/**
* Builds the full MomHamiltonian matrix
*/
void MomHamiltonian::BuildFullHam()
{
if( !baseUp.size() || !baseDown.size() )
{
std::cerr << "Build base before building MomHamiltonian" << std::endl;
return;
}
blockmat.resize(L);
for(int B=0; B<L; B++)
{
int cur_dim = mombase[B].size();
blockmat[B].reset(new double [cur_dim*cur_dim]);
for(int i=0; i<cur_dim; i++)
{
int a = mombase[B][i].first;
int b = mombase[B][i].second;
for(int j=i; j<cur_dim; j++)
{
int c = mombase[B][j].first;
int d = mombase[B][j].second;
blockmat[B][j+cur_dim*i] = U/L*interaction(a,b,c,d);
blockmat[B][i+cur_dim*j] = blockmat[B][j+cur_dim*i];
}
blockmat[B][i+cur_dim*i] += J * (hopping(baseUp[a]) + hopping(baseDown[b]));
}
}
}
/**
* This methods calculates the eigenvalues for a range of U values. The interval is [Ubegin, Uend]
* with a stepsize of step. The resulting data is written to a file in the HDF5 file format.
* @param Ubegin the startpoint for U
* @param Uend the endpoint for U. We demand Ubegin < Uend
* @param step the stepsize to use for U
* @param filename the name of the file write to written the eigenvalues to
*/
void MomHamiltonian::GenerateData(double Ubegin, double Uend, double step, std::string filename)
{
double Ucur = Ubegin;
if( !baseUp.size() || !baseDown.size() )
BuildBase();
std::vector<double> eigenvalues(dim);
hid_t file_id, group_id, dataset_id, dataspace_id, attribute_id;
herr_t status;
file_id = H5Fcreate(filename.c_str(), H5F_ACC_TRUNC, H5P_DEFAULT, H5P_DEFAULT);
HDF5_STATUS_CHECK(file_id);
group_id = H5Gcreate(file_id, "run", H5P_DEFAULT, H5P_DEFAULT, H5P_DEFAULT);
HDF5_STATUS_CHECK(group_id);
dataspace_id = H5Screate(H5S_SCALAR);
attribute_id = H5Acreate (group_id, "L", H5T_STD_I32LE, dataspace_id, H5P_DEFAULT, H5P_DEFAULT);
status = H5Awrite (attribute_id, H5T_NATIVE_INT, &L );
HDF5_STATUS_CHECK(status);
status = H5Aclose(attribute_id);
HDF5_STATUS_CHECK(status);
attribute_id = H5Acreate (group_id, "Nu", H5T_STD_I32LE, dataspace_id, H5P_DEFAULT, H5P_DEFAULT);
status = H5Awrite (attribute_id, H5T_NATIVE_INT, &Nu );
HDF5_STATUS_CHECK(status);
status = H5Aclose(attribute_id);
HDF5_STATUS_CHECK(status);
attribute_id = H5Acreate (group_id, "Nd", H5T_STD_I32LE, dataspace_id, H5P_DEFAULT, H5P_DEFAULT);
status = H5Awrite (attribute_id, H5T_NATIVE_INT, &Nd );
HDF5_STATUS_CHECK(status);
status = H5Aclose(attribute_id);
HDF5_STATUS_CHECK(status);
attribute_id = H5Acreate (group_id, "J", H5T_IEEE_F64LE, dataspace_id, H5P_DEFAULT, H5P_DEFAULT);
status = H5Awrite (attribute_id, H5T_NATIVE_DOUBLE, &J );
HDF5_STATUS_CHECK(status);
status = H5Aclose(attribute_id);
HDF5_STATUS_CHECK(status);
attribute_id = H5Acreate (group_id, "Ubegin", H5T_IEEE_F64LE, dataspace_id, H5P_DEFAULT, H5P_DEFAULT);
status = H5Awrite (attribute_id, H5T_NATIVE_DOUBLE, &Ubegin );
HDF5_STATUS_CHECK(status);
status = H5Aclose(attribute_id);
HDF5_STATUS_CHECK(status);
attribute_id = H5Acreate (group_id, "Uend", H5T_IEEE_F64LE, dataspace_id, H5P_DEFAULT, H5P_DEFAULT);
status = H5Awrite (attribute_id, H5T_NATIVE_DOUBLE, &Uend );
HDF5_STATUS_CHECK(status);
status = H5Aclose(attribute_id);
HDF5_STATUS_CHECK(status);
attribute_id = H5Acreate (group_id, "Ustep", H5T_IEEE_F64LE, dataspace_id, H5P_DEFAULT, H5P_DEFAULT);
status = H5Awrite (attribute_id, H5T_NATIVE_DOUBLE, &step );
HDF5_STATUS_CHECK(status);
status = H5Aclose(attribute_id);
HDF5_STATUS_CHECK(status);
status = H5Sclose(dataspace_id);
HDF5_STATUS_CHECK(status);
status = H5Gclose(group_id);
HDF5_STATUS_CHECK(status);
status = H5Fclose(file_id);
HDF5_STATUS_CHECK(status);
std::vector<double> diagonalelements(dim);
std::vector< std::unique_ptr<double []> > offdiag;
offdiag.resize(L);
// make sure that we don't rebuild the whole hamiltonian every time.
