void tm_times_Hopping_Matrix(const int ieo, spinor * const l, spinor * const k, double complex const cfactor) {
# ifdef XLC
# pragma disjoint(*l, *k)
# endif
# ifdef _GAUGE_COPY
if(g_update_gauge_copy) {
update_backward_gauge(g_gauge_field);
}
# endif
# if (defined MPI)
xchange_field(k, ieo);
# endif
# ifdef OMP
# pragma omp parallel
{
# endif
# define _MUL_G5_CMPLX
# if (defined BGQ && defined XLC)
complex double ALIGN bla = cfactor;
vector4double ALIGN cf = vec_ld2(0, (double*) &bla);
# elif (defined SSE2 || defined SSE3)
_Complex double ALIGN cf = cfactor;
# endif
# include "operator/hopping_body_dbl.c"
# undef _MUL_G5_CMPLX
# ifdef OMP
} /* OpenMP closing brace */
# endif
return;
}
void Hopping_Matrix(const int ieo, spinor * const l, spinor * const k) {
# ifdef XLC
# pragma disjoint(*l, *k)
# endif
# ifdef _GAUGE_COPY
if(g_update_gauge_copy) {
update_backward_gauge(g_gauge_field);
}
# endif
# if (defined TM_USE_MPI && !(defined _NO_COMM))
xchange_field(k, ieo);
# endif
# ifdef TM_USE_OMP
# pragma omp parallel
{
# endif
# include "operator/hopping_body_dbl.c"
# ifdef TM_USE_OMP
} /* OpenMP closing brace */
# endif
return;
}
void tm_sub_Hopping_Matrix(const int ieo, spinor * const l, spinor * p, spinor * const k,
complex double const cfactor) {
# ifdef XLC
# pragma disjoint(*l, *k)
# endif
# ifdef _GAUGE_COPY
if(g_update_gauge_copy) {
update_backward_gauge(g_gauge_field);
}
# endif
# if (defined TM_USE_MPI)
xchange_field(k, ieo);
# endif
# ifdef TM_USE_OMP
# pragma omp parallel
{
# endif
# define _TM_SUB_HOP
spinor * pn;
# if (defined BGQ && defined XLC)
complex double ALIGN bla = cfactor;
vector4double ALIGN cf = vec_ld2(0, (double*) &bla);
# elif (defined SSE2 || defined SSE3)
_Complex double ALIGN cf = cfactor;
su3_vector ALIGN psi, psi2;
# endif
# include "operator/hopping_body_dbl.c"
# undef _TM_SUB_HOP
# ifdef TM_USE_OMP
} /* OpenMP closing brace */
# endif
return;
}
void deriv_Sb_D_psi(spinor * const l, spinor * const k,
hamiltonian_field_t * const hf, const double factor) {
int ix,iy, iz;
int ioff,ioff2,icx,icy, icz;
su3 * restrict up ALIGN;
su3 * restrict um ALIGN;
su3adj * restrict ddd;
static su3adj der;
static su3 v1,v2;
static su3_vector psia,psib,phia,phib;
static spinor rr;
spinor * restrict r ALIGN;
spinor * restrict sp ALIGN;
spinor * restrict sm ALIGN;
/* We have 32 registers available */
double _Complex reg00, reg01, reg02, reg03, reg04, reg05;
double _Complex reg10, reg11, reg12, reg13, reg14, reg15;
/* For su3 matrix, use reg00 for missing register */
double _Complex v00, v01, v02, v10, v11, v12, v20, v21;
/* The following contains the left spinor (12 regs) and the final */
/* su3 matrix to trace over */
double _Complex r00, r01, r02, r10, r11, r12, r20, r21, r22,
r30, r31, r32;
#ifdef _KOJAK_INST
# pragma pomp inst begin(derivSb)
#endif
#pragma disjoint(*r, *sp, *sm, *up, *um, *ddd)
__alignx(16, l);
__alignx(16, k);
if(ieo==0) {
ioff=0;
}
else {
ioff=(VOLUME+RAND)/2;
}
ioff2=(VOLUME+RAND)/2-ioff;
/* for parallelization */
#ifdef MPI
xchange_field(k, ieo);
xchange_field(l, (ieo+1)%2);
#endif
/************** loop over all lattice sites ****************/
ix=ioff;
iy=g_iup[ix][0]; icy=iy;
sp = k + icy;
_prefetch_spinor(sp);
up=&hf->gaugefield[ix][0];
_prefetch_su3(up);
for(icx = ioff; icx < (VOLUME+ioff); icx++){
/* load left vector r and */
/* multiply with gamma5 */
r = l + (icx-ioff);
ix=icx;
/*********************** direction +0 ********************/
ddd = &hf->derivative[ix][0];
_bgl_load_r0((*r).s0);
_bgl_load_r1((*r).s1);
_bgl_load_minus_r2((*r).s2);
_bgl_load_minus_r3((*r).s3);
_bgl_load_reg0((*sp).s0);
_bgl_load_reg0_up((*sp).s1);
_bgl_load_reg1((*sp).s2);
_bgl_load_reg1_up((*sp).s3);
_bgl_add_to_reg0_reg1();
_bgl_add_to_reg0_up_reg1_up();
_bgl_add_r0_to_r2_reg1();
_bgl_add_r1_to_r3_reg1_up();
iy=g_idn[ix][0]; icy=iy;
sm = k + icy;
_prefetch_spinor(sm);
um=&hf->gaugefield[iy][0];
_prefetch_su3(um);
_bgl_tensor_product_and_add();
/* result in v now */
_bgl_su3_times_v_dagger(*up);
/* result in r now */
_bgl_complex_times_r(ka0);
_bgl_trace_lambda_mul_add_assign((*ddd), 2.*factor);
/************** direction -0 ****************************/
ddd = &hf->derivative[iy][0];
_bgl_load_r0((*r).s0);
_bgl_load_r1((*r).s1);
_bgl_load_minus_r2((*r).s2);
_bgl_load_minus_r3((*r).s3);
_bgl_load_reg0((*sm).s0);
_bgl_load_reg0_up((*sm).s1);
_bgl_load_reg1((*sm).s2);
_bgl_load_reg1_up((*sm).s3);
_bgl_sub_from_reg0_reg1();
_bgl_sub_from_reg0_up_reg1_up();
_bgl_sub_from_r0_r2_reg1();
_bgl_sub_from_r1_r3_reg1_up();
iy=g_iup[ix][1]; icy=[iy];
sp = k + icy;
_prefetch_spinor(sp);
up=&hf->gaugefield[ix][1];
_prefetch_su3(up);
_bgl_tensor_product_and_add_d();
/* result in v now */
_bgl_su3_times_v_dagger(*um);
/* result in r now */
_bgl_complex_times_r(ka0);
_bgl_trace_lambda_mul_add_assign((*ddd), 2.*factor);
/*************** direction +1 **************************/
ddd = &hf->derivative[ix][1];
_bgl_load_r0((*r).s0);
_bgl_load_r1((*r).s1);
_bgl_load_minus_r2((*r).s2);
_bgl_load_minus_r3((*r).s3);
_bgl_load_reg0((*sp).s0);
_bgl_load_reg0_up((*sp).s1);
_bgl_load_reg1((*sp).s2);
_bgl_load_reg1_up((*sp).s3);
_bgl_i_mul_add_to_reg0_reg1_up();
_bgl_i_mul_add_to_reg0_up_reg1();
_bgl_i_mul_add_r0_to_r3_reg1();
_bgl_i_mul_add_r1_to_r2_reg1_up();
iy=g_idn[ix][1]; icy=iy;
sm = k + icy;
_prefetch_spinor(sm);
um=&hf->gaugefield[iy][1];
_prefetch_su3(um);
_bgl_tensor_product_and_add();
/* result in v now */
_bgl_su3_times_v_dagger(*up);
/* result in r now */
_bgl_complex_times_r(ka1);
_bgl_trace_lambda_mul_add_assign((*ddd), 2.*factor);
/**************** direction -1 *************************/
ddd = &hf->derivative[iy][1];
_bgl_load_r0((*r).s0);
_bgl_load_r1((*r).s1);
_bgl_load_minus_r2((*r).s2);
_bgl_load_minus_r3((*r).s3);
_bgl_load_reg0((*sp).s0);
_bgl_load_reg0_up((*sp).s1);
_bgl_load_reg1((*sp).s2);
_bgl_load_reg1_up((*sp).s3);
_bgl_i_mul_sub_from_reg0_reg1_up();
_bgl_i_mul_sub_from_reg0_up_reg1();
_bgl_i_mul_sub_from_r0_r3_reg1();
_bgl_i_mul_sub_from_r1_r2_reg1_up();
iy=g_iup[ix][2]; icy=iy;
sp = k + icy;
_prefetch_spinor(sp);
up=&hf->gaugefield[ix][2];
_prefetch_su3(up);
_bgl_tensor_product_and_add_d();
/* result in v now */
_bgl_su3_times_v_dagger(*um);
/* result in r now */
_bgl_complex_times_r(ka1);
_bgl_trace_lambda_mul_add_assign((*ddd), 2.*factor);
/*************** direction +2 **************************/
ddd = &hf->derivative[ix][2];
_bgl_load_r0((*r).s0);
_bgl_load_r1((*r).s1);
_bgl_load_minus_r2((*r).s2);
_bgl_load_minus_r3((*r).s3);
_bgl_load_reg0((*sp).s0);
_bgl_load_reg0_up((*sp).s1);
_bgl_load_reg1((*sp).s2);
_bgl_load_reg1_up((*sp).s3);
_bgl_add_to_reg0_reg1_up();
_bgl_sub_from_reg0_up_reg1();
_bgl_add_r0_to_r3_reg1();
_bgl_sub_from_r1_r2_reg1_up();
iy=g_idn[ix][2]; icy=iy;
sm = k + icy;
_prefetch_spinor(sm);
um=&hf->gaugefield[iy][2];
_prefetch_su3(um);
_bgl_tensor_product_and_add();
/* result in v now */
_bgl_su3_times_v_dagger(*up);
/* result in r now */
_bgl_complex_times_r(ka2);
_bgl_trace_lambda_mul_add_assign((*ddd), 2.*factor);
/***************** direction -2 ************************/
ddd = &hf->derivative[iy][2];
_bgl_load_r0((*r).s0);
_bgl_load_r1((*r).s1);
_bgl_load_minus_r2((*r).s2);
_bgl_load_minus_r3((*r).s3);
_bgl_load_reg0((*sp).s0);
_bgl_load_reg0_up((*sp).s1);
_bgl_load_reg1((*sp).s2);
_bgl_load_reg1_up((*sp).s3);
_bgl_sub_from_reg0_reg1_up();
_bgl_add_to_reg0_up_reg1();
_bgl_sub_from_r0_r3_reg1();
_bgl_add_r1_to_r2_reg1_up();
iy=g_iup[ix][3]; icy=iy;
sp = k + icy;
_prefetch_spinor(sp);
up=&hf->gaugefield[ix][3];
_prefetch_su3(up);
_bgl_tensor_product_and_add_d();
/* result in v now */
_bgl_su3_times_v_dagger(*um);
/* result in r now */
_bgl_complex_times_r(ka1);
_bgl_trace_lambda_mul_add_assign(*ddd, 2.*factor);
/****************** direction +3 ***********************/
ddd = &hf->derivative[ix][3];
_bgl_load_r0((*r).s0);
_bgl_load_r1((*r).s1);
_bgl_load_minus_r2((*r).s2);
_bgl_load_minus_r3((*r).s3);
_bgl_load_reg0((*sp).s0);
_bgl_load_reg0_up((*sp).s1);
_bgl_load_reg1((*sp).s2);
_bgl_load_reg1_up((*sp).s3);
_bgl_i_mul_add_to_reg0_reg1();
_bgl_i_mul_sub_from_reg0_up_reg1_up();
_bgl_i_mul_add_r0_to_r2_reg1();
_bgl_i_mul_sub_from_r1_r3_reg1_up();
iy=g_idn[ix][3]; icy=iy;
sm = k + icy;
_prefetch_spinor(sm);
um=&hf->gaugefield[iy][3];
_prefetch_su3(um);
_bgl_tensor_product_and_add();
/* result in v now */
_bgl_su3_times_v_dagger(*up);
/* result in r now */
_bgl_complex_times_r(ka3);
_bgl_trace_lambda_mul_add_assign((*ddd), 2.*factor);
/***************** direction -3 ************************/
ddd = &hf->derivative[iy][3];
_bgl_load_r0((*r).s0);
_bgl_load_r1((*r).s1);
_bgl_load_minus_r2((*r).s2);
_bgl_load_minus_r3((*r).s3);
_bgl_load_reg0((*sp).s0);
_bgl_load_reg0_up((*sp).s1);
_bgl_load_reg1((*sp).s2);
_bgl_load_reg1_up((*sp).s3);
_bgl_i_mul_sub_from_reg0_reg1();
_bgl_i_mul_add_to_reg0_up_reg1_up();
_bgl_i_mul_sub_from_r0_r2_reg1();
_bgl_i_mul_add_r1_to_r3_reg1_up();
/* something wrong here...*/
icz=icx+1;
if(icz==((VOLUME+RAND)/2+ioff)) icz=ioff;
iz=icz;
iy=g_iup[iz][0]; icy=iy;
sp = k + icy;
_prefetch_spinor(sp);
up=&hf->gaugefield[iz][0];
_prefetch_su3(up);
_bgl_tensor_product_and_add_d();
/* result in v now */
_bgl_su3_times_v_dagger(*um);
/* result in r now */
_bgl_complex_times_r(ka3);
_bgl_trace_lambda_mul_add_assign((*ddd), 2.*factor);
/****************** end of loop ************************/
}
#ifdef _KOJAK_INST
#pragma pomp inst end(derivSb)
#endif
}
int main(int argc, char **argv) {
