,6 @@ void FixLbFluid::initialize_feq(void)
{
int i,j,k,p;
MPI_Request requests[8];
MPI_Status statuses[8];
int numrequests;
// If using the standary LB integrator, do not need to send feqn.
@ -1870,7 +1868,7 @@ void FixLbFluid::initialize_feq(void)
MPI_Isend(&feqn[subNbx-2][1][1][0],1,passxf,comm->procneigh[0][1],20,world,&requests[6]);
MPI_Irecv(&feqn[subNbx-1][1][1][0],1,passxf,comm->procneigh[0][1],10,world,&requests[7]);
}
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
for(i=0; i<numrequests; i++)
requests[i]=MPI_REQUEST_NULL;
@ -1884,7 +1882,7 @@ void FixLbFluid::initialize_feq(void)
MPI_Isend(&feqn[0][subNby-2][1][0],1,passyf,comm->procneigh[1][1],20,world,&requests[6]);
MPI_Irecv(&feqn[0][subNby-1][1][0],1,passyf,comm->procneigh[1][1],10,world,&requests[7]);
}
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
for(i=0; i<numrequests; i++)
requests[i]=MPI_REQUEST_NULL;
@ -1898,7 +1896,7 @@ void FixLbFluid::initialize_feq(void)
MPI_Isend(&feqn[0][0][subNbz-2][0],1,passzf,comm->procneigh[2][1],20,world,&requests[6]);
MPI_Irecv(&feqn[0][0][subNbz-1][0],1,passzf,comm->procneigh[2][1],10,world,&requests[7]);
}
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
//Save feqold.
if(typeLB == 2){
@ -1926,7 +1924,7 @@ void FixLbFluid::initialize_feq(void)
MPI_Irecv(&feqoldn[0][1][1][0],1,passxf,comm->procneigh[0][0],20,world,&requests[5]);
MPI_Isend(&feqoldn[subNbx-2][1][1][0],1,passxf,comm->procneigh[0][1],20,world,&requests[6]);
MPI_Irecv(&feqoldn[subNbx-1][1][1][0],1,passxf,comm->procneigh[0][1],10,world,&requests[7]);
MPI_Waitall(8,requests,statuses);
MPI_Waitall(8,requests,MPI_STATUS_IGNORE);
for(i=0; i<8; i++)
requests[i]=MPI_REQUEST_NULL;
@ -1938,7 +1936,7 @@ void FixLbFluid::initialize_feq(void)
MPI_Irecv(&feqoldn[0][0][1][0],1,passyf,comm->procneigh[1][0],20,world,&requests[5]);
MPI_Isend(&feqoldn[0][subNby-2][1][0],1,passyf,comm->procneigh[1][1],20,world,&requests[6]);
MPI_Irecv(&feqoldn[0][subNby-1][1][0],1,passyf,comm->procneigh[1][1],10,world,&requests[7]);
MPI_Waitall(8,requests,statuses);
MPI_Waitall(8,requests,MPI_STATUS_IGNORE);
for(i=0; i<8; i++)
requests[i]=MPI_REQUEST_NULL;
@ -1950,7 +1948,7 @@ void FixLbFluid::initialize_feq(void)
MPI_Irecv(&feqoldn[0][0][0][0],1,passzf,comm->procneigh[2][0],20,world,&requests[5]);
MPI_Isend(&feqoldn[0][0][subNbz-2][0],1,passzf,comm->procneigh[2][1],20,world,&requests[6]);
MPI_Irecv(&feqoldn[0][0][subNbz-1][0],1,passzf,comm->procneigh[2][1],10,world,&requests[7]);
MPI_Waitall(8,requests,statuses);
MPI_Waitall(8,requests,MPI_STATUS_IGNORE);
}
parametercalc_full();
}
@ -2325,9 +2323,7 @@ void FixLbFluid::equilibriumdist19(int xstart, int xend, int ystart, int yend, i
void FixLbFluid::parametercalc_full(void)
{
MPI_Request requests[4];
MPI_Status statuses[4];
MPI_Request requests2[12];
MPI_Status statuses2[12];
int numrequests;
int i;
@ -2343,7 +2339,7 @@ void FixLbFluid::parametercalc_full(void)
MPI_Isend(&f_lb[subNbx-2][1][1][0],1,passxf,comm->procneigh[0][1],20,world,&requests[2]);
MPI_Irecv(&f_lb[subNbx-1][1][1][0],1,passxf,comm->procneigh[0][1],10,world,&requests[3]);
parametercalc_part(1,subNbx-1,1,subNby-1,1,subNbz-1);
MPI_Waitall(4,requests,statuses);
MPI_Waitall(4,requests,MPI_STATUS_IGNORE);
for(i=0; i<4; i++)
requests[i]=MPI_REQUEST_NULL;
@ -2353,7 +2349,7 @@ void FixLbFluid::parametercalc_full(void)
MPI_Irecv(&f_lb[0][subNby-1][1][0],1,passyf,comm->procneigh[1][1],10,world,&requests[3]);
parametercalc_part(0,1,1,subNby-1,1,subNbz-1);
parametercalc_part(subNbx-1,subNbx,1,subNby-1,1,subNbz-1);
MPI_Waitall(4,requests,statuses);
MPI_Waitall(4,requests,MPI_STATUS_IGNORE);
for(i=0; i<4; i++)
requests[i]=MPI_REQUEST_NULL;
@ -2363,7 +2359,7 @@ void FixLbFluid::parametercalc_full(void)
MPI_Irecv(&f_lb[0][0][subNbz-1][0],1,passzf,comm->procneigh[2][1],10,world,&requests[3]);
parametercalc_part(0,subNbx,0,1,1,subNbz-1);
parametercalc_part(0,subNbx,subNby-1,subNby,1,subNbz-1);
MPI_Waitall(4,requests,statuses);
MPI_Waitall(4,requests,MPI_STATUS_IGNORE);
parametercalc_part(0,subNbx,0,subNby,0,1);
parametercalc_part(0,subNbx,0,subNby,subNbz-1,subNbz);
@ -2391,7 +2387,7 @@ void FixLbFluid::parametercalc_full(void)
MPI_Irecv(&density_lb[subNbx+1][0][0],1,passxrho,comm->procneigh[0][1],50,world,&requests2[10]);
MPI_Irecv(&density_lb[subNbx+2][0][0],1,passxrho,comm->procneigh[0][0],60,world,&requests2[11]);
}
MPI_Waitall(numrequests,requests2,statuses2);
MPI_Waitall(numrequests,requests2,MPI_STATUS_IGNORE);
for(i=0; i<numrequests; i++)
requests2[i]=MPI_REQUEST_NULL;
@ -2409,7 +2405,7 @@ void FixLbFluid::parametercalc_full(void)
MPI_Irecv(&density_lb[0][subNby+1][0],1,passyrho,comm->procneigh[1][1],50,world,&requests2[10]);
MPI_Irecv(&density_lb[0][subNby+2][0],1,passyrho,comm->procneigh[1][0],60,world,&requests2[11]);
}
MPI_Waitall(numrequests,requests2,statuses2);
MPI_Waitall(numrequests,requests2,MPI_STATUS_IGNORE);
for(i=0; i<12; i++)
requests2[i]=MPI_REQUEST_NULL;
@ -2438,7 +2434,7 @@ void FixLbFluid::parametercalc_full(void)
requestcount=requestcount+3;
}
}
MPI_Waitall(requestcount,requests2,statuses2);
MPI_Waitall(requestcount,requests2,MPI_STATUS_IGNORE);
}
@ -2670,7 +2666,6 @@ void FixLbFluid::update_full15(void)
MPI_Request req_send15,req_recv15;
MPI_Request req_send25,req_recv25;
MPI_Request requests[8];
MPI_Status statuses[8];
int numrequests;
double tmp1;
MPI_Status status;
@ -2706,7 +2701,7 @@ void FixLbFluid::update_full15(void)
MPI_Irecv(&feqn[subNbx-1][1][1][0],1,passxf,comm->procneigh[0][1],10,world,&requests[7]);
}
update_periodic(2,subNbx-2,2,subNby-2,2,subNbz-2);
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
for(i=0; i<numrequests; i++)
@ -2723,7 +2718,7 @@ void FixLbFluid::update_full15(void)
}
update_periodic(1,2,2,subNby-2,2,subNbz-2);
update_periodic(subNbx-2,subNbx-1,2,subNby-2,2,subNbz-2);
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
for(i=0; i<numrequests; i++)
requests[i]=MPI_REQUEST_NULL;
@ -2739,7 +2734,7 @@ void FixLbFluid::update_full15(void)
}
update_periodic(1,subNbx-1,1,2,2,subNbz-2);
update_periodic(1,subNbx-1,subNby-2,subNby-1,2,subNbz-2);
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
if(typeLB==1){
update_periodic(1,subNbx-1,1,subNby-1,1,2);
@ -2981,7 +2976,7 @@ void FixLbFluid::update_full15(void)
MPI_Irecv(&feqn[subNbx-1][1][1][0],1,passxf,comm->procneigh[0][1],10,world,&requests[7]);
}
update_periodic(2,subNbx-2,2,subNby-2,2,subNbz-2);
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
for(i=0; i<numrequests; i++)
requests[i]=MPI_REQUEST_NULL;
@ -2997,7 +2992,7 @@ void FixLbFluid::update_full15(void)
}
update_periodic(1,2,2,subNby-2,2,subNbz-2);
update_periodic(subNbx-2,subNbx-1,2,subNby-2,2,subNbz-2);
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
for(i=0; i<numrequests; i++)
requests[i]=MPI_REQUEST_NULL;
@ -3013,7 +3008,7 @@ void FixLbFluid::update_full15(void)
}
update_periodic(1,subNbx-1,1,2,2,subNbz-2);
update_periodic(1,subNbx-1,subNby-2,subNby-1,2,subNbz-2);
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
update_periodic(1,subNbx-1,1,subNby-1,1,2);
update_periodic(1,subNbx-1,1,subNby-1,subNbz-2,subNbz-1);
@ -3031,7 +3026,6 @@ void FixLbFluid::update_full19(void)
MPI_Request req_send15,req_recv15;
MPI_Request req_send25,req_recv25;
MPI_Request requests[8];
MPI_Status statuses[8];
int numrequests;
double tmp1,tmp2,tmp3;
MPI_Status status;
@ -3067,7 +3061,7 @@ void FixLbFluid::update_full19(void)
MPI_Irecv(&feqn[subNbx-1][1][1][0],1,passxf,comm->procneigh[0][1],10,world,&requests[7]);
}
update_periodic(2,subNbx-2,2,subNby-2,2,subNbz-2);
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
for(i=0; i<numrequests; i++)
requests[i]=MPI_REQUEST_NULL;
@ -3083,7 +3077,7 @@ void FixLbFluid::update_full19(void)
}
update_periodic(1,2,2,subNby-2,2,subNbz-2);
update_periodic(subNbx-2,subNbx-1,2,subNby-2,2,subNbz-2);
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
for(i=0; i<numrequests; i++)
requests[i]=MPI_REQUEST_NULL;
@ -3099,7 +3093,7 @@ void FixLbFluid::update_full19(void)
}
update_periodic(1,subNbx-1,1,2,2,subNbz-2);
update_periodic(1,subNbx-1,subNby-2,subNby-1,2,subNbz-2);
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
if(typeLB==1){
update_periodic(1,subNbx-1,1,subNby-1,1,2);
@ -3331,7 +3325,7 @@ void FixLbFluid::update_full19(void)
MPI_Irecv(&feqn[subNbx-1][1][1][0],1,passxf,comm->procneigh[0][1],10,world,&requests[7]);
}
update_periodic(2,subNbx-2,2,subNby-2,2,subNbz-2);
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
for(i=0; i<numrequests; i++)
requests[i]=MPI_REQUEST_NULL;
@ -3347,7 +3341,7 @@ void FixLbFluid::update_full19(void)
}
update_periodic(1,2,2,subNby-2,2,subNbz-2);
update_periodic(subNbx-2,subNbx-1,2,subNby-2,2,subNbz-2);
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
for(i=0; i<numrequests; i++)
requests[i]=MPI_REQUEST_NULL;
@ -3363,7 +3357,7 @@ void FixLbFluid::update_full19(void)
}
update_periodic(1,subNbx-1,1,2,2,subNbz-2);
update_periodic(1,subNbx-1,subNby-2,subNby-1,2,subNbz-2);
MPI_Waitall(numrequests,requests,statuses);
MPI_Waitall(numrequests,requests,MPI_STATUS_IGNORE);
update_periodic(1,subNbx-1,1,subNby-1,1,2);
update_periodic(1,subNbx-1,1,subNby-1,subNbz-2,subNbz-1);

