int main(){
su3_matrix FW, U, V, W, Fdiag, ILambda;
int i,j;
printf("Enter V\n");
for(i = 0; i < 3; i++)
for(j = 0; j < 3; j++)
scanf("%lf %lf",&V.e[i][j].real, &V.e[i][j].imag);
printf("Enter U\n");
for(i = 0; i < 3; i++)
for(j = 0; j < 3; j++)
scanf("%lf %lf",&U.e[i][j].real, &U.e[i][j].imag);
printf("Enter FW\n");
for(i = 0; i < 3; i++)
for(j = 0; j < 3; j++)
scanf("%lf %lf",&FW.e[i][j].real, &FW.e[i][j].imag);
stout_smear( &W, &V, &U);
printf("Result W\n");
for(i = 0; i < 3; i++){
for(j = 0; j < 3; j++)
printf("(%f, %f) ",W.e[i][j].real, W.e[i][j].imag);
printf("\n");
}
stout_force_terms(&Fdiag, &ILambda, &V, &U, &FW);
printf("Result Fdiag\n");
for(i = 0; i < 3; i++){
for(j = 0; j < 3; j++)
printf("(%f, %f) ",Fdiag.e[i][j].real, Fdiag.e[i][j].imag);
printf("\n");
}
printf("Result ILambda\n");
for(i = 0; i < 3; i++){
for(j = 0; j < 3; j++)
printf("(%f, %f) ",ILambda.e[i][j].real, ILambda.e[i][j].imag);
printf("\n");
}
}
int main(int argc,char *argv[]) {
FILE *parameterfile=NULL;
int c, j, is=0, ic=0;
int x, X, y, Y, z, Z, t, tt, i, sum;
char * filename = NULL;
char datafilename[50];
char parameterfilename[50];
char conf_filename[50];
char * input_filename = NULL;
double plaquette_energy, nrm;
double * norm;
struct stout_parameters params_smear;
#ifdef _GAUGE_COPY
int kb=0;
#endif
#ifdef MPI
double atime=0., etime=0.;
#endif
#ifdef _KOJAK_INST
#pragma pomp inst init
#pragma pomp inst begin(main)
#endif
DUM_DERI = 6;
/* DUM_DERI + 2 is enough (not 7) */
DUM_SOLVER = DUM_DERI+2;
DUM_MATRIX = DUM_SOLVER+6;
/* DUM_MATRIX + 2 is enough (not 6) */
NO_OF_SPINORFIELDS = DUM_MATRIX+2;
verbose = 0;
g_use_clover_flag = 0;
g_nr_of_psf = 1;
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?f:o:")) != -1) {
switch (c) {
case 'f':
input_filename = calloc(200, sizeof(char));
strcpy(input_filename,optarg);
break;
case 'o':
filename = calloc(200, sizeof(char));
strcpy(filename,optarg);
break;
case 'h':
case '?':
default:
usage();
break;
}
}
if(input_filename == NULL){
input_filename = "hmc.input";
}
if(filename == NULL){
filename = "output";
}
/* Read the input file */
read_input(input_filename);
/* here we want no even/odd preconditioning */
even_odd_flag = 0;
/* this DBW2 stuff is not needed for the inversion ! */
g_rgi_C1 = 0;
if(Nsave == 0){
Nsave = 1;
}
tmlqcd_mpi_init(argc, argv);
g_dbw2rand = 0;
#ifndef MPI
g_dbw2rand = 0;
#endif
#ifdef _GAUGE_COPY
j = init_gauge_field(VOLUMEPLUSRAND, 1);
#else
j = init_gauge_field(VOLUMEPLUSRAND, 0);
#endif
if ( j!= 0) {
fprintf(stderr, "Not enough memory for gauge_fields! Aborting...\n");
exit(-1);
}
j = init_geometry_indices(VOLUMEPLUSRAND);
if ( j!= 0) {
fprintf(stderr, "Not enough memory for geometry indices! Aborting...\n");
exit(-1);
}
if(even_odd_flag) {
j = init_spinor_field(VOLUMEPLUSRAND/2, NO_OF_SPINORFIELDS);
}
else {
j = init_spinor_field(VOLUMEPLUSRAND, NO_OF_SPINORFIELDS);
}
if ( j!= 0) {