// store the hopping and interaction part seperate so we can just
// add them in every step
#pragma omp parallel for
for(int B=0; B<L; B++)
{
int cur_dim = mombase[B].size();
int offset = 0;
for(int tmp=0; tmp<B; tmp++)
offset += mombase[tmp].size();
offdiag[B].reset(new double [cur_dim*cur_dim]);
for(int i=0; i<cur_dim; i++)
{
int a = mombase[B][i].first;
int b = mombase[B][i].second;
diagonalelements[offset+i] = hopping(baseUp[a]) + hopping(baseDown[b]);
for(int j=i; j<cur_dim; j++)
{
int c = mombase[B][j].first;
int d = mombase[B][j].second;
offdiag[B][j+cur_dim*i] = 1.0/L*interaction(a,b,c,d);
offdiag[B][i+cur_dim*j] = offdiag[B][j+cur_dim*i];
}
}
}
while(Ucur <= Uend)
{
std::cout << "U = " << Ucur << std::endl;
setU(Ucur);
// make hamiltonian
#pragma omp parallel for
for(int B=0; B<L; B++)
{
int cur_dim = mombase[B].size();
int offset = 0;
for(int tmp=0; tmp<B; tmp++)
offset += mombase[tmp].size();
std::memcpy(blockmat[B].get(),offdiag[B].get(),cur_dim*cur_dim*sizeof(double));
int tmp = cur_dim*cur_dim;
int inc = 1;
dscal_(&tmp,&Ucur,blockmat[B].get(),&inc);
for(int i=0; i<cur_dim; i++)
blockmat[B][i+cur_dim*i] += diagonalelements[offset+i];
}
#pragma omp parallel for
for(int B=0; B<L; B++)
{
int dim = mombase[B].size();
int offset = 0;
for(int tmp=0; tmp<B; tmp++)
offset += mombase[tmp].size();
Diagonalize(dim, blockmat[B].get(), &eigenvalues[offset], false);
}
hid_t U_id;
std::stringstream name;
name << std::setprecision(5) << std::fixed << "/run/" << Ucur;
file_id = H5Fopen(filename.c_str(), H5F_ACC_RDWR, H5P_DEFAULT);
HDF5_STATUS_CHECK(file_id);
U_id = H5Gcreate(file_id, name.str().c_str(), H5P_DEFAULT, H5P_DEFAULT, H5P_DEFAULT);
HDF5_STATUS_CHECK(U_id);
for(int B=0; B<L; B++)
{
int dim = mombase[B].size();
int offset = 0;
for(int tmp=0; tmp<B; tmp++)
offset += mombase[tmp].size();
hsize_t dimarr = dim;
dataspace_id = H5Screate_simple(1, &dimarr, NULL);
std::stringstream cur_block;
cur_block << B;
dataset_id = H5Dcreate(U_id, cur_block.str().c_str(), H5T_IEEE_F64LE, dataspace_id, H5P_DEFAULT, H5P_DEFAULT, H5P_DEFAULT);
status = H5Dwrite(dataset_id, H5T_NATIVE_DOUBLE, H5S_ALL, H5S_ALL, H5P_DEFAULT, &eigenvalues[offset] );
HDF5_STATUS_CHECK(status);
status = H5Sclose(dataspace_id);
HDF5_STATUS_CHECK(status);
status = H5Dclose(dataset_id);
HDF5_STATUS_CHECK(status);
}
status = H5Gclose(U_id);
HDF5_STATUS_CHECK(status);
status = H5Fclose(file_id);
HDF5_STATUS_CHECK(status);
Ucur += step;
}
}
int main(int argc, char **argv)
{
int meascount[MAX_NKAP];
int prompt, count1, count2;
Real avm_iters[MAX_NKAP];
double starttime, endtime;
int MaxMR, restart_flag;
Real RsdMR; /******/
int spin, color, nk; /******/
int max_prop;
int cl_cg = CL_CG;
double ssplaq, stplaq;
FILE *fp_m_out = NULL; /*** meson IO stuff **/
int fb_m_out = 0; /*** meson IO stuff **/
w_prop_file *fp_in_w[MAX_NKAP]; /* For propagator files */
w_prop_file *fp_out_w[MAX_NKAP]; /* For propagator files */
double g_time ;
int i ;
int MinMR;
/*** variables required for the static variational code ***/
int nodata = 0 ;
complex *meson = NULL;
/****** start of the execution of the code ************/
initialize_machine(&argc, &argv);
/* Remap standard I/O */
if(remap_stdio_from_args(argc, argv) == 1)terminate(1);
g_sync();
/* set up */
prompt = setup_h();
/**DEBUG***/
#ifdef DEBUGDEF
light_quark_pion(0) ;
#endif
/* loop over input sets */
while( readin(prompt) == 0)
{
if( fixflag == COULOMB_GAUGE_FIX)
{
if(this_node == 0)
printf("Fixing to Coulomb gauge\n");