11
src/USER-MISC/fix_ave_spatial_sphere.cpp
View File

@ -602,7 +602,6 @@ void FixAveSpatialSphere::end_of_step()
}
//bin each atom
double **x= atom->x;
int *mask= atom->mask;
int nlocal= atom->nlocal;
@ -869,10 +868,8 @@ void FixAveSpatialSphere::setup_bins()
}
//set the bin coordinates
for(int i= 0; i < nbins; i++) {
for(int i= 0; i < nbins; i++)
coord[i]= r_min + (i+0.5)*deltar;
double temp_r= r_min + (i+1)*deltar;
}
set_bin_volumes();
}
@ -942,17 +939,13 @@ void FixAveSpatialSphere::bin_atoms()
int *mask = atom->mask;
int nlocal = atom->nlocal;
double *boxlo,*boxhi,*prd,*prd_half;
double *prd,*prd_half;
//deal with the periodic boundary conditions
if(domain->periodicity[0] || domain->periodicity[1] || domain->periodicity[2]) {
if(scaleflag == REDUCED) {
boxlo= domain->boxlo_lamda;
boxhi= domain->boxhi_lamda;
prd= domain->prd_lamda;
prd_half= domain->prd_half_lamda;
} else {
boxlo= domain->boxlo;
boxhi= domain->boxhi;
prd= domain->prd;
prd_half= domain->prd_half;
}

2
src/USER-MISC/fix_ave_spatial_sphere.h
View File

@ -1,4 +1,4 @@
/* ----------------------------------------------------------------------
/* -*- c++ -*- ----------------------------------------------------------
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
http://lammps.sandia.gov, Sandia National Laboratories
Steve Plimpton, sjplimp@sandia.gov

12
src/USER-MISC/fix_imd.cpp
View File

@ -713,12 +713,12 @@ void FixIMD::setup(int)
taginthash_init(hashtable, num_coords);
idmap = (void *)hashtable;
MPI_Status status;
MPI_Request request;
int tmp, ndata;
struct commdata *buf = static_cast<struct commdata *>(comm_buf);
if (me == 0) {
MPI_Status status;
MPI_Request request;
tagint *taglist = new tagint[num_coords];
int numtag=0; /* counter to map atom tags to a 0-based consecutive index list */
@ -763,7 +763,7 @@ void FixIMD::setup(int)
}
}
/* blocking receive to wait until it is our turn to send data. */
MPI_Recv(&tmp, 0, MPI_INT, 0, 0, world, &status);
MPI_Recv(&tmp, 0, MPI_INT, 0, 0, world, MPI_STATUS_IGNORE);
MPI_Rsend(comm_buf, nme*size_one, MPI_BYTE, 0, 0, world);
}
@ -1012,12 +1012,12 @@ void FixIMD::post_force(int vflag)
comm_buf = memory->smalloc(maxbuf,"imd:comm_buf");
}
MPI_Status status;
MPI_Request request;
int tmp, ndata;
buf = static_cast<struct commdata *>(comm_buf);
if (me == 0) {
MPI_Status status;
MPI_Request request;
/* collect data into new array. we bypass the IMD API to save
* us one extra copy of the data. */
msglen = 3*sizeof(float)*num_coords+IMDHEADERSIZE;
@ -1145,7 +1145,7 @@ void FixIMD::post_force(int vflag)
}
}
/* blocking receive to wait until it is our turn to send data. */
MPI_Recv(&tmp, 0, MPI_INT, 0, 0, world, &status);
MPI_Recv(&tmp, 0, MPI_INT, 0, 0, world, MPI_STATUS_IGNORE);
MPI_Rsend(comm_buf, nme*size_one, MPI_BYTE, 0, 0, world);
}