fprintf(stderr, "Not enough memory for spinor fields! Aborting...\n");
exit(-1);
}
g_mu = g_mu1;
if(g_proc_id == 0){
/*construct the filenames for the observables and the parameters*/
strcpy(datafilename,filename); strcat(datafilename,".data");
strcpy(parameterfilename,filename); strcat(parameterfilename,".para");
parameterfile=fopen(parameterfilename, "w");
write_first_messages(parameterfile, 0, 1);
}
/* define the geometry */
geometry();
/* define the boundary conditions for the fermion fields */
boundary();
#ifdef _USE_HALFSPINOR
j = init_dirac_halfspinor();
if ( j!= 0) {
fprintf(stderr, "Not enough memory for halffield! Aborting...\n");
exit(-1);
}
if(g_sloppy_precision_flag == 1) {
j = init_dirac_halfspinor32();
if ( j!= 0) {
fprintf(stderr, "Not enough memory for 32-Bit halffield! Aborting...\n");
exit(-1);
}
}
# if (defined _PERSISTENT)
init_xchange_halffield();
# endif
#endif
norm = (double*)calloc(3.*LX/2.+T/2., sizeof(double));
for(j=0;j<Nmeas; j++) {
sprintf(conf_filename,"%s.%.4d", gauge_input_filename, nstore);
if (g_proc_id == 0){
printf("Reading Gauge field from file %s\n", conf_filename); fflush(stdout);
}
read_lime_gauge_field(conf_filename);
if (g_proc_id == 0){
printf("done!\n"); fflush(stdout);
}
#ifdef MPI
xchange_gauge();
#endif
#ifdef _GAUGE_COPY
update_backward_gauge();
#endif
/* Compute minimal eigenvalues, if wanted */
if(compute_evs != 0) {
eigenvalues(&no_eigenvalues, 1000, eigenvalue_precision, 0, compute_evs, nstore, even_odd_flag);
}
/*compute the energy of the gauge field*/
plaquette_energy = measure_gauge_action();
if(g_proc_id == 0) {
printf("The plaquette value is %e\n", plaquette_energy/(6.*VOLUME*g_nproc)); fflush(stdout);
}
if (use_stout_flag == 1){
params_smear.rho = stout_rho;
params_smear.iterations = stout_no_iter;
if (stout_smear((su3_tuple*)(g_gauge_field[0]), ¶ms_smear, (su3_tuple*)(g_gauge_field[0])) != 0)
exit(1) ;
g_update_gauge_copy = 1;
g_update_gauge_energy = 1;
g_update_rectangle_energy = 1;
plaquette_energy = measure_gauge_action();
if (g_proc_id == 0) {
printf("# The plaquette value after stouting is %e\n", plaquette_energy / (6.*VOLUME*g_nproc));
fflush(stdout);
}
}
source_spinor_field(g_spinor_field[0], g_spinor_field[1], 0, 0);
convert_eo_to_lexic(g_spinor_field[DUM_DERI], g_spinor_field[0], g_spinor_field[1]);
D_psi(g_spinor_field[DUM_DERI+1], g_spinor_field[DUM_DERI]);
if(even_odd_flag) {
i = invert_eo(g_spinor_field[2], g_spinor_field[3], g_spinor_field[0], g_spinor_field[1],
solver_precision, max_solver_iterations, solver_flag, g_relative_precision_flag,
sub_evs_cg_flag, even_odd_flag);
convert_eo_to_lexic(g_spinor_field[DUM_DERI+1], g_spinor_field[2], g_spinor_field[3]);
}
for(i = 0; i < 3*LX/2+T/2; i++){
norm[i] = 0.;
}
for(x = 0; x < LX; x++){
if(x > LX/2) X = LX-x;
else X = x;
for(y = 0; y < LY; y++){
if(y > LY/2) Y = LY-y;
else Y = y;
for(z = 0; z < LZ; z++){
if(z > LZ/2) Z = LZ-z;
else Z = z;
for(t = 0; t < T; t++){