g_time = -dclock();
gaugefix(TUP,(Real)1.5,500,GAUGE_FIX_TOL);
g_time += dclock();
if(this_node==0)printf("Time to gauge fix = %e\n",g_time);
invalidate_this_clov(gen_clov);
}
else
if(this_node == 0)printf("COULOMB GAUGE FIXING SKIPPED.\n");
/* save lattice if requested */
if( saveflag != FORGET )
{
save_lattice( saveflag, savefile, stringLFN );
}
/* call plaquette measuring process */
d_plaquette(&ssplaq, &stplaq);
if (this_node == 0)
printf("START %e %e\n",(double) ssplaq, (double) stplaq);
/******* set up code for the static variational calculation *****/
if( nkap == 1 )
{
nodata = nt*nosmear*144 ;
/** reserve memory for the smeared meson correlators on each node ****/
if( ( meson = (complex *) calloc( (size_t) nodata, sizeof(complex) ) ) == NULL )
{
printf("ERROR: could not reserve buffer space for the meson smearing functions\n");
terminate(1);
}
/** call a number of set up routines for the static variational code ***/
setup_vary(meson, nodata);
} /*** end of set up section for the static-variational calculation ***/
/***DEBUG check_calc_matrix() ; ****/
starttime=dclock();
MaxMR = niter;
RsdMR = (Real) sqrt((double) rsqprop);
if (this_node == 0)
printf("Residue=%e\n",(double) RsdMR);
for (nk = 0; nk < nkap; nk++)
{
avm_iters[nk] = 0.0;
meascount[nk] = 0;
}
max_prop = 12;
count1 = 0;
count2 = 0;
for (spin = start_spin; spin < 4; spin++)
{
for (color = 0; color < 3; color++)
{
count1++;
if (count1 == 1)
color += start_color;
for (nk = 0; nk < nkap; nk++)
{
count2++;
if (count2 == 1)
nk += start_kap;
kappa = cappa[nk];
meascount[nk]++;
/* open file for wilson propagators */
fp_in_w[nk] = r_open_wprop(startflag_w[nk], startfile_w[nk]);
if ((spin + color) == 0)
{
/*** first pass of the code **/
fp_out_w[nk] = w_open_wprop(saveflag_w[nk], savefile_w[nk],
wqs.type);
/* open file for meson output and write the header */
if (saveflag_m == SAVE_MESON_ASCII)
{
fp_m_out = w_ascii_m_i(savefile_m[nk], max_prop);
fb_m_out = -1; /* i.e. file is NOT binary */
}
else if (saveflag_m == SAVE_MESON_BINARY)
{
fb_m_out = w_binary_m_i(savefile_m[nk], max_prop);
fp_m_out = NULL; /* i.e. file is NOT ascii */
}
else
{
if( this_node == 0 )
printf("ERROR in main saveflag_m = %d is out of range in initial opening\n",saveflag_m) ;
terminate(1);
}
} /*** end of spin =0 && color == 0 **/
else
{
fp_out_w[nk] = w_open_wprop(saveflag_w[nk], savefile_w[nk],
wqs.type);
/* open file for meson output for appending output*/
if (saveflag_m == SAVE_MESON_ASCII)
{
fp_m_out = a_ascii_m_i(savefile_m[nk], max_prop);
fb_m_out = -1; /* i.e. file is NOT binary */
}
if (saveflag_m == SAVE_MESON_BINARY)
{
fb_m_out = a_binary_m_i(savefile_m[nk], max_prop);
fp_m_out = NULL; /* i.e. file is NOT ascii */
}
else
{
if( this_node == 0 )
printf("ERROR in main saveflag_m = %d is out of range in appending opening\n",saveflag_m) ;
terminate(1);
}
} /*** end of spin && color not equal to zero ***/
if (this_node == 0)
printf("color=%d spin=%d kappa=%f nk=%d\n", color, spin, (double) kappa, nk);
/* load psi if requested */
init_qs(&wqstmp2);
reload_wprop_sc_to_site(startflag_w[nk], fp_in_w[nk],&wqstmp2,
spin, color, F_OFFSET(psi),1);
if (nk == 0 || count2 == 1)
restart_flag = flag;
else
restart_flag = 1;
/* Conjugate gradient inversion uses site structure
temporary"chi" */
/* Complete the source structure */
wqs.color = color;
wqs.spin = spin;
wqs.parity = EVENANDODD;
/* For wilson_info */
wqstmp = wqs;
/* If we are starting fresh, we want to set a mininum number of