2
src/USER-MISC/fix_ipi.h
View File

@ -1,4 +1,4 @@
/* ----------------------------------------------------------------------
/* -*- c++ -*- ----------------------------------------------------------
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
http://lammps.sandia.gov, Sandia National Laboratories
Steve Plimpton, sjplimp@sandia.gov

8
src/USER-MISC/fix_pimd.cpp
View File

@ -602,8 +602,6 @@ void FixPIMD::comm_init()
void FixPIMD::comm_exec(double **ptr)
{
MPI_Status status;
int nlocal = atom->nlocal;
if(nlocal > max_nlocal)
@ -629,7 +627,7 @@ void FixPIMD::comm_exec(double **ptr)
int nsend;
MPI_Sendrecv( &(nlocal), 1, MPI_INT, plan_send[iplan], 0,
&(nsend), 1, MPI_INT, plan_recv[iplan], 0, universe->uworld, &status);
&(nsend), 1, MPI_INT, plan_recv[iplan], 0, universe->uworld, MPI_STATUS_IGNORE);
// allocate arrays
@ -643,7 +641,7 @@ void FixPIMD::comm_exec(double **ptr)
// send tags
MPI_Sendrecv( atom->tag, nlocal, MPI_INT, plan_send[iplan], 0,
tag_send, nsend, MPI_INT, plan_recv[iplan], 0, universe->uworld, &status);
tag_send, nsend, MPI_INT, plan_recv[iplan], 0, universe->uworld, MPI_STATUS_IGNORE);
// wrap positions
@ -671,7 +669,7 @@ void FixPIMD::comm_exec(double **ptr)
// sendrecv x
MPI_Sendrecv( buf_send, nsend*3, MPI_DOUBLE, plan_recv[iplan], 0,
buf_recv, nlocal*3, MPI_DOUBLE, plan_send[iplan], 0, universe->uworld, &status);
buf_recv, nlocal*3, MPI_DOUBLE, plan_send[iplan], 0, universe->uworld, MPI_STATUS_IGNORE);
// copy x

2
src/USER-MISC/fix_pimd.h
View File

@ -1,4 +1,4 @@
/* ----------------------------------------------------------------------
/* -*- c++ -*- ----------------------------------------------------------
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
http://lammps.sandia.gov, Sandia National Laboratories
Steve Plimpton, sjplimp@sandia.gov

2
src/USER-MISC/fix_srp.h
View File

@ -1,4 +1,4 @@
/* ----------------------------------------------------------------------
/* -*- c++ -*- ----------------------------------------------------------
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
http://lammps.sandia.gov, Sandia National Laboratories
Steve Plimpton, sjplimp@sandia.gov

2
src/USER-MISC/pair_srp.h
View File

@ -1,4 +1,4 @@
/* ----------------------------------------------------------------------
/* -*- c++ -*- ----------------------------------------------------------
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
http://lammps.sandia.gov, Sandia National Laboratories
Steve Plimpton, sjplimp@sandia.gov

6
src/USER-MOLFILE/dump_molfile.cpp
View File

@ -234,10 +234,10 @@ void DumpMolfile::write()
sort();
int tmp,nlines;
MPI_Status status;
MPI_Request request;
if (me == 0) {
MPI_Status status;
MPI_Request request;
for (int iproc = 0; iproc < nprocs; iproc++) {
if (iproc) {
MPI_Irecv(buf,maxbuf*size_one,MPI_DOUBLE,iproc,0,world,&request);
@ -251,7 +251,7 @@ void DumpMolfile::write()
}
} else {
MPI_Recv(&tmp,0,MPI_INT,0,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,0,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(buf,nme*size_one,MPI_DOUBLE,0,0,world);
}
}

8
src/USER-QMMM/fix_qmmm.cpp
View File

@ -395,8 +395,6 @@ void FixQMMM::exchange_positions()
}
if (qmmm_role == QMMM_ROLE_MASTER) {
MPI_Status status;
MPI_Request request;
int i,tmp;
/* check and potentially grow local communication buffers. */
@ -408,6 +406,8 @@ void FixQMMM::exchange_positions()
struct commdata *buf = static_cast<struct commdata *>(comm_buf);
if (comm->me == 0) {
MPI_Status status;
MPI_Request request;
// insert local atoms into comm buffer
for (i=0; i<nlocal; ++i) {
if (mask[i] & groupbit) {
@ -642,8 +642,6 @@ void FixQMMM::init()
taginthash_init(qm_hash, num_qm);
qm_idmap = (void *)qm_hash;
MPI_Status status;
MPI_Request request;
const int nlocal = atom->nlocal;
int i, j, tmp, ndata, qm_ntag;
tagint *tag = atom->tag;
@ -651,6 +649,8 @@ void FixQMMM::init()
struct commdata *buf = static_cast<struct commdata *>(comm_buf);
if (me == 0) {
MPI_Status status;
MPI_Request request;
tagint *qm_taglist = new tagint[num_qm];
qm_ntag = 0;
for (i=0; i < nlocal; ++i) {

5
src/USER-REAXC/fix_reaxc_bonds.cpp
View File

@ -255,7 +255,6 @@ void FixReaxCBonds::RecvBuffer(double *buf, int nbuf, int nbuf_local,
double cutof3 = reaxc->control->bg_cut;
MPI_Request irequest, irequest2;
MPI_Status istatus;
if (me == 0 ){
fprintf(fp,"# Timestep " BIGINT_FORMAT " \n",ntimestep);
@ -275,7 +274,7 @@ void FixReaxCBonds::RecvBuffer(double *buf, int nbuf, int nbuf_local,
nlocal_tmp = nlocal;
} else {
MPI_Irecv(&buf[0],nbuf,MPI_DOUBLE,inode,0,world,&irequest);
MPI_Wait(&irequest,&istatus);
MPI_Wait(&irequest,MPI_STATUS_IGNORE);
nlocal_tmp = nint(buf[0]);
}
j = 2;
@ -308,7 +307,7 @@ void FixReaxCBonds::RecvBuffer(double *buf, int nbuf, int nbuf_local,
}
} else {
MPI_Isend(&buf[0],nbuf_local,MPI_DOUBLE,0,0,world,&irequest2);
MPI_Wait(&irequest2,&istatus);
MPI_Wait(&irequest2,MPI_STATUS_IGNORE);
}
if(me ==0) fprintf(fp,"# \n");

1
src/USER-REAXC/reaxc_traj.cpp
View File

@ -264,7 +264,6 @@ int Write_Init_Desc( reax_system *system, control_params *control,
{
int i, me, np, cnt, buffer_len, buffer_req;
reax_atom *p_atom;
//MPI_Request request;
MPI_Status status;
me = system->my_rank;

5
src/comm.cpp
View File

@ -407,14 +407,13 @@ void Comm::set_proc_grid(int outflag)
// recv 3d proc grid of another partition if my 3d grid depends on it
if (recv_from_partition >= 0) {
MPI_Status status;
if (me == 0) {
MPI_Recv(other_procgrid,3,MPI_INT,
universe->root_proc[recv_from_partition],0,
universe->uworld,&status);
universe->uworld,MPI_STATUS_IGNORE);
MPI_Recv(other_coregrid,3,MPI_INT,
universe->root_proc[recv_from_partition],0,
universe->uworld,&status);
universe->uworld,MPI_STATUS_IGNORE);
}
MPI_Bcast(other_procgrid,3,MPI_INT,0,world);
MPI_Bcast(other_coregrid,3,MPI_INT,0,world);