if(t > T/2) tt = T - t;
else tt = t;
sum = X + Y + Z + tt;
_spinor_norm_sq(nrm, g_spinor_field[DUM_DERI+1][ g_ipt[t][x][y][z] ]);
/* _spinor_norm_sq(nrm, qprop[0][0][1][ g_ipt[t][x][y][z] ]); */
printf("%e %e\n", g_spinor_field[DUM_DERI+1][ g_ipt[t][x][y][z] ].s0.c0.re, g_spinor_field[DUM_DERI+1][ g_ipt[t][x][y][z] ].s0.c0.im);
nrm = sqrt( nrm );
printf("%1.12e\n", nrm);
if(nrm > norm[sum]) norm[sum] = nrm;
}
}
}
}
for(i = 0; i < 3*L/2+T/2; i++){
printf("%d %1.12e\n", i, norm[i]);
}
printf("\n");
nstore+=Nsave;
}
#ifdef MPI
MPI_Finalize();
#endif
free_gauge_field();
free_geometry_indices();
free_spinor_field();
free_moment_field();
return(0);
#ifdef _KOJAK_INST
#pragma pomp inst end(main)
#endif
}
static void
load_Y_from_V(info_t *info, hisq_auxiliary_t *aux, int umethod){
int dir,i;
su3_matrix tmat;
su3_matrix *U_link = aux->U_link;
su3_matrix *V_link = aux->V_link;
su3_matrix *Y_unitlink = aux->Y_unitlink;
/* Temporary. FIX THIS! */
int r0[4] = {0, 0, 0, 0};
double dtime;
dtime=-dclock();
int nhits = 100; // this is just a guess
Real tol = 1.0e-5; // this is just a guess
info->final_flop = 0.;
switch(umethod){
case UNITARIZE_NONE:
//node0_printf("WARNING: UNITARIZE_NONE\n");
FORALLFIELDSITES_OMP(i,private(dir))for(dir=XUP;dir<=TUP;dir++)
Y_unitlink[4*i+dir] = V_link[4*i+dir];
END_LOOP_OMP;
break;
case UNITARIZE_APE:
node0_printf("UNITARIZE_APE: derivative is not ready for this method\n");
terminate(0);
FORALLFIELDSITES_OMP(i,private(dir,tmat))for(dir=XUP;dir<=TUP;dir++){
/* Use partially reunitarized link for guess */
tmat = V_link[4*i+dir];
reunit_su3(&tmat);
project_su3(&tmat, &(V_link[4*i+dir]), nhits, tol);
Y_unitlink[4*i+dir] = tmat;
}
END_LOOP_OMP;
break;
case UNITARIZE_ROOT:
//node0_printf("WARNING: UNITARIZE_ROOT is performed\n");
// REPHASING IS NOT NEEDED IN THIS ROUTINE BUT THIS
// HAS TO BE CHECKED FIRST FOR NONTRIVIAL V_link ARRAYS
/* rephase (out) V_link array */
rephase_field_offset( V_link, OFF, NULL , r0);
FORALLFIELDSITES_OMP(i,private(dir,tmat))for(dir=XUP;dir<=TUP;dir++){
/* unitarize - project on U(3) */
u3_unitarize( &( V_link[4*i+dir] ), &tmat );
Y_unitlink[4*i+dir] = tmat;
}
END_LOOP_OMP;
/* rephase (in) Y_unitlink array */
rephase_field_offset( Y_unitlink, ON, NULL , r0);
//printf("UNITARIZATION RESULT\n");
//dumpmat( &( V_link[TUP][3] ) );
//dumpmat( &( Y_unitlink[TUP][3] ) );
break;
case UNITARIZE_RATIONAL:
//node0_printf("WARNING: UNITARIZE_RATIONAL is performed\n");
// REPHASING IS NOT NEEDED IN THIS ROUTINE BUT THIS
// HAS TO BE CHECKED FIRST FOR NONTRIVIAL V_link ARRAYS
/* rephase (out) V_link array */
rephase_field_offset( V_link, OFF, NULL , r0);
FORALLFIELDSITES_OMP(i,private(dir,tmat))for(dir=XUP;dir<=TUP;dir++){
/* unitarize - project on U(3) */
u3_unitarize_rational( &( V_link[4*i+dir] ), &tmat );
Y_unitlink[4*i+dir] = tmat;
}
END_LOOP_OMP;
rephase_field_offset( V_link, ON, NULL , r0);
/* rephase (in) Y_unitlink array */