int c, i, mu, nu;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix, iix;
int dxm[4], dxn[4], ixpm, ixpn;
int sid;
double *disc = (double*)NULL;
double *disc2 = (double*)NULL;
double *work = (double*)NULL;
double q[4], fnorm;
int verbose = 0;
int do_gt = 0;
char filename[100], contype[200];
double ratime, retime;
double plaq, _2kappamu, hpe3_coeff, onepmutilde2, mutilde2;
double spinor1[24], spinor2[24], U_[18], U1_[18], U2_[18];
double *gauge_trafo=(double*)NULL;
complex w, w1, w2, *cp1, *cp2, *cp3;
FILE *ofs;
fftw_complex *in=(fftw_complex*)NULL;
#ifdef MPI
fftwnd_mpi_plan plan_p, plan_m;
#else
fftwnd_plan plan_p, plan_m;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vgf:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'g':
do_gt = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
if(filename_set==0) strcpy(filename, "cvc.input");
fprintf(stdout, "# Reading input from file %s\n", filename);
read_input_parser(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.n");
usage();
}
/* initialize MPI parameters */
mpi_init(argc, argv);
/* initialize fftw */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
#ifdef MPI
plan_p = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_BACKWARD, FFTW_MEASURE);
plan_m = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_FORWARD, FFTW_MEASURE);
fftwnd_mpi_local_sizes(plan_p, &T, &Tstart, &l_LX_at, &l_LXstart_at, &FFTW_LOC_VOLUME);
#else
plan_p = fftwnd_create_plan(4, dims, FFTW_BACKWARD, FFTW_MEASURE | FFTW_IN_PLACE);
plan_m = fftwnd_create_plan(4, dims, FFTW_FORWARD, FFTW_MEASURE | FFTW_IN_PLACE);
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
#endif
fprintf(stdout, "# [%2d] fftw parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(2);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(1);
}
geometry();
/* read the gauge field */
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
#ifdef MPI
xchange_gauge();
#endif
/* measure the plaquette */
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value: %25.16e\n", plaq);
if(do_gt==1) {
/***********************************
* initialize gauge transformation
***********************************/
init_gauge_trafo(&gauge_trafo, 1.);
apply_gt_gauge(gauge_trafo);
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value after gauge trafo: %25.16e\n", plaq);
}
/****************************************
* allocate memory for the spinor fields
****************************************/
no_fields = 3;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/****************************************
* allocate memory for the contractions
****************************************/
disc = (double*)calloc( 8*VOLUME, sizeof(double));
if( disc == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
disc2 = (double*)calloc( 8*VOLUME, sizeof(double));
if( disc2 == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc2\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
for(ix=0; ix<8*VOLUME; ix++) disc2[ix] = 0.;
work = (double*)calloc(48*VOLUME, sizeof(double));
if( work == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for work\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
/****************************************
* prepare Fourier transformation arrays
****************************************/
in = (fftw_complex*)malloc(FFTW_LOC_VOLUME*sizeof(fftw_complex));
if(in==(fftw_complex*)NULL) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(4);
}
/************************************************
* HPE: calculate coeff. of 3rd order term
************************************************/
_2kappamu = 2. * g_kappa * g_mu;
onepmutilde2 = 1. + _2kappamu * _2kappamu;
mutilde2 = _2kappamu * _2kappamu;
hpe3_coeff = 16. * g_kappa*g_kappa*g_kappa*g_kappa * (1. + 6. * mutilde2 + mutilde2*mutilde2) / onepmutilde2 / onepmutilde2 / onepmutilde2 / onepmutilde2;
/*
hpe3_coeff = 8. * g_kappa*g_kappa*g_kappa * \
(1. + 6.*_2kappamu*_2kappamu + _2kappamu*_2kappamu*_2kappamu*_2kappamu) / (1. + _2kappamu*_2kappamu) / (1. + _2kappamu*_2kappamu) / (1. + _2kappamu*_2kappamu) / (1. + _2kappamu*_2kappamu);
*/
fprintf(stdout, "hpe3_coeff = %25.16e\n", hpe3_coeff);
/************************************************
* HPE: calculate the plaquette terms
************************************************/
for(ix=0; ix<VOLUME; ix++) {
for(mu=0; mu<4; mu++) {
for(i=1; i<4; i++) {
nu = (mu+i)%4;
_cm_eq_cm_ti_cm(U1_, g_gauge_field+_GGI(ix,mu), g_gauge_field+_GGI(g_iup[ix][mu],nu) );
_cm_eq_cm_ti_cm(U2_, g_gauge_field+_GGI(ix,nu), g_gauge_field+_GGI(g_iup[ix][nu],mu) );
_cm_eq_cm_ti_cm_dag(U_, U1_, U2_);
_co_eq_tr_cm(&w1, U_);
iix = g_idn[ix][nu];
_cm_eq_cm_ti_cm(U1_, g_gauge_field+_GGI(iix,mu), g_gauge_field+_GGI(g_iup[iix][mu],nu) );
_cm_eq_cm_ti_cm(U2_, g_gauge_field+_GGI(iix,nu), g_gauge_field+_GGI(g_iup[iix][nu],mu) );
_cm_eq_cm_ti_cm_dag(U_, U1_, U2_);
_co_eq_tr_cm(&w2, U_);
disc2[_GWI(mu,ix,VOLUME)+1] += hpe3_coeff * (w1.im - w2.im);
/*
_cm_eq_cm_ti_cm(U1_, g_gauge_field+_GGI(g_idn[ix][nu],nu), g_gauge_field+_GGI(ix,mu) );
_cm_eq_cm_ti_cm(U2_, g_gauge_field+_GGI(g_idn[ix][nu],mu), g_gauge_field+_GGI(g_iup[g_idn[ix][nu]][mu], nu) );
_cm_eq_cm_ti_cm_dag(U_, U1_, U2_);
_co_eq_tr_cm(&w2, U_);
disc2[_GWI(mu,ix,VOLUME)+1] += hpe3_coeff * (w1.im + w2.im);
*/
/* fprintf(stdout, "mu=%1d, ix=%5d, nu=%1d, w1=%25.16e +i %25.16e; w2=%25.16e +i %25.16e\n",
mu, ix, nu, w1.re, w1.im, w2.re, w2.im); */
} /* of nu */
/****************************************
* - in case lattice size equals 4
* calculate additional loop term
* - _NOTE_ the possible minus sign from
* the fermionic boundary conditions
****************************************/
if(dims[mu]==4) {
wilson_loop(&w, ix, mu, dims[mu]);
fnorm = -64. * g_kappa*g_kappa*g_kappa*g_kappa / onepmutilde2 / onepmutilde2 / onepmutilde2 / onepmutilde2;
disc2[_GWI(mu,ix,VOLUME)+1] += fnorm * w.im;
/* fprintf(stdout, "loop contribution: ix=%5d, mu=%2d, fnorm=%25.16e, w=%25.16e\n", ix, mu, fnorm, w.im); */
}
/*
fprintf(stdout, "-------------------------------------------\n");
fprintf(stdout, "disc2[ix=%d,mu=%d] = %25.16e +i %25.16e\n", ix, mu, disc2[_GWI(mu,ix,VOLUME)], disc2[_GWI(mu,ix,VOLUME)+1]);
fprintf(stdout, "-------------------------------------------\n");
*/
}
}
/*
sprintf(filename, "avc_disc_hpe5_3rd.%.4d", Nconf);
ofs = fopen(filename, "w");
for(ix=0; ix<VOLUME; ix++) {
for(mu=0; mu<4; mu++) {
fprintf(ofs, "%6d%3d%25.16e\t%25.16e\n", ix, mu, disc[_GWI(mu,ix,VOLUME)], disc[_GWI(mu,ix,VOLUME)+1]);
}
}
fclose(ofs);
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
*/
/*
for(x0=0; x0<T; x0++) {
for(x1=0; x1<LX; x1++) {
for(x2=0; x2<LY; x2++) {
for(x3=0; x3<LZ; x3++) {
ix = g_ipt[x0][x1][x2][x3];
for(mu=0; mu<4; mu++) {
dxm[0]=0; dxm[1]=0; dxm[2]=0; dxm[3]=0; dxm[mu]=1;
for(i=1; i<4; i++) {
nu = (mu+i)%4;
dxn[0]=0; dxn[1]=0; dxn[2]=0; dxn[3]=0; dxn[nu]=1;
ixpm = g_ipt[(x0+dxm[0]+T)%T][(x1+dxm[1]+LX)%LX][(x2+dxm[2]+LY)%LY][(x3+dxm[3]+LZ)%LZ];
ixpn = g_ipt[(x0+dxn[0]+T)%T][(x1+dxn[1]+LX)%LX][(x2+dxn[2]+LY)%LY][(x3+dxn[3]+LZ)%LZ];
_cm_eq_cm_ti_cm(U1_, g_gauge_field + 72*ix+18*mu, g_gauge_field + 72*ixpm+18*nu );
_cm_eq_cm_ti_cm(U2_, g_gauge_field + 72*ix+18*nu, g_gauge_field + 72*ixpn+18*mu );
_cm_eq_cm_ti_cm_dag(U_, U1_, U2_);
_co_eq_tr_cm(&w1, U_);
ixpm = g_ipt[(x0+dxm[0]-dxn[0]+T)%T][(x1+dxm[1]-dxn[1]+LX)%LX][(x2+dxm[2]-dxn[2]+LY)%LY][(x3+dxm[3]-dxn[3]+LZ)%LZ];
ixpn = g_ipt[(x0-dxn[0]+T)%T][(x1-dxn[1]+LX)%LX][(x2-dxn[2]+LY)%LY][(x3-dxn[3]+LZ)%LZ];
_cm_eq_cm_ti_cm(U1_, g_gauge_field + 72*ixpn+18*nu, g_gauge_field + 72*ix+18*mu);
_cm_eq_cm_ti_cm(U2_, g_gauge_field + 72*ixpn+18*mu, g_gauge_field + 72*ixpm+18*nu);
_cm_eq_cm_ti_cm_dag(U_, U1_, U2_);
_co_eq_tr_cm(&w2, U_);
disc2[_GWI(mu,ix,VOLUME)+1] += hpe3_coeff * (w1.im + w2.im);
fprintf(stdout, "mu=%1d, ix=%5d, nu=%1d, w1=%25.16e; w2=%25.16e\n", mu, ix, nu, w1.im, w2.im);
}
fprintf(stdout, "-------------------------------------------\n");
fprintf(stdout, "disc2[ix=%d,mu=%d] = %25.16e +i %25.16e\n", ix, mu, disc2[_GWI(mu,ix,VOLUME)], disc2[_GWI(mu,ix,VOLUME)+1]);
fprintf(stdout, "-------------------------------------------\n");
}
}
}
}
}
*/
/***********************************************
* start loop on source id.s
***********************************************/
for(sid=g_sourceid; sid<=g_sourceid2; sid+=g_sourceid_step) {
/* read the new propagator */
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, Nconf, sid);
if(read_lime_spinor(g_spinor_field[2], filename, 0) != 0) break;
}
else if(format==1) {
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, Nconf, sid);
if(read_cmi(g_spinor_field[2], filename) != 0) break;
}
xchange_field(g_spinor_field[2]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "time to read prop.: %e seconds\n", retime-ratime);
if(do_gt==1) {
/******************************************
* gauge transform the propagators for sid
******************************************/
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_cm_ti_fv(spinor1, gauge_trafo+18*ix, g_spinor_field[2]+_GSI(ix));
_fv_eq_fv(g_spinor_field[2]+_GSI(ix), spinor1);
}
xchange_field(g_spinor_field[2]);
}
count++;
/************************************************
* calculate the source: apply Q_phi_tbc
************************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
Q_phi_tbc(g_spinor_field[0], g_spinor_field[2]);
xchange_field(g_spinor_field[0]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "time to calculate source: %e seconds\n", retime-ratime);