62
src/comm_brick.cpp
View File

@ -444,7 +444,6 @@ void CommBrick::forward_comm(int dummy)
{
int n;
MPI_Request request;
MPI_Status status;
AtomVec *avec = atom->avec;
double **x = atom->x;
double *buf;
@ -465,7 +464,7 @@ void CommBrick::forward_comm(int dummy)
n = avec->pack_comm(sendnum[iswap],sendlist[iswap],
buf_send,pbc_flag[iswap],pbc[iswap]);
if (n) MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
if (size_forward_recv[iswap]) MPI_Wait(&request,&status);
if (size_forward_recv[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
} else if (ghost_velocity) {
if (size_forward_recv[iswap])
MPI_Irecv(buf_recv,size_forward_recv[iswap],MPI_DOUBLE,
@ -473,7 +472,7 @@ void CommBrick::forward_comm(int dummy)
n = avec->pack_comm_vel(sendnum[iswap],sendlist[iswap],
buf_send,pbc_flag[iswap],pbc[iswap]);
if (n) MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
if (size_forward_recv[iswap]) MPI_Wait(&request,&status);
if (size_forward_recv[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
avec->unpack_comm_vel(recvnum[iswap],firstrecv[iswap],buf_recv);
} else {
if (size_forward_recv[iswap])
@ -482,7 +481,7 @@ void CommBrick::forward_comm(int dummy)
n = avec->pack_comm(sendnum[iswap],sendlist[iswap],
buf_send,pbc_flag[iswap],pbc[iswap]);
if (n) MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
if (size_forward_recv[iswap]) MPI_Wait(&request,&status);
if (size_forward_recv[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
avec->unpack_comm(recvnum[iswap],firstrecv[iswap],buf_recv);
}
@ -513,7 +512,6 @@ void CommBrick::reverse_comm()
{
int n;
MPI_Request request;
MPI_Status status;
AtomVec *avec = atom->avec;
double **f = atom->f;
double *buf;
@ -534,14 +532,14 @@ void CommBrick::reverse_comm()
MPI_Send(buf,size_reverse_send[iswap],MPI_DOUBLE,
recvproc[iswap],0,world);
}
if (size_reverse_recv[iswap]) MPI_Wait(&request,&status);
if (size_reverse_recv[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
} else {
if (size_reverse_recv[iswap])
MPI_Irecv(buf_recv,size_reverse_recv[iswap],MPI_DOUBLE,
sendproc[iswap],0,world,&request);
n = avec->pack_reverse(recvnum[iswap],firstrecv[iswap],buf_send);
if (n) MPI_Send(buf_send,n,MPI_DOUBLE,recvproc[iswap],0,world);
if (size_reverse_recv[iswap]) MPI_Wait(&request,&status);
if (size_reverse_recv[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
}
avec->unpack_reverse(sendnum[iswap],sendlist[iswap],buf_recv);
@ -576,7 +574,6 @@ void CommBrick::exchange()
double **x;
double *sublo,*subhi;
MPI_Request request;
MPI_Status status;
AtomVec *avec = atom->avec;
// clear global->local map for owned and ghost atoms
@ -644,11 +641,11 @@ void CommBrick::exchange()
if (procgrid[dim] == 1) nrecv = 0;
else {
MPI_Sendrecv(&nsend,1,MPI_INT,procneigh[dim][0],0,
&nrecv1,1,MPI_INT,procneigh[dim][1],0,world,&status);
&nrecv1,1,MPI_INT,procneigh[dim][1],0,world,MPI_STATUS_IGNORE);
nrecv = nrecv1;
if (procgrid[dim] > 2) {
MPI_Sendrecv(&nsend,1,MPI_INT,procneigh[dim][1],0,
&nrecv2,1,MPI_INT,procneigh[dim][0],0,world,&status);
&nrecv2,1,MPI_INT,procneigh[dim][0],0,world,MPI_STATUS_IGNORE);
nrecv += nrecv2;
}
if (nrecv > maxrecv) grow_recv(nrecv);
@ -656,13 +653,13 @@ void CommBrick::exchange()
MPI_Irecv(buf_recv,nrecv1,MPI_DOUBLE,procneigh[dim][1],0,
world,&request);
MPI_Send(buf_send,nsend,MPI_DOUBLE,procneigh[dim][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
if (procgrid[dim] > 2) {
MPI_Irecv(&buf_recv[nrecv1],nrecv2,MPI_DOUBLE,procneigh[dim][0],0,
world,&request);
MPI_Send(buf_send,nsend,MPI_DOUBLE,procneigh[dim][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
}
@ -701,7 +698,6 @@ void CommBrick::borders()
double **x;
double *buf,*mlo,*mhi;
MPI_Request request;
MPI_Status status;
AtomVec *avec = atom->avec;
// do swaps over all 3 dimensions
@ -816,12 +812,12 @@ void CommBrick::borders()
if (sendproc[iswap] != me) {
MPI_Sendrecv(&nsend,1,MPI_INT,sendproc[iswap],0,
&nrecv,1,MPI_INT,recvproc[iswap],0,world,&status);
&nrecv,1,MPI_INT,recvproc[iswap],0,world,MPI_STATUS_IGNORE);
if (nrecv*size_border > maxrecv) grow_recv(nrecv*size_border);
if (nrecv) MPI_Irecv(buf_recv,nrecv*size_border,MPI_DOUBLE,
recvproc[iswap],0,world,&request);
if (n) MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
if (nrecv) MPI_Wait(&request,&status);
if (nrecv) MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else {
nrecv = nsend;
@ -872,7 +868,6 @@ void CommBrick::forward_comm_pair(Pair *pair)
int iswap,n;
double *buf;
MPI_Request request;
MPI_Status status;
int nsize = pair->comm_forward;
@ -892,7 +887,7 @@ void CommBrick::forward_comm_pair(Pair *pair)
recvproc[iswap],0,world,&request);
if (sendnum[iswap])
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
if (recvnum[iswap]) MPI_Wait(&request,&status);
if (recvnum[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else buf = buf_send;
@ -912,7 +907,6 @@ void CommBrick::reverse_comm_pair(Pair *pair)
int iswap,n;
double *buf;
MPI_Request request;
MPI_Status status;
int nsize = MAX(pair->comm_reverse,pair->comm_reverse_off);
@ -931,7 +925,7 @@ void CommBrick::reverse_comm_pair(Pair *pair)
world,&request);
if (recvnum[iswap])
MPI_Send(buf_send,n,MPI_DOUBLE,recvproc[iswap],0,world);
if (sendnum[iswap]) MPI_Wait(&request,&status);
if (sendnum[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else buf = buf_send;
@ -955,7 +949,6 @@ void CommBrick::forward_comm_fix(Fix *fix, int size)
int iswap,n,nsize;
double *buf;
MPI_Request request;
MPI_Status status;
if (size) nsize = size;
else nsize = fix->comm_forward;
@ -976,7 +969,7 @@ void CommBrick::forward_comm_fix(Fix *fix, int size)
world,&request);
if (sendnum[iswap])
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
if (recvnum[iswap]) MPI_Wait(&request,&status);
if (recvnum[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else buf = buf_send;
@ -1000,7 +993,6 @@ void CommBrick::reverse_comm_fix(Fix *fix, int size)
int iswap,n,nsize;
double *buf;
MPI_Request request;
MPI_Status status;
if (size) nsize = size;
else nsize = fix->comm_reverse;
@ -1020,7 +1012,7 @@ void CommBrick::reverse_comm_fix(Fix *fix, int size)
world,&request);
if (recvnum[iswap])
MPI_Send(buf_send,n,MPI_DOUBLE,recvproc[iswap],0,world);
if (sendnum[iswap]) MPI_Wait(&request,&status);
if (sendnum[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else buf = buf_send;
@ -1040,7 +1032,6 @@ void CommBrick::forward_comm_compute(Compute *compute)
int iswap,n;
double *buf;
MPI_Request request;
MPI_Status status;
int nsize = compute->comm_forward;
@ -1060,7 +1051,7 @@ void CommBrick::forward_comm_compute(Compute *compute)
world,&request);
if (sendnum[iswap])
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
if (recvnum[iswap]) MPI_Wait(&request,&status);
if (recvnum[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else buf = buf_send;
@ -1080,7 +1071,6 @@ void CommBrick::reverse_comm_compute(Compute *compute)
int iswap,n;
double *buf;
MPI_Request request;
MPI_Status status;
int nsize = compute->comm_reverse;
@ -1099,7 +1089,7 @@ void CommBrick::reverse_comm_compute(Compute *compute)
world,&request);
if (recvnum[iswap])
MPI_Send(buf_send,n,MPI_DOUBLE,recvproc[iswap],0,world);
if (sendnum[iswap]) MPI_Wait(&request,&status);
if (sendnum[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else buf = buf_send;
@ -1119,7 +1109,6 @@ void CommBrick::forward_comm_dump(Dump *dump)
int iswap,n;
double *buf;
MPI_Request request;
MPI_Status status;
int nsize = dump->comm_forward;
@ -1139,7 +1128,7 @@ void CommBrick::forward_comm_dump(Dump *dump)
world,&request);
if (sendnum[iswap])
MPI_Send(buf_send,n,MPI_DOUBLE,sendproc[iswap],0,world);
if (recvnum[iswap]) MPI_Wait(&request,&status);
if (recvnum[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else buf = buf_send;
@ -1159,7 +1148,6 @@ void CommBrick::reverse_comm_dump(Dump *dump)
int iswap,n;
double *buf;
MPI_Request request;
MPI_Status status;
int nsize = dump->comm_reverse;
@ -1178,7 +1166,7 @@ void CommBrick::reverse_comm_dump(Dump *dump)
world,&request);
if (recvnum[iswap])
MPI_Send(buf_send,n,MPI_DOUBLE,recvproc[iswap],0,world);
if (sendnum[iswap]) MPI_Wait(&request,&status);
if (sendnum[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else buf = buf_send;
@ -1197,7 +1185,6 @@ void CommBrick::forward_comm_array(int nsize, double **array)
int i,j,k,m,iswap,last;
double *buf;
MPI_Request request;
MPI_Status status;
// insure send/recv bufs are big enough for nsize
// based on smax/rmax from most recent borders() invocation
@ -1229,7 +1216,7 @@ void CommBrick::forward_comm_array(int nsize, double **array)
if (sendnum[iswap])
MPI_Send(buf_send,nsize*sendnum[iswap],MPI_DOUBLE,
sendproc[iswap],0,world);
if (recvnum[iswap]) MPI_Wait(&request,&status);
if (recvnum[iswap]) MPI_Wait(&request,MPI_STATUS_IGNORE);
buf = buf_recv;
} else buf = buf_send;
@ -1251,7 +1238,6 @@ int CommBrick::exchange_variable(int n, double *inbuf, double *&outbuf)
{
int nsend,nrecv,nrecv1,nrecv2;
MPI_Request request;
MPI_Status status;
nrecv = n;
if (nrecv > maxrecv) grow_recv(nrecv);
@ -1271,11 +1257,11 @@ int CommBrick::exchange_variable(int n, double *inbuf, double *&outbuf)
nsend = nrecv;
MPI_Sendrecv(&nsend,1,MPI_INT,procneigh[dim][0],0,
&nrecv1,1,MPI_INT,procneigh[dim][1],0,world,&status);
&nrecv1,1,MPI_INT,procneigh[dim][1],0,world,MPI_STATUS_IGNORE);
nrecv += nrecv1;
if (procgrid[dim] > 2) {
MPI_Sendrecv(&nsend,1,MPI_INT,procneigh[dim][1],0,
&nrecv2,1,MPI_INT,procneigh[dim][0],0,world,&status);
&nrecv2,1,MPI_INT,procneigh[dim][0],0,world,MPI_STATUS_IGNORE);
nrecv += nrecv2;
} else nrecv2 = 0;
@ -1284,13 +1270,13 @@ int CommBrick::exchange_variable(int n, double *inbuf, double *&outbuf)
MPI_Irecv(&buf_recv[nsend],nrecv1,MPI_DOUBLE,procneigh[dim][1],0,
world,&request);
MPI_Send(buf_recv,nsend,MPI_DOUBLE,procneigh[dim][0],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
if (procgrid[dim] > 2) {
MPI_Irecv(&buf_recv[nsend+nrecv1],nrecv2,MPI_DOUBLE,procneigh[dim][0],0,
world,&request);
MPI_Send(buf_recv,nsend,MPI_DOUBLE,procneigh[dim][1],0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
}