rephase_field_offset( Y_unitlink, ON, NULL , r0);
//printf("UNITARIZATION RESULT\n");
//dumpmat( &( V_link[TUP][3] ) );
//dumpmat( &( Y_unitlink[TUP][3] ) );
break;
case UNITARIZE_ANALYTIC:
//node0_printf("WARNING: UNITARIZE_ANALYTIC is performed\n");
// REPHASING IS NOT NEEDED IN THIS ROUTINE BUT THIS
// HAS TO BE CHECKED FIRST FOR NONTRIVIAL V_link ARRAYS
/* rephase (out) V_link array */
rephase_field_offset( V_link, OFF, NULL , r0);
FORALLFIELDSITES_OMP(i,private(dir,tmat))for(dir=XUP;dir<=TUP;dir++){
/* unitarize - project on U(3) */
#ifdef MILC_GLOBAL_DEBUG
#ifdef HISQ_REUNITARIZATION_DEBUG
u3_unitarize_analytic_index( &( V_link[4*i+dir] ), &tmat, i, dir );
#else /* HISQ_REUNITARIZATION_DEBUG */
u3_unitarize_analytic( info, &( V_link[4*i+dir] ), &tmat );
#endif /* HISQ_REUNITARIZATION_DEBUG */
#else /* MILC_GLOBAL_DEBUG */
u3_unitarize_analytic( info, &( V_link[4*i+dir] ), &tmat );
#endif /* MILC_GLOBAL_DEBUG */
Y_unitlink[4*i+dir] = tmat;
}
END_LOOP_OMP;
/* rephase (in) V_link array */
rephase_field_offset( V_link, ON, NULL , r0);
/* rephase (in) Y_unitlink array */
rephase_field_offset( Y_unitlink, ON, NULL , r0);
//printf("UNITARIZATION RESULT\n");
//dumpmat( &( V_link[TUP][3] ) );
//dumpmat( &( Y_unitlink[TUP][3] ) );
break;
case UNITARIZE_STOUT:
rephase_field_offset( U_link, OFF, NULL , r0);
rephase_field_offset( V_link, OFF, NULL , r0);
FORALLFIELDSITES_OMP(i,private(dir))for(dir=XUP;dir<=TUP;dir++){
/* unitarize - project on SU(3) with "stout" procedure */
stout_smear( &( Y_unitlink[4*i+dir] ),
&( V_link[4*i+dir] ),
&( U_link[4*i+dir] ) );
}
END_LOOP_OMP;
rephase_field_offset( U_link, ON, NULL , r0);
rephase_field_offset( V_link, ON, NULL , r0);
/* rephase (in) Y_unitlink array */
rephase_field_offset( Y_unitlink, ON, NULL , r0);
//printf("UNITARIZATION RESULT\n");
//dumpmat( &( V_link[TUP][3] ) );
//dumpmat( &( Y_unitlink[TUP][3] ) );
break;
case UNITARIZE_HISQ:
node0_printf("UNITARIZE_HISQ: not ready!\n");
terminate(0);
break;
default:
node0_printf("Unknown unitarization method\n"); terminate(0);
} /* umethod */
dtime += dclock();
info->final_sec = dtime;
}
int main(int argc, char *argv[])
{
FILE *parameterfile = NULL;
int c, j, i, ix = 0, isample = 0, op_id = 0;
char * filename = NULL;
char datafilename[50];
char parameterfilename[50];
char conf_filename[50];
char * input_filename = NULL;
double plaquette_energy;
struct stout_parameters params_smear;
spinor **s, *s_;
#ifdef _KOJAK_INST
#pragma pomp inst init
#pragma pomp inst begin(main)
#endif
#if (defined SSE || defined SSE2 || SSE3)
signal(SIGILL, &catch_ill_inst);
#endif
DUM_DERI = 8;
DUM_MATRIX = DUM_DERI + 5;
NO_OF_SPINORFIELDS = DUM_MATRIX + 2;
verbose = 0;
g_use_clover_flag = 0;
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vVf:o:")) != -1) {
switch (c) {
case 'f':
input_filename = calloc(200, sizeof(char));
strcpy(input_filename, optarg);
break;
case 'o':
filename = calloc(200, sizeof(char));