/************************************************
* HPE: apply BH5
************************************************/
BH5(g_spinor_field[1], g_spinor_field[2]);
/* add new contractions to (existing) disc */
# ifdef MPI
ratime = MPI_Wtime();
# else
ratime = (double)clock() / CLOCKS_PER_SEC;
# endif
for(mu=0; mu<4; mu++) { /* loop on Lorentz index of the current */
iix = _GWI(mu,0,VOLUME);
for(ix=0; ix<VOLUME; ix++) { /* loop on lattice sites */
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix, mu)], &co_phase_up[mu]);
/* first contribution */
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(ix)], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
/* second contribution */
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(g_iup[ix][mu])], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
iix += 2;
} /* of ix */
} /* of mu */
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to contract cvc: %e seconds\n", retime-ratime);
/************************************************
* save results for count = multiple of Nsave
************************************************/
if(count%Nsave == 0) {
if(g_cart_id == 0) fprintf(stdout, "save results for count = %d\n", count);
fnorm = 1. / ( (double)count * g_prop_normsqr );
if(g_cart_id==0) fprintf(stdout, "# X-fnorm = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(ix=0; ix<VOLUME; ix++) {
work[_GWI(mu,ix,VOLUME) ] = disc[_GWI(mu,ix,VOLUME) ] * fnorm + disc2[_GWI(mu,ix,VOLUME) ];
work[_GWI(mu,ix,VOLUME)+1] = disc[_GWI(mu,ix,VOLUME)+1] * fnorm + disc2[_GWI(mu,ix,VOLUME)+1];
}
}
/* save the result in position space */
sprintf(filename, "cvc_hpe5_X.%.4d.%.4d", Nconf, count);
sprintf(contype, "cvc-disc-all-hpe-05-X");
write_lime_contraction(work, filename, 64, 4, contype, Nconf, count);
/*
sprintf(filename, "cvc_hpe5_Xascii.%.4d.%.4d", Nconf, count);
write_contraction(work, NULL, filename, 4, 2, 0);
*/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/* Fourier transform data, copy to work */
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(work+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
memcpy((void*)(work+_GWI(4+mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
memcpy((void*)in, (void*)(work+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(work+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
} /* of mu =0 ,..., 3*/
fnorm = 1. / (double)(T_global*LX*LY*LZ);
if(g_cart_id==0) fprintf(stdout, "# P-fnorm = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(work+_GWI(mu,0,VOLUME));
cp2 = (complex*)(work+_GWI(4+nu,0,VOLUME));
cp3 = (complex*)(work+_GWI(8+4*mu+nu,0,VOLUME));
for(x0=0; x0<T; x0++) {
q[0] = (double)(x0+Tstart) / (double)T_global;
for(x1=0; x1<LX; x1++) {
q[1] = (double)(x1) / (double)LX;
for(x2=0; x2<LY; x2++) {
q[2] = (double)(x2) / (double)LY;
for(x3=0; x3<LZ; x3++) {
q[3] = (double)(x3) / (double)LZ;
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI * (q[mu]-q[nu]) );
w.im = sin( M_PI * (q[mu]-q[nu]) );
_co_eq_co_ti_co(&w1, cp1, cp2);
_co_eq_co_ti_co(cp3, &w1, &w);
_co_ti_eq_re(cp3, fnorm);
cp1++; cp2++; cp3++;
}
}
}
}
}
}
/* save the result in momentum space */
sprintf(filename, "cvc_hpe5_P.%.4d.%.4d", Nconf, count);
sprintf(contype, "cvc-disc-all-hpe-05-P");
write_lime_contraction(work+_GWI(8,0,VOLUME), filename, 64, 16, contype, Nconf, count);
/*
sprintf(filename, "cvc_hpe5_Pascii.%.4d.%.4d", Nconf, count);
write_contraction(work+_GWI(8,0,VOLUME), NULL, filename, 16, 2, 0);
*/
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to save cvc results: %e seconds\n", retime-ratime);
} /* of count % Nsave == 0 */
} /* of loop on sid */
/***********************************************
* free the allocated memory, finalize
***********************************************/
free(g_gauge_field);
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field);
free_geometry();
fftw_free(in);
free(disc);
free(work);
#ifdef MPI
fftwnd_mpi_destroy_plan(plan_p);
fftwnd_mpi_destroy_plan(plan_m);
MPI_Finalize();
#else
fftwnd_destroy_plan(plan_p);
fftwnd_destroy_plan(plan_m);
#endif
return(0);
}
int main(int argc, char **argv) {
int c, i, mu, nu;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix, iix;
int dxm[4], dxn[4], ixpm, ixpn;
int sid;
double *disc = (double*)NULL;
double *work = (double*)NULL;
double q[4], fnorm;
int verbose = 0;
int do_gt = 0;
char filename[100];
double ratime, retime;
double plaq, _2kappamu, hpe3_coeff, onepmutilde2, mutilde2;
double spinor1[24], spinor2[24], U_[18], U1_[18], U2_[18];
double *gauge_trafo=(double*)NULL;
complex w, w1, w2, *cp1, *cp2, *cp3;
FILE *ofs;
fftw_complex *in=(fftw_complex*)NULL;
#ifdef MPI
fftwnd_mpi_plan plan_p, plan_m;
int *status;
#else
fftwnd_plan plan_p, plan_m;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vgf:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'g':
do_gt = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
set_default_input_values();
if(filename_set==0) strcpy(filename, "cvc.input");
/* read the input file */
read_input(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.n");
usage();
}
/* initialize MPI parameters */
mpi_init(argc, argv);
#ifdef MPI
if((status = (int*)calloc(g_nproc, sizeof(int))) == (int*)NULL) {
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(7);
}
#endif
/* initialize fftw */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
#ifdef MPI
plan_p = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_BACKWARD, FFTW_MEASURE);
plan_m = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_FORWARD, FFTW_MEASURE);
fftwnd_mpi_local_sizes(plan_p, &T, &Tstart, &l_LX_at, &l_LXstart_at, &FFTW_LOC_VOLUME);
#else
plan_p = fftwnd_create_plan(4, dims, FFTW_BACKWARD, FFTW_MEASURE | FFTW_IN_PLACE);
plan_m = fftwnd_create_plan(4, dims, FFTW_FORWARD, FFTW_MEASURE | FFTW_IN_PLACE);
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
#endif
fprintf(stdout, "# [%2d] fftw parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(2);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(1);
}
geometry();
/* read the gauge field */
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
xchange_gauge();
/* measure the plaquette */
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value: %25.16e\n", plaq);
if(do_gt==1) {
/***********************************
* initialize gauge transformation
***********************************/
init_gauge_trafo(&gauge_trafo, 1.);
apply_gt_gauge(gauge_trafo);
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value after gauge trafo: %25.16e\n", plaq);
}
/****************************************
* allocate memory for the spinor fields
****************************************/
no_fields = 3;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/****************************************
* allocate memory for the contractions
****************************************/
disc = (double*)calloc( 8*VOLUME, sizeof(double));
if( disc == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
work = (double*)calloc(48*VOLUME, sizeof(double));
if( work == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for work\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(3);
}
/****************************************
* prepare Fourier transformation arrays
****************************************/
in = (fftw_complex*)malloc(FFTW_LOC_VOLUME*sizeof(fftw_complex));
if(in==(fftw_complex*)NULL) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(4);
}
/***********************************************
* start loop on source id.s
***********************************************/
for(sid=g_sourceid; sid<=g_sourceid2; sid++) {
/********************************
* read the first propagator
********************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, Nconf, sid);
if(read_lime_spinor(g_spinor_field[2], filename, 0) != 0) break;
}
else if(format==1) {
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, Nconf, sid);
if(read_cmi(g_spinor_field[2], filename) != 0) break;
}
xchange_field(g_spinor_field[2]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "time to read prop.: %e seconds\n", retime-ratime);
if(do_gt==1) {
/******************************************
* gauge transform the propagators for sid
******************************************/
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_cm_ti_fv(spinor1, gauge_trafo+18*ix, g_spinor_field[2]+_GSI(ix));
_fv_eq_fv(g_spinor_field[2]+_GSI(ix), spinor1);
}
xchange_field(g_spinor_field[2]);
}
/************************************************
* calculate the source: apply Q_phi_tbc
************************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
Q_phi_tbc(g_spinor_field[0], g_spinor_field[2]);
xchange_field(g_spinor_field[0]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "time to calculate source: %e seconds\n", retime-ratime);
/************************************************
* HPE: apply BH5
************************************************/
BH5(g_spinor_field[1], g_spinor_field[2]);
for(ix=0; ix<8*VOLUME; ix++) {disc[ix] = 0.;}
/* add new contractions to (existing) disc */
# ifdef MPI
ratime = MPI_Wtime();
# else
ratime = (double)clock() / CLOCKS_PER_SEC;
# endif
for(mu=0; mu<4; mu++) { /* loop on Lorentz index of the current */
iix = _GWI(mu,0,VOLUME);
for(ix=0; ix<VOLUME; ix++) { /* loop on lattice sites */
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix, mu)], &co_phase_up[mu]);
/* first contribution */