53
src/comm_tiled.cpp
View File

@ -436,9 +436,7 @@ void CommTiled::setup()
if (nmax > maxreqstat) {
maxreqstat = nmax;
delete [] requests;
delete [] statuses;
requests = new MPI_Request[maxreqstat];
statuses = new MPI_Status[maxreqstat];
}
}
@ -450,7 +448,6 @@ void CommTiled::setup()
void CommTiled::forward_comm(int dummy)
{
int i,irecv,n,nsend,nrecv;
MPI_Status status;
AtomVec *avec = atom->avec;
double **x = atom->x;
@ -483,7 +480,7 @@ void CommTiled::forward_comm(int dummy)
x[firstrecv[iswap][nrecv]],pbc_flag[iswap][nsend],
pbc[iswap][nsend]);
}
if (recvother[iswap]) MPI_Waitall(nrecv,requests,statuses);
if (recvother[iswap]) MPI_Waitall(nrecv,requests,MPI_STATUS_IGNORE);
} else if (ghost_velocity) {
if (recvother[iswap]) {
@ -507,7 +504,7 @@ void CommTiled::forward_comm(int dummy)
}
if (recvother[iswap]) {
for (i = 0; i < nrecv; i++) {
MPI_Waitany(nrecv,requests,&irecv,&status);
MPI_Waitany(nrecv,requests,&irecv,MPI_STATUS_IGNORE);
avec->unpack_comm_vel(recvnum[iswap][irecv],firstrecv[iswap][irecv],
&buf_recv[size_forward*
forward_recv_offset[iswap][irecv]]);
@ -536,7 +533,7 @@ void CommTiled::forward_comm(int dummy)
}
if (recvother[iswap]) {
for (i = 0; i < nrecv; i++) {
MPI_Waitany(nrecv,requests,&irecv,&status);
MPI_Waitany(nrecv,requests,&irecv,MPI_STATUS_IGNORE);
avec->unpack_comm(recvnum[iswap][irecv],firstrecv[iswap][irecv],
&buf_recv[size_forward*
forward_recv_offset[iswap][irecv]]);
@ -554,7 +551,6 @@ void CommTiled::forward_comm(int dummy)
void CommTiled::reverse_comm()
{
int i,irecv,n,nsend,nrecv;
MPI_Status status;
AtomVec *avec = atom->avec;
double **f = atom->f;
@ -588,7 +584,7 @@ void CommTiled::reverse_comm()
}
if (sendother[iswap]) {
for (i = 0; i < nsend; i++) {
MPI_Waitany(nsend,requests,&irecv,&status);
MPI_Waitany(nsend,requests,&irecv,MPI_STATUS_IGNORE);
avec->unpack_reverse(sendnum[iswap][irecv],sendlist[iswap][irecv],
&buf_recv[size_reverse*
reverse_recv_offset[iswap][irecv]]);
@ -617,7 +613,7 @@ void CommTiled::reverse_comm()
}
if (sendother[iswap]) {
for (i = 0; i < nsend; i++) {
MPI_Waitany(nsend,requests,&irecv,&status);
MPI_Waitany(nsend,requests,&irecv,MPI_STATUS_IGNORE);
avec->unpack_reverse(sendnum[iswap][irecv],sendlist[iswap][irecv],
&buf_recv[size_reverse*
reverse_recv_offset[iswap][irecv]]);
@ -736,7 +732,7 @@ void CommTiled::exchange()
exchproc[dim][m],0,world,&requests[m]);
for (m = 0; m < nexch; m++)
MPI_Send(&nsend,1,MPI_INT,exchproc[dim][m],0,world);
MPI_Waitall(nexch,requests,statuses);
MPI_Waitall(nexch,requests,MPI_STATUS_IGNORE);
nrecv = 0;
for (m = 0; m < nexch; m++) nrecv += exchnum[dim][m];
@ -750,7 +746,7 @@ void CommTiled::exchange()
}
for (m = 0; m < nexch; m++)
MPI_Send(buf_send,nsend,MPI_DOUBLE,exchproc[dim][m],0,world);
MPI_Waitall(nexch,requests,statuses);
MPI_Waitall(nexch,requests,MPI_STATUS_IGNORE);
// check incoming atoms to see if I own it and they are in my box
// if so, add to my list
@ -870,7 +866,7 @@ void CommTiled::borders()
for (m = 0; m < nsend; m++)
MPI_Send(&sendnum[iswap][m],1,MPI_INT,sendproc[iswap][m],0,world);
if (sendself[iswap]) recvnum[iswap][nrecv] = sendnum[iswap][nsend];
if (recvother[iswap]) MPI_Waitall(nrecv,requests,statuses);
if (recvother[iswap]) MPI_Waitall(nrecv,requests,MPI_STATUS_IGNORE);
// setup other per swap/proc values from sendnum and recvnum
@ -931,7 +927,7 @@ void CommTiled::borders()
buf_send);
}
if (recvother[iswap]) {
MPI_Waitall(nrecv,requests,statuses);
MPI_Waitall(nrecv,requests,MPI_STATUS_IGNORE);
for (m = 0; m < nrecv; m++)
avec->unpack_border_vel(recvnum[iswap][m],firstrecv[iswap][m],
&buf_recv[size_border*
@ -960,7 +956,7 @@ void CommTiled::borders()
buf_send);
}
if (recvother[iswap]) {
MPI_Waitall(nrecv,requests,statuses);
MPI_Waitall(nrecv,requests,MPI_STATUS_IGNORE);
for (m = 0; m < nrecv; m++)
avec->unpack_border(recvnum[iswap][m],firstrecv[iswap][m],
&buf_recv[size_border*
@ -997,7 +993,6 @@ void CommTiled::borders()
void CommTiled::forward_comm_pair(Pair *pair)
{
int i,irecv,n,nsend,nrecv;
MPI_Status status;
int nsize = pair->comm_forward;
@ -1029,7 +1024,7 @@ void CommTiled::forward_comm_pair(Pair *pair)
}
if (recvother[iswap]) {
for (i = 0; i < nrecv; i++) {
MPI_Waitany(nrecv,requests,&irecv,&status);
MPI_Waitany(nrecv,requests,&irecv,MPI_STATUS_IGNORE);
pair->unpack_forward_comm(recvnum[iswap][irecv],firstrecv[iswap][irecv],
&buf_recv[nsize*
forward_recv_offset[iswap][irecv]]);
@ -1046,7 +1041,6 @@ void CommTiled::forward_comm_pair(Pair *pair)
void CommTiled::reverse_comm_pair(Pair *pair)
{
int i,irecv,n,nsend,nrecv;
MPI_Status status;
int nsize = MAX(pair->comm_reverse,pair->comm_reverse_off);
@ -1075,7 +1069,7 @@ void CommTiled::reverse_comm_pair(Pair *pair)