strcpy(filename, optarg);
break;
case 'v':
verbose = 1;
break;
case 'V':
fprintf(stdout,"%s %s\n",PACKAGE_STRING,git_hash);
exit(0);
break;
case 'h':
case '?':
default:
usage();
break;
}
}
if (input_filename == NULL) {
input_filename = "invert.input";
}
if (filename == NULL) {
filename = "output";
}
/* Read the input file */
if( (j = read_input(input_filename)) != 0) {
fprintf(stderr, "Could not find input file: %s\nAborting...\n", input_filename);
exit(-1);
}
/* this DBW2 stuff is not needed for the inversion ! */
if (g_dflgcr_flag == 1) {
even_odd_flag = 0;
}
g_rgi_C1 = 0;
if (Nsave == 0) {
Nsave = 1;
}
if (g_running_phmc) {
NO_OF_SPINORFIELDS = DUM_MATRIX + 8;
}
tmlqcd_mpi_init(argc, argv);
g_dbw2rand = 0;
/* starts the single and double precision random number */
/* generator */
start_ranlux(rlxd_level, random_seed);
/* we need to make sure that we don't have even_odd_flag = 1 */
/* if any of the operators doesn't use it */
/* in this way even/odd can still be used by other operators */
for(j = 0; j < no_operators; j++) if(!operator_list[j].even_odd_flag) even_odd_flag = 0;
#ifndef MPI
g_dbw2rand = 0;
#endif
#ifdef _GAUGE_COPY
j = init_gauge_field(VOLUMEPLUSRAND, 1);
#else
j = init_gauge_field(VOLUMEPLUSRAND, 0);
#endif
if (j != 0) {
fprintf(stderr, "Not enough memory for gauge_fields! Aborting...\n");
exit(-1);
}
j = init_geometry_indices(VOLUMEPLUSRAND);
if (j != 0) {
fprintf(stderr, "Not enough memory for geometry indices! Aborting...\n");
exit(-1);
}
if (no_monomials > 0) {
if (even_odd_flag) {
j = init_monomials(VOLUMEPLUSRAND / 2, even_odd_flag);
}
else {
j = init_monomials(VOLUMEPLUSRAND, even_odd_flag);
}
if (j != 0) {
fprintf(stderr, "Not enough memory for monomial pseudo fermion fields! Aborting...\n");
exit(-1);
}
}
if (even_odd_flag) {
j = init_spinor_field(VOLUMEPLUSRAND / 2, NO_OF_SPINORFIELDS);
}
else {
j = init_spinor_field(VOLUMEPLUSRAND, NO_OF_SPINORFIELDS);
}
if (j != 0) {
fprintf(stderr, "Not enough memory for spinor fields! Aborting...\n");
exit(-1);
}
if (g_running_phmc) {
j = init_chi_spinor_field(VOLUMEPLUSRAND / 2, 20);
if (j != 0) {
fprintf(stderr, "Not enough memory for PHMC Chi fields! Aborting...\n");
exit(-1);
}
}
g_mu = g_mu1;
if (g_cart_id == 0) {
/*construct the filenames for the observables and the parameters*/
strcpy(datafilename, filename);
strcat(datafilename, ".data");
strcpy(parameterfilename, filename);
strcat(parameterfilename, ".para");
parameterfile = fopen(parameterfilename, "w");
write_first_messages(parameterfile, 1);
fclose(parameterfile);
}
/* define the geometry */
geometry();
/* define the boundary conditions for the fermion fields */
boundary(g_kappa);
phmc_invmaxev = 1.;
init_operators();
/* this could be maybe moved to init_operators */
#ifdef _USE_HALFSPINOR
j = init_dirac_halfspinor();
if (j != 0) {
fprintf(stderr, "Not enough memory for halffield! Aborting...\n");
exit(-1);
}
if (g_sloppy_precision_flag == 1) {