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(ix)], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
/* second contribution */
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(g_iup[ix][mu])], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
iix += 2;
} /* of ix */
} /* of mu */
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "[%2d] time to contract cvc: %e seconds\n", g_cart_id, retime-ratime);
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/* Fourier transform data, copy to work */
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(disc+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(work+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
}
/********************************
* read the second propagator
********************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, Nconf, sid+g_resume);
if(read_lime_spinor(g_spinor_field[2], filename, 0) != 0) break;
}
else if(format==1) {
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, Nconf, sid+g_resume);
if(read_cmi(g_spinor_field[2], filename) != 0) break;
}
xchange_field(g_spinor_field[2]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "time to read prop.: %e seconds\n", retime-ratime);
if(do_gt==1) {
/******************************************
* gauge transform the propagators for sid
******************************************/
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_cm_ti_fv(spinor1, gauge_trafo+18*ix, g_spinor_field[2]+_GSI(ix));
_fv_eq_fv(g_spinor_field[2]+_GSI(ix), spinor1);
}
xchange_field(g_spinor_field[2]);
}
/************************************************
* calculate the source: apply Q_phi_tbc
************************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
Q_phi_tbc(g_spinor_field[0], g_spinor_field[2]);
xchange_field(g_spinor_field[0]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "time to calculate source: %e seconds\n", retime-ratime);
/************************************************
* HPE: apply BH5
************************************************/
BH5(g_spinor_field[1], g_spinor_field[2]);
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
/* add new contractions to (existing) disc */
# ifdef MPI
ratime = MPI_Wtime();
# else
ratime = (double)clock() / CLOCKS_PER_SEC;
# endif
for(mu=0; mu<4; mu++) { /* loop on Lorentz index of the current */
iix = _GWI(mu,0,VOLUME);
for(ix=0; ix<VOLUME; ix++) { /* loop on lattice sites */
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix, mu)], &co_phase_up[mu]);
/* first contribution */
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(ix)], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
/* second contribution */
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(g_iup[ix][mu])], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
iix += 2;
} /* of ix */
} /* of mu */
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "[%2d] time to contract cvc: %e seconds\n", g_cart_id, retime-ratime);
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/* Fourier transform data, copy to work */
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(disc+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
memcpy((void*)(work+_GWI(4+mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
}
fnorm = 1. / ((double)(T_global*LX*LY*LZ));
fprintf(stdout, "fnorm = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(work+_GWI(mu,0,VOLUME));
cp2 = (complex*)(work+_GWI(4+nu,0,VOLUME));
cp3 = (complex*)(work+_GWI(8+4*mu+nu,0,VOLUME));
for(x0=0; x0<T; x0++) {
q[0] = (double)(x0+Tstart) / (double)T_global;
for(x1=0; x1<LX; x1++) {
q[1] = (double)(x1) / (double)LX;
for(x2=0; x2<LY; x2++) {
q[2] = (double)(x2) / (double)LY;
for(x3=0; x3<LZ; x3++) {
q[3] = (double)(x3) / (double)LZ;
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI * (q[mu]-q[nu]) );
w.im = sin( M_PI * (q[mu]-q[nu]) );
_co_eq_co_ti_co(&w1, cp1, cp2);
_co_eq_co_ti_co(cp3, &w1, &w);
_co_ti_eq_re(cp3, fnorm);
cp1++; cp2++; cp3++;
}
}
}
}
}
}
/* save the result in momentum space */
sprintf(filename, "cvc_hpe5_ft.%.4d.%.2d", Nconf, sid);
write_contraction(work+_GWI(8,0,VOLUME), NULL, filename, 16, 0, 0);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "time to save cvc results: %e seconds\n", retime-ratime);
} /* of loop on sid */
/***********************************************
* free the allocated memory, finalize
***********************************************/
free(g_gauge_field);
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field);
free_geometry();
fftw_free(in);
free(disc);
free(work);
#ifdef MPI
fftwnd_mpi_destroy_plan(plan_p);
fftwnd_mpi_destroy_plan(plan_m);
free(status);
MPI_Finalize();
#else
fftwnd_destroy_plan(plan_p);
fftwnd_destroy_plan(plan_m);
#endif
return(0);
}
/* for ieo=0, k resides on odd sites and l on even sites */
void Hopping_Matrix(int ieo, spinor * const l, spinor * const k){
int ix,iy;
int ioff,ioff2,icx,icy;
su3 * restrict up, * restrict um;
spinor * restrict r, * restrict sp, * restrict sm;
spinor temp;
#ifdef _GAUGE_COPY
if(g_update_gauge_copy) {
update_backward_gauge();
}
#endif
/* for parallelization */
# if (defined MPI && !(defined _NO_COMM))
xchange_field(k, ieo);
# endif
if(k == l){
printf("Error in H_psi (simple.c):\n");
printf("Arguments k and l must be different\n");
printf("Program aborted\n");
exit(1);
}
if(ieo == 0){
ioff = 0;
}
else{
ioff = (VOLUME+RAND)/2;
}
ioff2 = (VOLUME+RAND)/2-ioff;
/**************** loop over all lattice sites ****************/
for (icx = ioff; icx < (VOLUME/2 + ioff); icx++){
ix=g_eo2lexic[icx];
r=l+(icx-ioff);
/*********************** direction +0 ************************/
iy=g_iup[ix][0]; icy=g_lexic2eosub[iy];
sp=k+icy;
# if ((defined _GAUGE_COPY))
up=&g_gauge_field_copy[icx][0];
# else
up=&g_gauge_field[ix][0];
# endif
_vector_add(psi,(*sp).s0,(*sp).s2);
_su3_multiply(chi,(*up),psi);
_complex_times_vector(psi,ka0,chi);
_vector_assign(temp.s0,psi);
_vector_assign(temp.s2,psi);
_vector_add(psi,(*sp).s1,(*sp).s3);
_su3_multiply(chi,(*up),psi);
_complex_times_vector(psi,ka0,chi);
_vector_assign(temp.s1,psi);
_vector_assign(temp.s3,psi);
/*********************** direction -0 ************************/
iy=g_idn[ix][0]; icy=g_lexic2eosub[iy];
sm=k+icy;
# if ((defined _GAUGE_COPY))
um = up+1;
# else
um=&g_gauge_field[iy][0];
# endif
_vector_sub(psi,(*sm).s0,(*sm).s2);
_su3_inverse_multiply(chi,(*um),psi);
_complexcjg_times_vector(psi,ka0,chi);
_vector_add_assign(temp.s0,psi);
_vector_sub_assign(temp.s2,psi);
_vector_sub(psi,(*sm).s1,(*sm).s3);
_su3_inverse_multiply(chi,(*um),psi);
_complexcjg_times_vector(psi,ka0,chi);
_vector_add_assign(temp.s1,psi);
_vector_sub_assign(temp.s3,psi);
/*********************** direction +1 ************************/
iy=g_iup[ix][1]; icy=g_lexic2eosub[iy];
sp=k+icy;
# if ((defined _GAUGE_COPY))
up=um+1;
# else
up+=1;
# endif
_vector_i_add(psi,(*sp).s0,(*sp).s3);
_su3_multiply(chi,(*up),psi);
_complex_times_vector(psi,ka1,chi);
_vector_add_assign(temp.s0,psi);
_vector_i_sub_assign(temp.s3,psi);
_vector_i_add(psi,(*sp).s1,(*sp).s2);
_su3_multiply(chi,(*up),psi);
_complex_times_vector(psi,ka1,chi);
_vector_add_assign(temp.s1,psi);
_vector_i_sub_assign(temp.s2,psi);
/*********************** direction -1 ************************/
iy=g_idn[ix][1]; icy=g_lexic2eosub[iy];
sm=k+icy;
# ifndef _GAUGE_COPY
um=&g_gauge_field[iy][1];
# else
um=up+1;
# endif
_vector_i_sub(psi,(*sm).s0,(*sm).s3);
_su3_inverse_multiply(chi,(*um),psi);
_complexcjg_times_vector(psi,ka1,chi);
_vector_add_assign(temp.s0,psi);
_vector_i_add_assign(temp.s3,psi);
_vector_i_sub(psi,(*sm).s1,(*sm).s2);
_su3_inverse_multiply(chi,(*um),psi);
_complexcjg_times_vector(psi,ka1,chi);
_vector_add_assign(temp.s1,psi);
_vector_i_add_assign(temp.s2,psi);
/*********************** direction +2 ************************/
iy=g_iup[ix][2]; icy=g_lexic2eosub[iy];
sp=k+icy;
# if ((defined _GAUGE_COPY))
up=um+1;
# else
up+=1;
# endif
_vector_add(psi,(*sp).s0,(*sp).s3);
_su3_multiply(chi,(*up),psi);
_complex_times_vector(psi,ka2,chi);
_vector_add_assign(temp.s0,psi);
_vector_add_assign(temp.s3,psi);
_vector_sub(psi,(*sp).s1,(*sp).s2);
_su3_multiply(chi,(*up),psi);
_complex_times_vector(psi,ka2,chi);
_vector_add_assign(temp.s1,psi);
_vector_sub_assign(temp.s2,psi);
/*********************** direction -2 ************************/
iy=g_idn[ix][2]; icy=g_lexic2eosub[iy];
sm=k+icy;
# ifndef _GAUGE_COPY
um = &g_gauge_field[iy][2];
# else
um = up +1;
# endif
_vector_sub(psi,(*sm).s0,(*sm).s3);
_su3_inverse_multiply(chi,(*um),psi);
_complexcjg_times_vector(psi,ka2,chi);
_vector_add_assign(temp.s0,psi);
_vector_sub_assign(temp.s3,psi);
_vector_add(psi,(*sm).s1,(*sm).s2);
_su3_inverse_multiply(chi,(*um),psi);
_complexcjg_times_vector(psi,ka2,chi);
_vector_add_assign(temp.s1,psi);
_vector_add_assign(temp.s2,psi);
/*********************** direction +3 ************************/
iy=g_iup[ix][3]; icy=g_lexic2eosub[iy];
sp=k+icy;
# if ((defined _GAUGE_COPY))
up=um+1;
# else
up+=1;
# endif
_vector_i_add(psi,(*sp).s0,(*sp).s2);
_su3_multiply(chi,(*up),psi);
_complex_times_vector(psi,ka3,chi);
_vector_add_assign(temp.s0,psi);
_vector_i_sub_assign(temp.s2,psi);
_vector_i_sub(psi,(*sp).s1,(*sp).s3);
_su3_multiply(chi,(*up),psi);
_complex_times_vector(psi,ka3,chi);
_vector_add_assign(temp.s1,psi);
_vector_i_add_assign(temp.s3,psi);
/*********************** direction -3 ************************/
iy=g_idn[ix][3]; icy=g_lexic2eosub[iy];
sm=k+icy;
# ifndef _GAUGE_COPY
um = &g_gauge_field[iy][3];
# else
um = up+1;