}
if (sendother[iswap]) {
for (i = 0; i < nsend; i++) {
MPI_Waitany(nsend,requests,&irecv,&status);
MPI_Waitany(nsend,requests,&irecv,MPI_STATUS_IGNORE);
pair->unpack_reverse_comm(sendnum[iswap][irecv],sendlist[iswap][irecv],
&buf_recv[nsize*
reverse_recv_offset[iswap][irecv]]);
@ -1096,7 +1090,6 @@ void CommTiled::reverse_comm_pair(Pair *pair)
void CommTiled::forward_comm_fix(Fix *fix, int size)
{
int i,irecv,n,nsize,nsend,nrecv;
MPI_Status status;
if (size) nsize = size;
else nsize = fix->comm_forward;
@ -1127,7 +1120,7 @@ void CommTiled::forward_comm_fix(Fix *fix, int size)
}
if (recvother[iswap]) {
for (i = 0; i < nrecv; i++) {
MPI_Waitany(nrecv,requests,&irecv,&status);
MPI_Waitany(nrecv,requests,&irecv,MPI_STATUS_IGNORE);
fix->unpack_forward_comm(recvnum[iswap][irecv],firstrecv[iswap][irecv],
&buf_recv[nsize*
forward_recv_offset[iswap][irecv]]);
@ -1148,7 +1141,6 @@ void CommTiled::forward_comm_fix(Fix *fix, int size)
void CommTiled::reverse_comm_fix(Fix *fix, int size)
{
int i,irecv,n,nsize,nsend,nrecv;
MPI_Status status;
if (size) nsize = size;
else nsize = fix->comm_reverse;
@ -1178,7 +1170,7 @@ void CommTiled::reverse_comm_fix(Fix *fix, int size)
}
if (sendother[iswap]) {
for (i = 0; i < nsend; i++) {
MPI_Waitany(nsend,requests,&irecv,&status);
MPI_Waitany(nsend,requests,&irecv,MPI_STATUS_IGNORE);
fix->unpack_reverse_comm(sendnum[iswap][irecv],sendlist[iswap][irecv],
&buf_recv[nsize*
reverse_recv_offset[iswap][irecv]]);
@ -1195,7 +1187,6 @@ void CommTiled::reverse_comm_fix(Fix *fix, int size)
void CommTiled::forward_comm_compute(Compute *compute)
{
int i,irecv,n,nsend,nrecv;
MPI_Status status;
int nsize = compute->comm_forward;
@ -1226,7 +1217,7 @@ void CommTiled::forward_comm_compute(Compute *compute)
}
if (recvother[iswap]) {
for (i = 0; i < nrecv; i++) {
MPI_Waitany(nrecv,requests,&irecv,&status);
MPI_Waitany(nrecv,requests,&irecv,MPI_STATUS_IGNORE);
compute->
unpack_forward_comm(recvnum[iswap][irecv],firstrecv[iswap][irecv],
&buf_recv[nsize*
@ -1244,7 +1235,6 @@ void CommTiled::forward_comm_compute(Compute *compute)
void CommTiled::reverse_comm_compute(Compute *compute)
{
int i,irecv,n,nsend,nrecv;
MPI_Status status;
int nsize = compute->comm_reverse;
@ -1273,7 +1263,7 @@ void CommTiled::reverse_comm_compute(Compute *compute)
}
if (sendother[iswap]) {
for (i = 0; i < nsend; i++) {
MPI_Waitany(nsend,requests,&irecv,&status);
MPI_Waitany(nsend,requests,&irecv,MPI_STATUS_IGNORE);
compute->
unpack_reverse_comm(sendnum[iswap][irecv],sendlist[iswap][irecv],
&buf_recv[nsize*
@ -1291,7 +1281,6 @@ void CommTiled::reverse_comm_compute(Compute *compute)
void CommTiled::forward_comm_dump(Dump *dump)
{
int i,irecv,n,nsend,nrecv;
MPI_Status status;
int nsize = dump->comm_forward;
@ -1322,7 +1311,7 @@ void CommTiled::forward_comm_dump(Dump *dump)
}
if (recvother[iswap]) {
for (i = 0; i < nrecv; i++) {
MPI_Waitany(nrecv,requests,&irecv,&status);
MPI_Waitany(nrecv,requests,&irecv,MPI_STATUS_IGNORE);
dump->unpack_forward_comm(recvnum[iswap][irecv],firstrecv[iswap][irecv],
&buf_recv[nsize*
forward_recv_offset[iswap][irecv]]);
@ -1339,7 +1328,6 @@ void CommTiled::forward_comm_dump(Dump *dump)
void CommTiled::reverse_comm_dump(Dump *dump)
{
int i,irecv,n,nsend,nrecv;
MPI_Status status;
int nsize = dump->comm_reverse;
@ -1368,7 +1356,7 @@ void CommTiled::reverse_comm_dump(Dump *dump)
}
if (sendother[iswap]) {
for (i = 0; i < nsend; i++) {
MPI_Waitany(nsend,requests,&irecv,&status);
MPI_Waitany(nsend,requests,&irecv,MPI_STATUS_IGNORE);
dump->unpack_reverse_comm(sendnum[iswap][irecv],sendlist[iswap][irecv],
&buf_recv[nsize*
reverse_recv_offset[iswap][irecv]]);
@ -1384,7 +1372,6 @@ void CommTiled::reverse_comm_dump(Dump *dump)
void CommTiled::forward_comm_array(int nsize, double **array)
{
int i,j,k,m,iatom,last,irecv,nsend,nrecv;
MPI_Status status;
// insure send/recv bufs are big enough for nsize
// based on smaxone/rmaxall from most recent borders() invocation
@ -1435,7 +1422,7 @@ void CommTiled::forward_comm_array(int nsize, double **array)
if (recvother[iswap]) {
for (i = 0; i < nrecv; i++) {
MPI_Waitany(nrecv,requests,&irecv,&status);
MPI_Waitany(nrecv,requests,&irecv,MPI_STATUS_IGNORE);
m = nsize*forward_recv_offset[iswap][irecv];
last = firstrecv[iswap][irecv] + recvnum[iswap][irecv];
for (iatom = firstrecv[iswap][irecv]; iatom < last; iatom++)
@ -1919,7 +1906,6 @@ void CommTiled::allocate_swap(int n)
maxreqstat = 0;
requests = NULL;
statuses = NULL;
for (int i = 0; i < n; i++) {
nprocmax[i] = DELTA_PROCS;
@ -2044,7 +2030,6 @@ void CommTiled::deallocate_swap(int n)
delete [] sendlist;
delete [] requests;
delete [] statuses;
delete [] nprocmax;