j = init_dirac_halfspinor32();
if (j != 0)
{
fprintf(stderr, "Not enough memory for 32-bit halffield! Aborting...\n");
exit(-1);
}
}
# if (defined _PERSISTENT)
if (even_odd_flag)
init_xchange_halffield();
# endif
#endif
for (j = 0; j < Nmeas; j++) {
sprintf(conf_filename, "%s.%.4d", gauge_input_filename, nstore);
if (g_cart_id == 0) {
printf("#\n# Trying to read gauge field from file %s in %s precision.\n",
conf_filename, (gauge_precision_read_flag == 32 ? "single" : "double"));
fflush(stdout);
}
if( (i = read_gauge_field(conf_filename)) !=0) {
fprintf(stderr, "Error %d while reading gauge field from %s\n Aborting...\n", i, conf_filename);
exit(-2);
}
if (g_cart_id == 0) {
printf("# Finished reading gauge field.\n");
fflush(stdout);
}
#ifdef MPI
xchange_gauge();
#endif
/*compute the energy of the gauge field*/
plaquette_energy = measure_gauge_action();
if (g_cart_id == 0) {
printf("# The computed plaquette value is %e.\n", plaquette_energy / (6.*VOLUME*g_nproc));
fflush(stdout);
}
if (use_stout_flag == 1){
params_smear.rho = stout_rho;
params_smear.iterations = stout_no_iter;
if (stout_smear((su3_tuple*)(g_gauge_field[0]), ¶ms_smear, (su3_tuple*)(g_gauge_field[0])) != 0)
exit(1) ;
g_update_gauge_copy = 1;
g_update_gauge_energy = 1;
g_update_rectangle_energy = 1;
plaquette_energy = measure_gauge_action();
if (g_cart_id == 0) {
printf("# The plaquette value after stouting is %e\n", plaquette_energy / (6.*VOLUME*g_nproc));
fflush(stdout);
}
}
if (reweighting_flag == 1) {
reweighting_factor(reweighting_samples, nstore);
}
/* Compute minimal eigenvalues, if wanted */
if (compute_evs != 0) {
eigenvalues(&no_eigenvalues, 5000, eigenvalue_precision,
0, compute_evs, nstore, even_odd_flag);
}
if (phmc_compute_evs != 0) {
#ifdef MPI
MPI_Finalize();
#endif
return(0);
}
/* Compute the mode number or topological susceptibility using spectral projectors, if wanted*/
if(compute_modenumber != 0 || compute_topsus !=0){
s_ = calloc(no_sources_z2*VOLUMEPLUSRAND+1, sizeof(spinor));
s = calloc(no_sources_z2, sizeof(spinor*));
if(s_ == NULL) {
printf("Not enough memory in %s: %d",__FILE__,__LINE__); exit(42);
}
if(s == NULL) {
printf("Not enough memory in %s: %d",__FILE__,__LINE__); exit(42);
}
for(i = 0; i < no_sources_z2; i++) {
#if (defined SSE3 || defined SSE2 || defined SSE)
s[i] = (spinor*)(((unsigned long int)(s_)+ALIGN_BASE)&~ALIGN_BASE)+i*VOLUMEPLUSRAND;
#else
s[i] = s_+i*VOLUMEPLUSRAND;
#endif
z2_random_spinor_field(s[i], VOLUME);
/* what is this here needed for?? */
/* spinor *aux_,*aux; */
/* #if ( defined SSE || defined SSE2 || defined SSE3 ) */
/* aux_=calloc(VOLUMEPLUSRAND+1, sizeof(spinor)); */
/* aux = (spinor *)(((unsigned long int)(aux_)+ALIGN_BASE)&~ALIGN_BASE); */
/* #else */
/* aux_=calloc(VOLUMEPLUSRAND, sizeof(spinor)); */
/* aux = aux_; */
/* #endif */
if(g_proc_id == 0) {
printf("source %d \n", i);
}
if(compute_modenumber != 0){
mode_number(s[i], mstarsq);