# endif
_vector_i_sub(psi,(*sm).s0,(*sm).s2);
_su3_inverse_multiply(chi,(*um),psi);
_complexcjg_times_vector(psi,ka3,chi);
_vector_add((*r).s0, temp.s0, psi);
_vector_i_add((*r).s2, temp.s2, psi);
_vector_i_add(psi,(*sm).s1,(*sm).s3);
_su3_inverse_multiply(chi,(*um),psi);
_complexcjg_times_vector(psi,ka3,chi);
_vector_add((*r).s1, temp.s1, psi);
_vector_i_sub((*r).s3, temp.s3, psi);
/************************ end of loop ************************/
}
}
int main(int argc, char **argv) {
int c, i, j, mu, nu;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix, iix, it;
int sid, status, gid;
double **corr=NULL, **corr2=NULL;
double *tcorr=NULL, *tcorr2=NULL;
double *work = (double*)NULL;
double q[4], fnorm;
int verbose = 0;
int do_gt = 0;
int nsource=0;
char filename[100], contype[200];
double ratime, retime;
double plaq;
double spinor1[24], spinor2[24], U_[18];
double *gauge_trafo=(double*)NULL;
double mom2, mom4;
complex w, w1, *cp1, *cp2, *cp3;
FILE *ofs;
#ifdef MPI
// MPI_Init(&argc, &argv);
fprintf(stderr, "[jc_ud_x] Error, only non-mpi version implemented\n");
exit(1);
#endif
while ((c = getopt(argc, argv, "h?f:")) != -1) {
switch (c) {
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
if(filename_set==0) strcpy(filename, "cvc.input");
fprintf(stdout, "# Reading input from file %s\n", filename);
read_input_parser(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.n");
usage();
}
fprintf(stdout, "\n**************************************************\n");
fprintf(stdout, "* jc_ud_x\n");
fprintf(stdout, "**************************************************\n\n");
/*********************************
* initialize MPI parameters
*********************************/
// mpi_init(argc, argv);
/* initialize */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
fprintf(stdout, "# [%2d] parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
exit(1);
}
geometry();
/*************************************************
* allocate mem for gauge field and spinor fields
*************************************************/
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
no_fields = 2;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/****************************************
* allocate memory for the contractions
****************************************/
nsource = (g_sourceid2 - g_sourceid + 1) / g_sourceid_step;
if(g_cart_id==0) fprintf(stdout, "# nsource = %d\n", nsource);
corr = (double**)calloc( nsource, sizeof(double*));
corr[0] = (double*)calloc( nsource*T*8, sizeof(double));
for(i=1;i<nsource;i++) corr[i] = corr[i-1] + 8*T;
corr2 = (double**)calloc( nsource, sizeof(double*));
corr2[0] = (double*)calloc( nsource*8*T, sizeof(double));
for(i=1;i<nsource;i++) corr2[i] = corr2[i-1] + 8*T;
tcorr = (double*)calloc(T*8, sizeof(double));
tcorr2 = (double*)calloc(T*8, sizeof(double));
/***********************************************
* start loop on gauge id.s
***********************************************/
for(gid=g_gaugeid; gid<=g_gaugeid2; gid++) {
sprintf(filename, "%s.%.4d", gaugefilename_prefix, gid);
if(g_cart_id==0) fprintf(stdout, "# reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
xchange_gauge();
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "# measured plaquette value: %25.16e\n", plaq);
/* reset disc to zero */
for(ix=0; ix<nsource*8*T; ix++) corr[0][ix] = 0.;
for(ix=0; ix<nsource*8*T; ix++) corr2[0][ix] = 0.;
count=0;
/***********************************************
* start loop on source id.s
***********************************************/
for(sid=g_sourceid; sid<=g_sourceid2; sid+=g_sourceid_step) {
/* read the new propagator to g_spinor_field[0] */
ratime = (double)clock() / CLOCKS_PER_SEC;
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, gid, sid);
if(read_lime_spinor(g_spinor_field[0], filename, 0) != 0) break;
}
else if(format==1) {
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, gid, sid);
if(read_cmi(g_spinor_field[0], filename) != 0) {
fprintf(stderr, "\nError from read_cmi\n");
break;
}
}
xchange_field(g_spinor_field[0]);
retime = (double)clock() / CLOCKS_PER_SEC;
if(g_cart_id==0) fprintf(stdout, "# time to read prop.: %e seconds\n", retime-ratime);
ratime = (double)clock() / CLOCKS_PER_SEC;
/* apply [1] = D_tm [0] */
Q_phi_tbc(g_spinor_field[1], g_spinor_field[0]);
xchange_field(g_spinor_field[1]);
retime = (double)clock() / CLOCKS_PER_SEC;
if(g_cart_id==0) fprintf(stdout, "# time to apply D_W: %e seconds\n", retime-ratime);
ratime = (double)clock() / CLOCKS_PER_SEC;
/* calculate real and imaginary part */
for(mu=0; mu<4; mu++) {
for(x0=0; x0<T; x0++) {
for(x1=0; x1<LX; x1++) {
for(x2=0; x2<LY; x2++) {
for(x3=0; x3<LZ; x3++) {
ix = g_ipt[x0][x1][x2][x3];
_cm_eq_cm_ti_co(U_, g_gauge_field+_GGI(ix,mu), &(co_phase_up[mu]));
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[0][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[1][_GSI(ix)], spinor2);
corr[count][2*(mu*T+x0) ] -= 0.5*w.re;
corr[count][2*(mu*T+x0)+1] -= 0.5*w.im;
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[0][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[1][_GSI(g_iup[ix][mu])], spinor2);
corr[count][2*(mu*T+x0) ] -= 0.5*w.re;
corr[count][2*(mu*T+x0)+1] -= 0.5*w.im;
_fv_eq_gamma_ti_fv(spinor1, mu, &g_spinor_field[0][_GSI(ix)]);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[1][_GSI(ix)], spinor1);
corr2[count][2*(mu*T+x0) ] -= w.re;
corr2[count][2*(mu*T+x0)+1] -= w.im;
}}}
}
} // of mu
count++;
} // of sid
retime = (double)clock() / CLOCKS_PER_SEC;
if(g_cart_id==0) fprintf(stdout, "# time to calculate contractions: %e seconds\n", retime-ratime);
for(ix=0;ix<8*T;ix++) tcorr[ix] = 0.;
for(ix=0;ix<8*T;ix++) tcorr2[ix] = 0.;
for(i=0;i<nsource-1;i++) {
for(j=i+1;j<nsource;j++) {
for(mu=0;mu<4;mu++) {
for(x0=0;x0<T;x0++) { // times at source
for(x1=0;x1<T;x1++) { // times at sink
it = (x1 - x0 + T) % T;
// conserved current
tcorr[2*(mu*T+it) ] += corr[i][2*(mu*T+x1)] * corr[j][2*(mu*T+x0) ] - corr[i][2*(mu*T+x1)+1] * corr[j][2*(mu*T+x0)+1];
tcorr[2*(mu*T+it)+1] += corr[i][2*(mu*T+x1)] * corr[j][2*(mu*T+x0)+1] + corr[i][2*(mu*T+x1)+1] * corr[j][2*(mu*T+x0) ];
tcorr[2*(mu*T+it) ] += corr[j][2*(mu*T+x1)] * corr[i][2*(mu*T+x0) ] - corr[j][2*(mu*T+x1)+1] * corr[i][2*(mu*T+x0)+1];
tcorr[2*(mu*T+it)+1] += corr[j][2*(mu*T+x1)] * corr[i][2*(mu*T+x0)+1] + corr[j][2*(mu*T+x1)+1] * corr[i][2*(mu*T+x0) ];
// local current
tcorr2[2*(mu*T+it) ] += corr2[i][2*(mu*T+x1)] * corr2[j][2*(mu*T+x0) ] - corr2[i][2*(mu*T+x1)+1] * corr2[j][2*(mu*T+x0)+1];
tcorr2[2*(mu*T+it)+1] += corr2[i][2*(mu*T+x1)] * corr2[j][2*(mu*T+x0)+1] + corr2[i][2*(mu*T+x1)+1] * corr2[j][2*(mu*T+x0) ];
tcorr2[2*(mu*T+it) ] += corr2[j][2*(mu*T+x1)] * corr2[i][2*(mu*T+x0) ] - corr2[j][2*(mu*T+x1)+1] * corr2[i][2*(mu*T+x0)+1];
tcorr2[2*(mu*T+it)+1] += corr2[j][2*(mu*T+x1)] * corr2[i][2*(mu*T+x0)+1] + corr2[j][2*(mu*T+x1)+1] * corr2[i][2*(mu*T+x0) ];
}}
}
}}
fnorm = 1. / ( g_prop_normsqr * g_prop_normsqr * (double)(LX*LY*LZ) * (double)(LX*LY*LZ) * nsource * (nsource-1));
if(g_cart_id==0) fprintf(stdout, "X-fnorm = %e\n", fnorm);
for(ix=0;ix<8*T;ix++) tcorr[ix] *= fnorm;
for(ix=0;ix<8*T;ix++) tcorr2[ix] *= fnorm;
/************************************************
* save results
************************************************/
if(g_cart_id == 0) fprintf(stdout, "# save results for gauge id %d and sid %d\n", gid, sid);
/* save the result in position space */
sprintf(filename, "jc_u_tp0.%.4d.%.4d", gid, sid);
ofs = fopen(filename, "w");
for(x0=0;x0<T;x0++) fprintf(ofs, "%d%25.16e%25.16e%25.16e%25.16e%25.16e%25.16e%25.16e%25.16e\n", x0,
tcorr[2*(0*T+x0)], tcorr[2*(0*T+x0)+1],
tcorr[2*(1*T+x0)], tcorr[2*(1*T+x0)+1],
tcorr[2*(2*T+x0)], tcorr[2*(2*T+x0)+1],
tcorr[2*(3*T+x0)], tcorr[2*(3*T+x0)+1]);
fclose(ofs);
} /* of loop on gid */
/***********************************************
* free the allocated memory, finalize
***********************************************/
free(g_gauge_field);
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field);
free_geometry();
free(corr);
free(corr2);
free(tcorr);
free(tcorr2);
return(0);
}
int main(int argc, char **argv) {
int c, i, mu, nu;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix, iix;
int sid, status, gid;
double *disc = (double*)NULL;
double *work = (double*)NULL;
double q[4], fnorm;
int verbose = 0;
int do_gt = 0;
char filename[100], contype[200];
double ratime, retime;
double plaq;
double spinor1[24], spinor2[24], U_[18];
double *gauge_trafo=(double*)NULL;
complex w, w1, *cp1, *cp2, *cp3;
FILE *ofs;
#ifdef MPI
// MPI_Init(&argc, &argv);
fprintf(stderr, "[jc_ud_x] Error, only non-mpi version implemented\n");
exit(1);
#endif
while ((c = getopt(argc, argv, "h?f:")) != -1) {
switch (c) {
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
if(filename_set==0) strcpy(filename, "cvc.input");
fprintf(stdout, "# Reading input from file %s\n", filename);
read_input_parser(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.n");
usage();
}
fprintf(stdout, "\n**************************************************\n");
fprintf(stdout, "* jc_ud_x\n");
fprintf(stdout, "**************************************************\n\n");
/*********************************
* initialize MPI parameters
*********************************/
// mpi_init(argc, argv);
/* initialize fftw */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
fprintf(stdout, "# [%2d] parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
exit(1);
}
geometry();