1
src/comm_tiled.h
View File

@ -92,7 +92,6 @@ class CommTiled : public Comm {
int maxreqstat; // max size of Request and Status vectors
MPI_Request *requests;
MPI_Status *statuses;
struct RCBinfo {
double mysplit[3][2]; // fractional RCB bounding box for one proc

2
src/compute_vcm_molecule.cpp
View File

@ -97,7 +97,6 @@ void ComputeVCMMolecule::compute_array()
{
tagint imol;
double massone;
double unwrap[3];
invoked_array = update->ntimestep;
@ -108,7 +107,6 @@ void ComputeVCMMolecule::compute_array()
int *mask = atom->mask;
tagint *molecule = atom->molecule;
int *type = atom->type;
imageint *image = atom->image;
double *mass = atom->mass;
double *rmass = atom->rmass;
int nlocal = atom->nlocal;

4
src/dump.cpp
View File

@ -383,7 +383,7 @@ void Dump::write()
if (flush_flag) fflush(fp);
} else {
MPI_Recv(&tmp,0,MPI_INT,fileproc,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,fileproc,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(buf,nme*size_one,MPI_DOUBLE,fileproc,0,world);
}
@ -404,7 +404,7 @@ void Dump::write()
if (flush_flag) fflush(fp);
} else {
MPI_Recv(&tmp,0,MPI_INT,fileproc,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,fileproc,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(sbuf,nsme,MPI_CHAR,fileproc,0,world);
}
}

2
src/fix_recenter.h
View File

@ -38,8 +38,8 @@ class FixRecenter : public Fix {
int group2bit,scaleflag;
int xflag,yflag,zflag;
int xinitflag,yinitflag,zinitflag;
double xcom,ycom,zcom,xinit,yinit,zinit,masstotal,distance,shift[3];
int nlevels_respa;
double xcom,ycom,zcom,xinit,yinit,zinit,masstotal,distance,shift[3];
};
}

1
src/fix_temp_csvr.cpp
View File

@ -49,6 +49,7 @@ FixTempCSVR::FixTempCSVR(LAMMPS *lmp, int narg, char **arg) :
nevery = 1;
global_freq = nevery;
dynamic_group_allow = 1;
tstr = NULL;
if (strstr(arg[3],"v_") == arg[3]) {

5
src/image.cpp
View File

@ -294,7 +294,6 @@ void Image::clear()
void Image::merge()
{
MPI_Request requests[3];
MPI_Status statuses[3];
int nhalf = 1;
while (nhalf < nprocs) nhalf *= 2;
@ -307,8 +306,8 @@ void Image::merge()
if (ssao)
MPI_Irecv(surfacecopy,npixels*2,MPI_DOUBLE,
me+nhalf,0,world,&requests[2]);
if (ssao) MPI_Waitall(3,requests,statuses);
else MPI_Waitall(2,requests,statuses);
if (ssao) MPI_Waitall(3,requests,MPI_STATUS_IGNORE);
else MPI_Waitall(2,requests,MPI_STATUS_IGNORE);
for (int i = 0; i < npixels; i++) {
if (depthBuffer[i] < 0 || (depthcopy[i] >= 0 &&

7
src/procmap.cpp
View File

@ -679,7 +679,6 @@ void ProcMap::output(char *file, int *procgrid, int ***grid2proc)
int tmp;
int vec[6];
char procname[MPI_MAX_PROCESSOR_NAME+1];
MPI_Status status;
vec[0] = me;
vec[1] = universe->me;
@ -696,9 +695,9 @@ void ProcMap::output(char *file, int *procgrid, int ***grid2proc)
for (int iproc = 0; iproc < nprocs; iproc++) {
if (iproc) {
MPI_Send(&tmp,0,MPI_INT,iproc,0,world);
MPI_Recv(vec,6,MPI_INT,iproc,0,world,&status);
MPI_Recv(vec,6,MPI_INT,iproc,0,world,MPI_STATUS_IGNORE);
MPI_Recv(procname,MPI_MAX_PROCESSOR_NAME+1,MPI_CHAR,
iproc,0,world,&status);
iproc,0,world,MPI_STATUS_IGNORE);
}
fprintf(fp,"%d %d %d: %d %d %d: %s\n",
@ -706,7 +705,7 @@ void ProcMap::output(char *file, int *procgrid, int ***grid2proc)
}
} else {
MPI_Recv(&tmp,0,MPI_INT,0,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,0,0,world,MPI_STATUS_IGNORE);
MPI_Send(vec,6,MPI_INT,0,0,world);
MPI_Send(procname,strlen(procname)+1,MPI_CHAR,0,0,world);
}

11
src/rcb.cpp
View File

@ -115,7 +115,6 @@ void RCB::compute(int dimension, int n, double **x, double *wt,
double tolerance;
MPI_Comm comm,comm_half;
MPI_Request request,request2;
MPI_Status status;
Median med,medme;
// create list of my Dots
@ -469,9 +468,9 @@ void RCB::compute(int dimension, int n, double **x, double *wt,
MPI_Send(&outgoing,1,MPI_INT,procpartner,0,world);
incoming = 0;
if (readnumber) {
MPI_Recv(&incoming,1,MPI_INT,procpartner,0,world,&status);
MPI_Recv(&incoming,1,MPI_INT,procpartner,0,world,MPI_STATUS_IGNORE);
if (readnumber == 2) {
MPI_Recv(&incoming2,1,MPI_INT,procpartner2,0,world,&status);
MPI_Recv(&incoming2,1,MPI_INT,procpartner2,0,world,MPI_STATUS_IGNORE);
incoming += incoming2;
}
}
@ -527,7 +526,7 @@ void RCB::compute(int dimension, int n, double **x, double *wt,
MPI_Send(NULL,0,MPI_INT,procpartner,0,world);
if (readnumber == 2) MPI_Send(NULL,0,MPI_INT,procpartner2,0,world);
}
MPI_Recv(NULL,0,MPI_INT,procpartner,0,world,&status);
MPI_Recv(NULL,0,MPI_INT,procpartner,0,world,MPI_STATUS_IGNORE);
// send dots to partner
@ -536,8 +535,8 @@ void RCB::compute(int dimension, int n, double **x, double *wt,
// wait until all dots are received
if (readnumber > 0) {
MPI_Wait(&request,&status);
if (readnumber == 2) MPI_Wait(&request2,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
if (readnumber == 2) MPI_Wait(&request2,MPI_STATUS_IGNORE);
}
ndot = ndotnew;