iterations */
if(startflag_w[nk] == FRESH)MinMR = nt/2; else MinMR = 0;
/* Load inversion control structure */
qic.prec = PRECISION;
qic.min = MinMR;
qic.max = MaxMR;
qic.nrestart = nrestart;
qic.parity = EVENANDODD;
qic.start_flag = restart_flag;
qic.nsrc = 1;
qic.resid = RsdMR;
qic.relresid = 0;
/* Load Dirac matrix parameters */
dwp.Kappa = kappa;
switch (cl_cg) {
case CG:
/* Load temporaries specific to inverter */
/* compute the propagator. Result in psi. */
avm_iters[nk] +=
(Real)wilson_invert_site_wqs(F_OFFSET(chi),F_OFFSET(psi),
w_source_h,&wqs,
cgilu_w_site,&qic,(void *)&dwp);
break;
case MR:
/* Load temporaries specific to inverter */
/* compute the propagator. Result in psi. */
avm_iters[nk] +=
(Real)wilson_invert_site_wqs(F_OFFSET(chi),F_OFFSET(psi),
w_source_h,&wqs,
mrilu_w_site,&qic,(void *)&dwp);
break;
default:
node0_printf("main(%d): Inverter choice %d not supported\n",
this_node,cl_cg);
}
/* save psi if requested */
save_wprop_sc_from_site( saveflag_w[nk],fp_out_w[nk], &wqstmp2,
spin,color,F_OFFSET(psi),1);
light_meson(F_OFFSET(psi), color, spin, wqs.type, fp_m_out, fb_m_out);
if (this_node == 0)
printf("Light mesons found\n");
/*** calculate the correlators required for the static variational code **/
if( nkap == 1 )
{
/** calculate the smeared meson correlators required for Bparam **/
calc_smeared_meson(meson, F_OFFSET(psi) , F_OFFSET(mp), color, spin);
/** calculate the object required for the 2-pt variational calculation ***/
buildup_strip(F_OFFSET(psi) , color, spin);
} /** end of the partial calculations for the variationl project ***/
/*
* find source again since mrilu overwrites it; for hopping
* expansion
*/
/* source must be of definite parity */
wqs.parity = source_parity;
w_source_h(F_OFFSET(chi), &wqs);
hopping(F_OFFSET(chi), F_OFFSET(mp), F_OFFSET(psi), nhop,
kappa_c, wqs.parity, color, spin, wqs.type,
fp_m_out, fb_m_out);
/* close files */
r_close_wprop(startflag_w[nk], fp_in_w[nk]);
w_close_wprop(saveflag_w[nk],fp_out_w[nk]);
if (saveflag_m == SAVE_MESON_ASCII)
w_ascii_m_f(fp_m_out, savefile_m[nk]);
else if (saveflag_m == SAVE_MESON_BINARY)
w_binary_m_f(fb_m_out, savefile_m[nk]);
if (spin == end_spin && color == end_color && nk == end_kap)
goto end_of_loops;
}
}
} /* end of loop over spin, color, kappa */
end_of_loops:
if (this_node == 0)
printf("RUNNING COMPLETED\n");
/**DEBUG***/
#ifdef DEBUGDEF
light_quark_pion(2) ;
#endif
for (nk = 0; nk < nkap; nk++)
{
if (meascount[nk] > 0)
{
if (this_node == 0)
printf("total mr iters for measurement= %e\n",
(double) avm_iters[nk]);
if (this_node == 0)
printf("average mr iters per spin-color= %e\n",
(double) avm_iters[nk] / (double) meascount[nk]);
}
}
endtime=dclock();
node0_printf("Time = %e seconds\n", (double) (endtime - starttime));
fflush(stdout);
/*** calculation section for the variational code *****/
if( nkap == 1 )
{
/** sum up the smeared meson correlators over all the nodes ***/
for(i=0 ; i < nodata ;++i)
{
g_complexsum(meson + i) ;
}
/* write the smeared correlators to a single disk file ***/
IF_MASTER
write_smear_mesonx(meson);
free(meson); /*** free up the memory for the b-parameter correlators ***/
calc_vary_matrix() ; /** calculate the static variational matrix **/
node0_printf(">> The end of the static variational code <<<<\n");
}/** end of the final static variational code *****/
node0_printf("Time = %e seconds\n",(double)(endtime-starttime));
fflush(stdout);
} /* end of while(prompt) */
return 0;
} /* end of main() */