/*************************************************
* allocate mem for gauge field and spinor fields
*************************************************/
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
no_fields = 2;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/****************************************
* allocate memory for the contractions
****************************************/
disc = (double*)calloc( 8*VOLUME, sizeof(double));
if( disc == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc\n");
exit(3);
}
/***********************************************
* start loop on gauge id.s
***********************************************/
for(gid=g_gaugeid; gid<=g_gaugeid2; gid++) {
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
sprintf(filename, "%s.%.4d", gaugefilename_prefix, gid);
if(g_cart_id==0) fprintf(stdout, "# reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
xchange_gauge();
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "# measured plaquette value: %25.16e\n", plaq);
/***********************************************
* start loop on source id.s
***********************************************/
for(sid=g_sourceid; sid<=g_sourceid2; sid+=g_sourceid_step) {
/* reset disc to zero */
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
/* read the new propagator to g_spinor_field[0] */
ratime = (double)clock() / CLOCKS_PER_SEC;
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, gid, sid);
if(read_lime_spinor(g_spinor_field[0], filename, 0) != 0) break;
}
else if(format==1) {
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, gid, sid);
if(read_cmi(g_spinor_field[0], filename) != 0) break;
}
xchange_field(g_spinor_field[0]);
retime = (double)clock() / CLOCKS_PER_SEC;
if(g_cart_id==0) fprintf(stdout, "# time to read prop.: %e seconds\n", retime-ratime);
ratime = (double)clock() / CLOCKS_PER_SEC;
/* apply D_W once, save in g_spinor_field[1] */
Hopping(g_spinor_field[1], g_spinor_field[0]);
for(ix=0; ix<VOLUME; ix++) {
_fv_pl_eq_fv(g_spinor_field[1]+_GSI(ix), g_spinor_field[0]+_GSI(ix));
_fv_ti_eq_re(g_spinor_field[1]+_GSI(ix), 1./(2.*g_kappa));
}
xchange_field(g_spinor_field[1]);
retime = (double)clock() / CLOCKS_PER_SEC;
if(g_cart_id==0) fprintf(stdout, "# time to apply D_W: %e seconds\n", retime-ratime);
ratime = (double)clock() / CLOCKS_PER_SEC;
/* calculate real and imaginary part */
for(mu=0; mu<4; mu++) {
for(ix=0; ix<VOLUME; ix++) {
_cm_eq_cm_ti_co(U_, g_gauge_field+_GGI(ix,mu), &(co_phase_up[mu]));
_fv_eq_gamma_ti_fv(spinor1, 5, g_spinor_field[0]+_GSI(g_iup[ix][mu]));
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_fv_eq_cm_ti_fv(spinor1, U_, spinor2);
_co_eq_fv_dag_ti_fv(&w, g_spinor_field[0]+_GSI(ix), spinor1);
disc[_GWI(mu,ix,VOLUME) ] = g_mu * w.im;
_fv_eq_gamma_ti_fv(spinor1, mu, g_spinor_field[1]+_GSI(g_iup[ix][mu]));
_fv_pl_eq_fv(spinor1, g_spinor_field[1]+_GSI(g_iup[ix][mu]));
_fv_eq_cm_ti_fv(spinor2, U_, spinor1);
_co_eq_fv_dag_ti_fv(&w, g_spinor_field[0]+_GSI(ix), spinor2);
disc[_GWI(mu,ix,VOLUME)+1] = w.im / 3.;
}
}
retime = (double)clock() / CLOCKS_PER_SEC;
if(g_cart_id==0) fprintf(stdout, "# time to calculate contractions: %e seconds\n", retime-ratime);
/************************************************
* save results
************************************************/
if(g_cart_id == 0) fprintf(stdout, "# save results for gauge id %d and sid %d\n", gid, sid);
/* save the result in position space */
fnorm = 1. / g_prop_normsqr;
if(g_cart_id==0) fprintf(stdout, "X-fnorm = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(ix=0; ix<VOLUME; ix++) {
disc[_GWI(mu,ix,VOLUME) ] *= fnorm;
disc[_GWI(mu,ix,VOLUME)+1] *= fnorm;
}
}
sprintf(filename, "jc_ud_x.%.4d.%.4d", gid, sid);
sprintf(contype, "jc-u_and_d-X");
write_lime_contraction(disc, filename, 64, 4, contype, gid, sid);
//sprintf(filename, "jc_ud_x.%.4d.%.4d.ascii", gid, sid);
//write_contraction (disc, NULL, filename, 4, 2, 0);
} /* of loop on sid */
} /* of loop on gid */
/***********************************************
* free the allocated memory, finalize
***********************************************/
free(g_gauge_field);
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field);
free_geometry();
free(disc);
return(0);
}
int main(int argc, char **argv) {
int c, i, mu, nu;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix, iix;
int sid, status;
double *disc = (double*)NULL;
double *data = (double*)NULL;
double *bias = (double*)NULL;
double *work = (double*)NULL;
double q[4], fnorm;
int verbose = 0;
char filename[100], contype[200];
double ratime, retime;
double plaq;
double spinor1[24], spinor2[24], U_[18];
complex w, w1, *cp1, *cp2, *cp3, *cp4;
fftw_complex *in=(fftw_complex*)NULL;
#ifdef MPI
fftwnd_mpi_plan plan_p, plan_m;
#else
fftwnd_plan plan_p, plan_m;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vf:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
if(filename_set==0) strcpy(filename, "cvc.input");
fprintf(stdout, "# Reading input from file %s\n", filename);
read_input_parser(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa <= 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.\n");
usage();
}
if(hpe_order%2==0 && hpe_order>0) {
if(g_proc_id==0) fprintf(stdout, "HPE order should be odd\n");
usage();
}
fprintf(stdout, "\n**************************************************\n"\
"* vp_disc_hpe_stoch_subtract with HPE of order %d\n"\
"**************************************************\n\n", hpe_order);
/*********************************
* initialize MPI parameters
*********************************/
mpi_init(argc, argv);
/* initialize fftw */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
#ifdef MPI
plan_p = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_BACKWARD, FFTW_MEASURE);
plan_m = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_FORWARD, FFTW_MEASURE);
fftwnd_mpi_local_sizes(plan_p, &T, &Tstart, &l_LX_at, &l_LXstart_at, &FFTW_LOC_VOLUME);
#else
plan_p = fftwnd_create_plan(4, dims, FFTW_BACKWARD, FFTW_MEASURE | FFTW_IN_PLACE);
plan_m = fftwnd_create_plan(4, dims, FFTW_FORWARD, FFTW_MEASURE | FFTW_IN_PLACE);
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
#endif
fprintf(stdout, "# [%2d] fftw parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(101);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(102);
}
geometry();
/************************************************
* read the gauge field, measure the plaquette
************************************************/
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "# reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
xchange_gauge();
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "# measured plaquette value: %25.16e\n", plaq);
/****************************************
* allocate memory for the spinor fields
****************************************/
no_fields = 3;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/****************************************
* allocate memory for the contractions
****************************************/
disc = (double*)calloc(16*VOLUME, sizeof(double));
if( disc == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(103);
}
data = (double*)calloc(16*VOLUME, sizeof(double));
if( data== (double*)NULL ) {
fprintf(stderr, "could not allocate memory for data\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(104);
}
for(ix=0; ix<16*VOLUME; ix++) data[ix] = 0.;
work = (double*)calloc(32*VOLUME, sizeof(double));
if( work == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for work\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(105);
}
bias = (double*)calloc(32*VOLUME, sizeof(double));
if( bias == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for bias\n");
# ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
# endif
exit(106);
}
for(ix=0; ix<32*VOLUME; ix++) bias[ix] = 0.;
/****************************************
* prepare Fourier transformation arrays
****************************************/
in = (fftw_complex*)malloc(FFTW_LOC_VOLUME*sizeof(fftw_complex));
if(in==(fftw_complex*)NULL) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(107);
}
/***********************************************
* start loop on source id.s
***********************************************/
for(sid=g_sourceid; sid<=g_sourceid2; sid+=g_sourceid_step) {
for(ix=0; ix<16*VOLUME; ix++) disc[ix] = 0.;
/* read the new propagator */
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, Nconf, sid);
if(read_lime_spinor(g_spinor_field[2], filename, 0) != 0) break;
}
else if(format==1) {
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, Nconf, sid);
if(read_cmi(g_spinor_field[2], filename) != 0) break;
}
xchange_field(g_spinor_field[2]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to read prop.: %e seconds\n", retime-ratime);
count++;
/************************************************
* calculate the source: apply Q_phi_tbc
************************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
Q_phi_tbc(g_spinor_field[0], g_spinor_field[2]);
xchange_field(g_spinor_field[0]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to calculate source: %e seconds\n", retime-ratime);
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/************************************************
* HPE: apply BH to order hpe_order+2
************************************************/
if(hpe_order>0) {
BHn(g_spinor_field[1], g_spinor_field[2], hpe_order+2);
} else {