}
if(compute_topsus !=0) {
top_sus(s[i], mstarsq);
}
}
free(s);
free(s_);
}
/* move to operators as well */
if (g_dflgcr_flag == 1) {
/* set up deflation blocks */
init_blocks(nblocks_t, nblocks_x, nblocks_y, nblocks_z);
/* the can stay here for now, but later we probably need */
/* something like init_dfl_solver called somewhere else */
/* create set of approximate lowest eigenvectors ("global deflation subspace") */
/* g_mu = 0.; */
/* boundary(0.125); */
generate_dfl_subspace(g_N_s, VOLUME);
/* boundary(g_kappa); */
/* g_mu = g_mu1; */
/* Compute little Dirac operators */
/* alt_block_compute_little_D(); */
if (g_debug_level > 0) {
check_projectors();
check_local_D();
}
if (g_debug_level > 1) {
check_little_D_inversion();
}
}
if(SourceInfo.type == 1) {
index_start = 0;
index_end = 1;
}
g_precWS=NULL;
if(use_preconditioning == 1){
/* todo load fftw wisdom */
#if (defined HAVE_FFTW ) && !( defined MPI)
loadFFTWWisdom(g_spinor_field[0],g_spinor_field[1],T,LX);
#else
use_preconditioning=0;
#endif
}
if (g_cart_id == 0) {
fprintf(stdout, "#\n"); /*Indicate starting of the operator part*/
}
for(op_id = 0; op_id < no_operators; op_id++) {
boundary(operator_list[op_id].kappa);
g_kappa = operator_list[op_id].kappa;
g_mu = 0.;
if(use_preconditioning==1 && PRECWSOPERATORSELECT[operator_list[op_id].solver]!=PRECWS_NO ){
printf("# Using preconditioning with treelevel preconditioning operator: %s \n",
precWSOpToString(PRECWSOPERATORSELECT[operator_list[op_id].solver]));
/* initial preconditioning workspace */
operator_list[op_id].precWS=(spinorPrecWS*)malloc(sizeof(spinorPrecWS));
spinorPrecWS_Init(operator_list[op_id].precWS,
operator_list[op_id].kappa,
operator_list[op_id].mu/2./operator_list[op_id].kappa,
-(0.5/operator_list[op_id].kappa-4.),
PRECWSOPERATORSELECT[operator_list[op_id].solver]);
g_precWS = operator_list[op_id].precWS;
if(PRECWSOPERATORSELECT[operator_list[op_id].solver] == PRECWS_D_DAGGER_D) {
fitPrecParams(op_id);
}
}
for(isample = 0; isample < no_samples; isample++) {
for (ix = index_start; ix < index_end; ix++) {
if (g_cart_id == 0) {
fprintf(stdout, "#\n"); /*Indicate starting of new index*/
}
/* we use g_spinor_field[0-7] for sources and props for the moment */
/* 0-3 in case of 1 flavour */
/* 0-7 in case of 2 flavours */
prepare_source(nstore, isample, ix, op_id, read_source_flag, source_location);
operator_list[op_id].inverter(op_id, index_start);
}
}
if(use_preconditioning==1 && operator_list[op_id].precWS!=NULL ){
/* free preconditioning workspace */
spinorPrecWS_Free(operator_list[op_id].precWS);
free(operator_list[op_id].precWS);
}
if(operator_list[op_id].type == OVERLAP){
free_Dov_WS();
}
}
nstore += Nsave;
}
#ifdef MPI
MPI_Finalize();
#endif
free_blocks();
free_dfl_subspace();
free_gauge_field();
free_geometry_indices();
free_spinor_field();
free_moment_field();
free_chi_spinor_field();
return(0);
#ifdef _KOJAK_INST
#pragma pomp inst end(main)