7
src/read_restart.cpp
View File

@ -369,7 +369,6 @@ void ReadRestart::command(int narg, char **arg)
MPI_Bcast(&procsperfile,1,MPI_INT,0,clustercomm);
int tmp,iproc;
MPI_Status status;
MPI_Request request;
for (int i = 0; i < procsperfile; i++) {
@ -389,12 +388,12 @@ void ReadRestart::command(int narg, char **arg)
if (i % nclusterprocs) {
iproc = me + (i % nclusterprocs);
MPI_Send(&n,1,MPI_INT,iproc,0,world);
MPI_Recv(&tmp,0,MPI_INT,iproc,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,iproc,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(buf,n,MPI_DOUBLE,iproc,0,world);
}
} else if (i % nclusterprocs == me - fileproc) {
MPI_Recv(&n,1,MPI_INT,fileproc,0,world,&status);
MPI_Recv(&n,1,MPI_INT,fileproc,0,world,MPI_STATUS_IGNORE);
if (n > maxbuf) {
maxbuf = n;
memory->destroy(buf);
@ -402,7 +401,7 @@ void ReadRestart::command(int narg, char **arg)
}
MPI_Irecv(buf,n,MPI_DOUBLE,fileproc,0,world,&request);
MPI_Send(&tmp,0,MPI_INT,fileproc,0,world);
MPI_Wait(&request,&status);
MPI_Wait(&request,MPI_STATUS_IGNORE);
}
if (i % nclusterprocs == me - fileproc) {

2
src/variable.h
View File

@ -1,4 +1,4 @@
/* ----------------------------------------------------------------------
/* -*- c++ -*- ----------------------------------------------------------
LAMMPS - Large-scale Atomic/Molecular Massively Parallel Simulator
http://lammps.sandia.gov, Sandia National Laboratories
Steve Plimpton, sjplimp@sandia.gov

49
src/write_data.cpp
View File

@ -328,10 +328,11 @@ void WriteData::atoms()
// all other procs wait for ping, send their chunk to proc 0
int tmp,recvrow;
MPI_Status status;
MPI_Request request;
if (me == 0) {
MPI_Status status;
MPI_Request request;
fprintf(fp,"\nAtoms # %s\n\n",atom->atom_style);
for (int iproc = 0; iproc < nprocs; iproc++) {
if (iproc) {
@ -346,7 +347,7 @@ void WriteData::atoms()
}
} else {
MPI_Recv(&tmp,0,MPI_INT,0,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,0,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(&buf[0][0],sendrow*ncol,MPI_DOUBLE,0,0,world);
}
@ -381,10 +382,11 @@ void WriteData::velocities()
// all other procs wait for ping, send their chunk to proc 0
int tmp,recvrow;
MPI_Status status;
MPI_Request request;
if (me == 0) {
MPI_Status status;
MPI_Request request;
fprintf(fp,"\nVelocities\n\n");
for (int iproc = 0; iproc < nprocs; iproc++) {
if (iproc) {
@ -399,7 +401,7 @@ void WriteData::velocities()
}
} else {
MPI_Recv(&tmp,0,MPI_INT,0,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,0,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(&buf[0][0],sendrow*ncol,MPI_DOUBLE,0,0,world);
}
@ -432,11 +434,12 @@ void WriteData::bonds()
// all other procs wait for ping, send their chunk to proc 0
int tmp,recvrow;
MPI_Status status;
MPI_Request request;
int index = 1;
if (me == 0) {
MPI_Status status;
MPI_Request request;
fprintf(fp,"\nBonds\n\n");
for (int iproc = 0; iproc < nprocs; iproc++) {
if (iproc) {
@ -452,7 +455,7 @@ void WriteData::bonds()
}
} else {
MPI_Recv(&tmp,0,MPI_INT,0,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,0,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(&buf[0][0],sendrow*ncol,MPI_LMP_TAGINT,0,0,world);
}
@ -485,11 +488,12 @@ void WriteData::angles()
// all other procs wait for ping, send their chunk to proc 0
int tmp,recvrow;
MPI_Status status;
MPI_Request request;
int index = 1;
if (me == 0) {
MPI_Status status;
MPI_Request request;
fprintf(fp,"\nAngles\n\n");
for (int iproc = 0; iproc < nprocs; iproc++) {
if (iproc) {
@ -505,7 +509,7 @@ void WriteData::angles()
}
} else {
MPI_Recv(&tmp,0,MPI_INT,0,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,0,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(&buf[0][0],sendrow*ncol,MPI_LMP_TAGINT,0,0,world);
}
@ -556,11 +560,12 @@ void WriteData::dihedrals()
// all other procs wait for ping, send their chunk to proc 0
int tmp,recvrow;
MPI_Status status;
MPI_Request request;
int index = 1;
if (me == 0) {
MPI_Status status;
MPI_Request request;
fprintf(fp,"\nDihedrals\n\n");
for (int iproc = 0; iproc < nprocs; iproc++) {
if (iproc) {
@ -576,7 +581,7 @@ void WriteData::dihedrals()
}
} else {
MPI_Recv(&tmp,0,MPI_INT,0,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,0,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(&buf[0][0],sendrow*ncol,MPI_LMP_TAGINT,0,0,world);
}
@ -627,11 +632,12 @@ void WriteData::impropers()
// all other procs wait for ping, send their chunk to proc 0
int tmp,recvrow;
MPI_Status status;
MPI_Request request;
int index = 1;
if (me == 0) {
MPI_Status status;
MPI_Request request;
fprintf(fp,"\nImpropers\n\n");
for (int iproc = 0; iproc < nprocs; iproc++) {
if (iproc) {
@ -647,7 +653,7 @@ void WriteData::impropers()
}
} else {
MPI_Recv(&tmp,0,MPI_INT,0,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,0,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(&buf[0][0],sendrow*ncol,MPI_LMP_TAGINT,0,0,world);
}
@ -680,11 +686,12 @@ void WriteData::fix(int ifix, int mth)
// all other procs wait for ping, send their chunk to proc 0
int tmp,recvrow;
MPI_Status status;
MPI_Request request;
int index = 1;
if (me == 0) {
MPI_Status status;
MPI_Request request;
modify->fix[ifix]->write_data_section_keyword(mth,fp);
for (int iproc = 0; iproc < nprocs; iproc++) {
if (iproc) {
@ -700,7 +707,7 @@ void WriteData::fix(int ifix, int mth)
}
} else {
MPI_Recv(&tmp,0,MPI_INT,0,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,0,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(&buf[0][0],sendrow*ncol,MPI_DOUBLE,0,0,world);
}

6
src/write_restart.cpp
View File

@ -393,10 +393,10 @@ void WriteRestart::write(char *file)
else {
int tmp,recv_size;
MPI_Status status;
MPI_Request request;
if (filewriter) {
MPI_Status status;
MPI_Request request;
for (int iproc = 0; iproc < nclusterprocs; iproc++) {
if (iproc) {
MPI_Irecv(buf,max_size,MPI_DOUBLE,me+iproc,0,world,&request);
@ -410,7 +410,7 @@ void WriteRestart::write(char *file)
fclose(fp);
} else {
MPI_Recv(&tmp,0,MPI_INT,fileproc,0,world,&status);
MPI_Recv(&tmp,0,MPI_INT,fileproc,0,world,MPI_STATUS_IGNORE);
MPI_Rsend(buf,send_size,MPI_DOUBLE,fileproc,0,world);
}
}