memcpy((void*)g_spinor_field[1], (void*)g_spinor_field[2], 24*VOLUMEPLUSRAND*sizeof(double));
}
/************************************************
* add new contractions to (existing) disc
************************************************/
for(mu=0; mu<4; mu++) {
iix = _GWI(mu,0,VOLUME);
for(ix=0; ix<VOLUME; ix++) {
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix, mu)], &co_phase_up[mu]);
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(ix)], spinor2);
disc[iix ] = -0.5 * w.re;
disc[iix+1] = -0.5 * w.im;
data[iix ] -= 0.5 * w.re;
data[iix+1] -= 0.5 * w.im;
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(g_iup[ix][mu])], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
data[iix ] -= 0.5 * w.re;
data[iix+1] -= 0.5 * w.im;
iix += 2;
}
}
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to contract cvc: %e seconds\n", retime-ratime);
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(disc+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
memcpy((void*)(disc+_GWI(4+mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
memcpy((void*)in, (void*)(disc+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(disc+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
} /* of mu =0 ,..., 3*/
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(disc+_GWI(mu, 0,VOLUME));
cp2 = (complex*)(disc+_GWI(4+nu, 0,VOLUME));
cp3 = (complex*)(bias+_GWI(4*mu+nu,0,VOLUME));
for(ix=0; ix<VOLUME; ix++) {
_co_eq_co_ti_co(&w1, cp1, cp2);
cp3->re += w1.re;
cp3->im += w1.im;
cp1++; cp2++; cp3++;
}
}}
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time for Fourier trafo and adding to bias: %e seconds\n",
retime-ratime);
} /* of loop on sid */
/************************************************
* save results for count == Nsave
************************************************/
if(count==Nsave) {
if(g_cart_id == 0) fprintf(stdout, "# save results for count = %d\n", count);
for(ix=0; ix<16*VOLUME; ix++) disc[ix] = 0.;
if(hpe_order>0) {
sprintf(filename, "vp_disc_hpe%.2d_loops_X.%.4d", hpe_order, Nconf);
if(g_cart_id==0) fprintf(stdout, "# reading loop part from file %s\n", filename);
if( (status = read_lime_contraction(disc, filename, 4, 0)) != 0 ) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(108);
}
}
/* save the result in position space */
fnorm = 1. / ( (double)count * g_prop_normsqr );
if(g_cart_id==0) fprintf(stdout, "# X-fnorm = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(ix=0; ix<VOLUME; ix++) {
work[_GWI(mu,ix,VOLUME) ] = data[_GWI(mu,ix,VOLUME) ] * fnorm + disc[_GWI(mu,ix,VOLUME) ];
work[_GWI(mu,ix,VOLUME)+1] = data[_GWI(mu,ix,VOLUME)+1] * fnorm + disc[_GWI(mu,ix,VOLUME)+1];
}
}
sprintf(filename, "vp_disc_hpe%.2d_subtracted_X.%.4d.%.4d", hpe_order, Nconf, count);
sprintf(contype, "cvc-disc-hpe-loops-%2d-to-%2d-stoch-subtracted-X", hpe_order, hpe_order+2);
write_lime_contraction(work, filename, 64, 4, contype, Nconf, count);
/*
sprintf(filename, "vp_disc_hpe%.2d_subtracted_X.%.4d.%.4d.ascii", hpe_order, Nconf, count);
write_contraction(work, NULL, filename, 4, 2, 0);
*/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(data+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
memcpy((void*)(data+_GWI(4+mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
memcpy((void*)in, (void*)(data+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(data+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
}
fnorm = 1. / ( g_prop_normsqr*g_prop_normsqr * (double)count * (double)(count-1) );
if(g_cart_id==0) fprintf(stdout, "# P-fnorm for purely stochastic part = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(data+_GWI(mu, 0,VOLUME));
cp2 = (complex*)(data+_GWI(4+nu, 0,VOLUME));
cp3 = (complex*)(work+_GWI(4*mu+nu,0,VOLUME));
cp4 = (complex*)(bias+_GWI(4*mu+nu,0,VOLUME));
for(ix=0; ix<VOLUME; ix++) {
_co_eq_co_ti_co(&w1, cp1, cp2);
cp3->re = ( w1.re - cp4->re ) * fnorm;
cp3->im = ( w1.im - cp4->im ) * fnorm;
cp1++; cp2++; cp3++; cp4++;
}
}}
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(disc+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
memcpy((void*)(disc+_GWI(4+mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
memcpy((void*)in, (void*)(disc+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(disc+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
}
fnorm = 1. / ( g_prop_normsqr * (double)count );
if(g_cart_id==0) fprintf(stdout, "# P-fnorm for mixed stochastic-loop part = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(data + _GWI(mu, 0,VOLUME));
cp2 = (complex*)(disc + _GWI(4+nu, 0,VOLUME));
cp3 = (complex*)(work + _GWI(4*mu+nu,0,VOLUME));
for(ix=0; ix<VOLUME; ix++) {
_co_eq_co_ti_co(&w1, cp1, cp2);
cp3->re += w1.re * fnorm;
cp3->im += w1.im * fnorm;
cp1++; cp2++; cp3++;
}
cp1 = (complex*)(disc + _GWI(mu, 0,VOLUME));
cp2 = (complex*)(data + _GWI(4+nu, 0,VOLUME));
cp3 = (complex*)(work + _GWI(4*mu+nu,0,VOLUME));
for(ix=0; ix<VOLUME; ix++) {
_co_eq_co_ti_co(&w1, cp1, cp2);
cp3->re += w1.re * fnorm;
cp3->im += w1.im * fnorm;
cp1++; cp2++; cp3++;
}
}}
fnorm = 1. / ( (double)T_global * (double)(LX*LY*LZ) );
if(g_cart_id==0) fprintf(stdout, "# P-fnorm for final estimator (1/T/V) = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(disc + _GWI(mu, 0,VOLUME));
cp2 = (complex*)(disc + _GWI(4+nu, 0,VOLUME));
cp3 = (complex*)(work + _GWI(4*mu+nu,0,VOLUME));
for(x0=0; x0<T; x0++) {
q[0] = (double)(x0+Tstart) / (double)T_global;
for(x1=0; x1<LX; x1++) {
q[1] = (double)x1 / (double)LX;
for(x2=0; x2<LY; x2++) {
q[2] = (double)x2 / (double)LY;
for(x3=0; x3<LZ; x3++) {
q[3] = (double)x3 / (double)LZ;
ix = g_ipt[x0][x1][x2][x3];
w.re = cos(M_PI * ( q[mu] - q[nu] ) );
w.im = sin(M_PI * ( q[mu] - q[nu] ) );
_co_eq_co_ti_co(&w1, cp1, cp2);
cp3->re += w1.re;
cp3->im += w1.im;
_co_eq_co_ti_co(&w1, cp3, &w);
cp3->re = w1.re * fnorm;
cp3->im = w1.im * fnorm;
cp1++; cp2++; cp3++;
}}}}
}}
sprintf(filename, "vp_disc_hpe%.2d_subtracted_P.%.4d.%.4d", hpe_order, Nconf, count);
sprintf(contype, "cvc-disc-hpe-loops-%2d-to-%2d-stoch-subtracted-P", hpe_order, hpe_order+2);
write_lime_contraction(work, filename, 64, 16, contype, Nconf, count);
/*
sprintf(filename, "vp_disc_hpe%.2d_subtracted_P.%.4d.%.4d.ascii", hpe_order, Nconf, count);
write_contraction(work, NULL, filename, 16, 2, 0);
*/
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to save cvc results: %e seconds\n", retime-ratime);
} /* of if count == Nsave */
/***********************************************
* free the allocated memory, finalize
***********************************************/
free(g_gauge_field);
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field);
free_geometry();
fftw_free(in);
free(disc);
free(bias);
free(data);
free(work);
#ifdef MPI
fftwnd_mpi_destroy_plan(plan_p);
fftwnd_mpi_destroy_plan(plan_m);
MPI_Finalize();
#else
fftwnd_destroy_plan(plan_p);
fftwnd_destroy_plan(plan_m);
#endif
return(0);
}
int main(int argc, char **argv) {
int c, i, mu;
int count = 0;
int filename_set = 0;
int l_LX_at, l_LXstart_at;
int x0, x1, ix, idx;
int VOL3;
int sid;
double *disc = (double*)NULL;
int verbose = 0;
char filename[100];
double ratime, retime;
double plaq;
double spinor1[24], spinor2[24];
double _2kappamu;
double *gauge_field_f=NULL, *gauge_field_timeslice=NULL;
double v4norm = 0., vvnorm = 0.;
complex w;
FILE *ofs1, *ofs2;
/* double sign_adj5[] = {-1., -1., -1., -1., +1., +1., +1., +1., +1., +1., -1., -1., -1., 1., -1., -1.}; */
double hopexp_coeff[8], addreal, addimag;
int gindex[] = { 5 , 1 , 2 , 3 , 6 ,10 ,11 ,12 , 4 , 7 , 8 , 9 , 0 ,15 , 14 ,13 };
int isimag[] = { 0 , 0 , 0 , 0 , 1 , 1 , 1 , 1 , 0 , 1 , 1 , 1 , 0 , 1 , 1 , 1 };
double gsign[] = {-1., 1., 1., 1., -1., 1., 1., 1., 1., 1., 1., 1., 1., 1., -1., 1.};
#ifdef MPI
MPI_Status status;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vgf:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
if(filename_set==0) strcpy(filename, "cvc.input");
fprintf(stdout, "# reading input from file %s\n", filename);
read_input_parser(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.n");
usage();
}
/* initialize MPI parameters */
mpi_init(argc, argv);
#ifdef MPI
T = T_global / g_nproc;
Tstart = g_cart_id * T;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
VOL3 = LX*LY*LZ;
#else
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
VOL3 = LX*LY*LZ;
#endif
fprintf(stdout, "# [%2d] parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(1);
}
geometry();
/* read the gauge field */
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
xchange_gauge();
/* measure the plaquette */
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "# measured plaquette value: %25.16e\n", plaq);
if(Nlong > -1) {
/* N_ape = 5; */
alpha_ape = 0.4;
if(g_cart_id==0) fprintf(stdout, "# apply fuzzing of gauge field and propagators with parameters:\n"\
"# Nlong = %d\n# N_ape = %d\n# alpha_ape = %f\n", Nlong, N_ape, alpha_ape);
alloc_gauge_field(&gauge_field_f, VOLUMEPLUSRAND);
if( (gauge_field_timeslice = (double*)malloc(72*VOL3*sizeof(double))) == (double*)NULL ) {
fprintf(stderr, "Error, could not allocate mem for gauge_field_timeslice\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(2);
}
for(x0=0; x0<T; x0++) {
memcpy((void*)gauge_field_timeslice, (void*)(g_gauge_field+_GGI(g_ipt[x0][0][0][0],0)), 72*VOL3*sizeof(double));
for(i=0; i<N_ape; i++) {
APE_Smearing_Step_Timeslice(gauge_field_timeslice, alpha_ape);
}
fuzzed_links_Timeslice(gauge_field_f, gauge_field_timeslice, Nlong, x0);
}
free(gauge_field_timeslice);