#endif
}
// Evolve momenta according to the rooted staggered force
THREADABLE_FUNCTION_7ARG(evolve_momenta_with_quark_force, quad_su3**,H, quad_su3**,conf, std::vector<std::vector<pseudofermion_t> >*,pf, theory_pars_t*,theory_pars, hmc_evol_pars_t*,simul_pars, std::vector<rat_approx_t>*,rat_appr, double,dt)
{
GET_THREAD_ID();
verbosity_lv2_master_printf("Evolving momenta with quark force, dt=%lg\n",dt);
//allocate forces
quad_su3 *F[2]={nissa_malloc("F0",loc_volh,quad_su3),nissa_malloc("F1",loc_volh,quad_su3)};
//compute the force
compute_quark_force(F,conf,pf,theory_pars,rat_appr,simul_pars->md_residue);
//#define DEBUG
#ifdef DEBUG
int par=1,ieo=1,mu=1;
double eps=1e-5;
//store initial link
su3 sto;
su3_copy(sto,conf[par][ieo][mu]);
//allocate smeared conf
quad_su3 *sme_conf[2];
for(int eo=0;eo<2;eo++) sme_conf[eo]=nissa_malloc("sme_conf",loc_volh+bord_volh+edge_volh,quad_su3);
//compute action before
double act_ori;
stout_smear(sme_conf,conf,&(theory_pars->stout_pars));
rootst_eoimpr_quark_action(&act_ori,sme_conf,theory_pars->nflavs,theory_pars->backfield,pf,simul_pars);
//store derivative
su3 nu_plus,nu_minus;
su3_put_to_zero(nu_plus);
su3_put_to_zero(nu_minus);
for(int igen=0;igen<NCOL*NCOL-1;igen++)
{
//prepare increment and change
su3 ba;
su3_prod_double(ba,gell_mann_matr[igen],eps/2);
su3 exp_mod;
safe_hermitian_exact_i_exponentiate(exp_mod,ba);
//change -, compute action
unsafe_su3_dag_prod_su3(conf[par][ieo][mu],exp_mod,sto);
double act_minus;
stout_smear(sme_conf,conf,&(theory_pars->stout_pars));
rootst_eoimpr_quark_action(&act_minus,sme_conf,theory_pars->nflavs,theory_pars->backfield,pf,simul_pars);
//change +, compute action
unsafe_su3_prod_su3(conf[par][ieo][mu],exp_mod,sto);
double act_plus;
stout_smear(sme_conf,conf,&(theory_pars->stout_pars));
rootst_eoimpr_quark_action(&act_plus,sme_conf,theory_pars->nflavs,theory_pars->backfield,pf,simul_pars);
//set back everything
su3_copy(conf[par][ieo][mu],sto);
//printf("plus: %+016.016le, ori: %+016.016le, minus: %+016.016le, eps: %lg\n",act_plus,act_ori,act_minus,eps);
double gr_plus=-(act_plus-act_ori)/eps;
double gr_minus=-(act_ori-act_minus)/eps;
su3_summ_the_prod_idouble(nu_plus,gell_mann_matr[igen],gr_plus);
su3_summ_the_prod_idouble(nu_minus,gell_mann_matr[igen],gr_minus);
}
//take the average
su3 nu;
su3_summ(nu,nu_plus,nu_minus);
su3_prodassign_double(nu,0.5);
master_printf("checking pure gauge force\n");
master_printf("an\n");
su3_print(F[par][ieo][mu]);
master_printf("nu\n");
su3_print(nu);
master_printf("nu_plus\n");
su3_print(nu_plus);
master_printf("nu_minus\n");
su3_print(nu_minus);
//crash("anna");
#endif
//evolve
for(int par=0;par<2;par++)
{
NISSA_PARALLEL_LOOP(ivol,0,loc_volh)
for(int mu=0;mu<NDIM;mu++)
for(int ic1=0;ic1<NCOL;ic1++)
for(int ic2=0;ic2<NCOL;ic2++)
complex_subt_the_prod_idouble(H[par][ivol][mu][ic1][ic2],F[par][ivol][mu][ic1][ic2],dt);
nissa_free(F[par]);
}
}