}
/* test: print the fuzzed APE smeared gauge field to stdout */
/*
for(ix=0; ix<36*VOLUME; ix++) {
fprintf(stdout, "%6d%25.16e%25.16e%25.16e%25.16e\n", ix, gauge_field_f[2*ix], gauge_field_f[2*ix+1], g_gauge_field[2*ix], g_gauge_field[2*ix+1]);
}
*/
/* allocate memory for the spinor fields */
no_fields = 4;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/* allocate memory for the contractions */
disc = (double*)calloc(4*16*T*2, sizeof(double));
if( disc==(double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(3);
}
for(ix=0; ix<4*32*T; ix++) disc[ix] = 0.;
if(g_cart_id==0) {
sprintf(filename, "cvc_2pt_disc_vv.%.4d", Nconf);
ofs1 = fopen(filename, "w");
sprintf(filename, "cvc_2pt_disc_v4.%.4d", Nconf);
ofs2 = fopen(filename, "w");
if(ofs1==(FILE*)NULL || ofs2==(FILE*)NULL) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(5);
}
}
/* add the HPE coefficients */
if(format==1) {
addimag = 2*g_kappa*g_mu/sqrt(1 + 4*g_kappa*g_kappa*g_mu*g_mu)* LX*LY*LZ*3*4*2.*g_kappa*g_kappa*4;
addreal = 1./sqrt(1 + 4*g_kappa*g_kappa*g_mu*g_mu)*LX*LY*LZ*3*4*2.*g_kappa*g_kappa*4;
v4norm = 1. / ( 8. * g_kappa * g_kappa );
vvnorm = g_mu / ( 4. * g_kappa );
} else {
addimag = 2*g_kappa*g_mu/sqrt(1 + 4*g_kappa*g_kappa*g_mu*g_mu)* LX*LY*LZ*3*4*2.*g_kappa*2;
addreal = 1./sqrt(1 + 4*g_kappa*g_kappa*g_mu*g_mu)*LX*LY*LZ*3*4*2.*g_kappa*2;
v4norm = 1. / ( 4. * g_kappa );
vvnorm = g_mu / ( 4. * g_kappa );
}
/* calculate additional contributions for 1 and gamma_5 */
_2kappamu = 2.*g_kappa*g_mu;
hopexp_coeff[0] = 24. * g_kappa * LX*LY*LZ / (1. + _2kappamu*_2kappamu);
hopexp_coeff[1] = 0.;
hopexp_coeff[2] = -768. * g_kappa*g_kappa*g_kappa * LX*LY*LZ * _2kappamu*_2kappamu /
( (1.+_2kappamu*_2kappamu)*(1.+_2kappamu*_2kappamu)*(1.+_2kappamu*_2kappamu) );
hopexp_coeff[3] = 0.;
hopexp_coeff[4] = 0.;
hopexp_coeff[5] = -24.*g_kappa * LX*LY*LZ * _2kappamu / (1. + _2kappamu*_2kappamu);
hopexp_coeff[6] = 0.;
hopexp_coeff[7] = -384. * g_kappa*g_kappa*g_kappa * LX*LY*LZ *
(1.-_2kappamu*_2kappamu)*_2kappamu /
( (1.+_2kappamu*_2kappamu)*(1.+_2kappamu*_2kappamu)*(1.+_2kappamu*_2kappamu) );
/* start loop on source id.s */
for(sid=g_sourceid; sid<=g_sourceid2; sid+=g_sourceid_step) {
for(ix=0; ix<4*32*T; ix++) disc[ix] = 0.;
/* read the new propagator */
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, Nconf, sid);
/* sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, Nconf, sid); */
if(read_lime_spinor(g_spinor_field[1], filename, 0) != 0) {
fprintf(stderr, "[%2d] Error, could not read from file %s\n", g_cart_id, filename);
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(4);
}
count++;
xchange_field(g_spinor_field[1]);
/* calculate the source: apply Q_phi_tbc */
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
Q_phi_tbc(g_spinor_field[0], g_spinor_field[1]);
xchange_field(g_spinor_field[0]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time to apply Q_tm %e seconds\n", retime-ratime);
/* apply gamma5_BdagH4_gamma5 */
gamma5_BdagH4_gamma5(g_spinor_field[2], g_spinor_field[0], g_spinor_field[3]);
/* attention: additional factor 2kappa because of CMI format */
/*
if(format==1) {
for(ix=0; ix<VOLUME; ix++) {
_fv_ti_eq_re(&g_spinor_field[2][_GSI(ix)], 2.*g_kappa);
}
}
*/
if(Nlong>-1) {
if(g_cart_id==0) fprintf(stdout, "# fuzzing propagator with Nlong = %d\n", Nlong);
memcpy((void*)g_spinor_field[3], (void*)g_spinor_field[1], 24*VOLUMEPLUSRAND*sizeof(double));
Fuzz_prop(gauge_field_f, g_spinor_field[3], Nlong);
}
/* add new contractions to disc */
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
for(x0=0; x0<T; x0++) { /* loop on time */
for(x1=0; x1<VOL3; x1++) { /* loop on sites in timeslice */
ix = x0*VOL3 + x1;
for(mu=0; mu<16; mu++) { /* loop on index of gamma matrix */
_fv_eq_gamma_ti_fv(spinor1, mu, &g_spinor_field[1][_GSI(ix)]);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[2][_GSI(ix)], spinor1);
disc[2*( x0*16+mu) ] += w.re;
disc[2*( x0*16+mu)+1] += w.im;
_fv_eq_gamma_ti_fv(spinor1, 5, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[1][_GSI(ix)], spinor2);
disc[2*(16*T + x0*16+mu) ] += w.re;
disc[2*(16*T + x0*16+mu)+1] += w.im;
if(Nlong>-1) {
_fv_eq_gamma_ti_fv(spinor1, mu, &g_spinor_field[3][_GSI(ix)]);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[2][_GSI(ix)], spinor1);
disc[2*(32*T + x0*16+mu) ] += w.re;
disc[2*(32*T + x0*16+mu)+1] += w.im;
_fv_eq_gamma_ti_fv(spinor1, 5, &g_spinor_field[3][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[1][_GSI(ix)], spinor2);
disc[2*(48*T + x0*16+mu) ] += w.re;
disc[2*(48*T + x0*16+mu)+1] += w.im;
}
}
}
}
if(g_cart_id==0) fprintf(stdout, "# addimag = %25.16e\n", addimag);
if(g_cart_id==0) fprintf(stdout, "# addreal = %25.16e\n", addreal);
for(x0=0; x0<T; x0++) {
disc[2*( x0*16+4) ] += addreal;
disc[2*( x0*16+5)+1] -= addimag;
/*
if(Nlong>-1) {
disc[2*(32*T + x0*16+4) ] += addreal;
disc[2*(32*T + x0*16+5)+1] -= addimag;
}
*/
}
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# contractions in %e seconds\n", retime-ratime);
/* write current disc to file */
if(g_cart_id==0) {
if(sid==g_sourceid) fprintf(ofs1, "#%6d%3d%3d%3d%3d\t%f\t%f\n", Nconf, T, LX, LY, LZ, g_kappa, g_mu);
if(sid==g_sourceid) fprintf(ofs2, "#%6d%3d%3d%3d%3d\t%f\t%f\n", Nconf, T, LX, LY, LZ, g_kappa, g_mu);
for(x0=0; x0<T; x0++) {
for(mu=0; mu<16; mu++) {
idx = gindex[mu];
ix = 16*x0 + idx;
if(isimag[mu]==0) {
fprintf(ofs2, "%6d%3d%4d%4d%25.16e%25.16e%25.16e%25.16e\n",
Nconf, mu, x0, sid,
gsign[mu]*disc[2* ix ]*v4norm, gsign[mu]*disc[2* ix +1]*v4norm,
gsign[mu]*disc[2*(32*T+ix)]*v4norm, gsign[mu]*disc[2*(32*T+ix)+1]*v4norm);
} else {
fprintf(ofs2, "%6d%3d%4d%4d%25.16e%25.16e%25.16e%25.16e\n",
Nconf, mu, x0, sid,
gsign[mu]*disc[2*( ix)+1]*v4norm, -gsign[mu]*disc[2* ix ]*v4norm,
gsign[mu]*disc[2*(32*T+ix)+1]*v4norm, -gsign[mu]*disc[2*(32*T+ix)]*v4norm);
}
}
}
for(x0=0; x0<T; x0++) {
for(mu=0; mu<16; mu++) {
idx = gindex[mu];
ix = 16*x0 + idx;
if(isimag[mu]==0) {
fprintf(ofs1, "%6d%3d%4d%4d%25.16e%25.16e%25.16e%25.16e\n",
Nconf, mu, x0, sid,
gsign[mu]*disc[2*(16*T+ix)+1]*vvnorm, -gsign[mu]*disc[2*(16*T+ix)]*vvnorm,
gsign[mu]*disc[2*(48*T+ix)+1]*vvnorm, -gsign[mu]*disc[2*(48*T+ix)]*vvnorm);
} else {
fprintf(ofs1, "%6d%3d%4d%4d%25.16e%25.16e%25.16e%25.16e\n",
Nconf, mu, x0, sid,
-gsign[mu]*disc[2*(16*T+ix)]*vvnorm, -gsign[mu]*disc[2*(16*T+ix)+1]*vvnorm,
-gsign[mu]*disc[2*(48*T+ix)]*vvnorm, -gsign[mu]*disc[2*(48*T+ix)+1]*vvnorm);
}
}
}
#ifdef MPI
for(c=1; c<g_nproc; c++) {
MPI_Recv(disc, 128*T, MPI_DOUBLE, c, 100+c, g_cart_grid, &status);
for(x0=0; x0<T; x0++) {
for(mu=0; mu<16; mu++) {
idx=gindex[mu];
ix = 16*x0 + idx;
if(isimag[mu]==0) {
fprintf(ofs2, "%6d%3d%4d%4d%25.16e%25.16e%25.16e%25.16e\n",
Nconf, mu, c*T+x0, sid,
gsign[mu]*disc[2* ix ]*v4norm, gsign[mu]*disc[2* ix +1]*v4norm,
gsign[mu]*disc[2*(32*T+ix)]*v4norm, gsign[mu]*disc[2*(32*T+ix)+1]*v4norm);
} else {
fprintf(ofs2, "%6d%3d%4d%4d%25.16e%25.16e%25.16e%25.16e\n",
Nconf, mu, c*T+x0, sid,
gsign[mu]*disc[2*( ix)+1]*v4norm, -gsign[mu]*disc[2* ix ]*v4norm,
gsign[mu]*disc[2*(32*T+ix)+1]*v4norm, -gsign[mu]*disc[2*(32*T+ix)]*v4norm);
}
}
}
for(x0=0; x0<T; x0++) {
for(mu=0; mu<16; mu++) {
idx = gindex[mu];
ix = 16*x0 + idx;
if(isimag[mu]==0) {
fprintf(ofs1, "%6d%3d%4d%4d%25.16e%25.16e%25.16e%25.16e\n",
Nconf, mu, c*T+x0, sid,
gsign[mu]*disc[2*(16*T+ix)+1]*vvnorm, -gsign[mu]*disc[2*(16*T+ix)]*vvnorm,
gsign[mu]*disc[2*(48*T+ix)+1]*vvnorm, -gsign[mu]*disc[2*(48*T+ix)]*vvnorm);
} else {
fprintf(ofs1, "%6d%3d%4d%4d%25.16e%25.16e%25.16e%25.16e\n",
Nconf, mu, c*T+x0, sid,
-gsign[mu]*disc[2*(16*T+ix)]*vvnorm, -gsign[mu]*disc[2*(16*T+ix)+1]*vvnorm,
-gsign[mu]*disc[2*(48*T+ix)]*vvnorm, -gsign[mu]*disc[2*(48*T+ix)+1]*vvnorm);
}
}
}
}
#endif
}
#ifdef MPI
else {
for(c=1; c<g_nproc; c++) {
if(g_cart_id==c) {
MPI_Send(disc, 128*T, MPI_DOUBLE, 0, 100+c, g_cart_grid);
}
}
}
#endif
} /* of loop on sid */
if(g_cart_id==0) { fclose(ofs1); fclose(ofs2); }
if(g_cart_id==0) {
fprintf(stdout, "# contributions from HPE:\n");
fprintf(stdout, "(1) X = id\t%25.16e%25.16e\n"\
" \t%25.16e%25.16e\n"\
"(2) X = 5\t%25.16e%25.16e\n"\
" \t%25.16e%25.16e\n",
hopexp_coeff[0], hopexp_coeff[1], hopexp_coeff[2], hopexp_coeff[3],
hopexp_coeff[4], hopexp_coeff[5], hopexp_coeff[6], hopexp_coeff[7]);
}
/* free the allocated memory, finalize */
free(g_gauge_field); g_gauge_field=(double*)NULL;
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field); g_spinor_field=(double**)NULL;
free_geometry();
free(disc);
if(Nlong>-1) free(gauge_field_f);
#ifdef MPI
MPI_Finalize();
#endif
return(0);
}
int main(int argc, char **argv) {
int c, i, mu, nu;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix, iix;
int sx0, sx1, sx2, sx3;
int sid;
double *disc = (double*)NULL;
double *disc2 = (double*)NULL;
double *work = (double*)NULL;
double q[4], fnorm;
double cvc_lnuy[8];
double *gauge_trafo=(double*)NULL;
double unit_trace[2], D_trace[2];
int verbose = 0;
int do_gt = 0;
char filename[100];
double ratime, retime;
double plaq;
double spinor1[24], spinor2[24], U_[18];
complex w, w1, *cp1, *cp2, *cp3;
FILE *ofs;
fftw_complex *in=(fftw_complex*)NULL;
#ifdef MPI
fftwnd_mpi_plan plan_p;
int *status;
#else
fftwnd_plan plan_p;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vgf:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'g':
do_gt = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
set_default_input_values();
if(filename_set==0) strcpy(filename, "cvc.input");
/* read the input file */
read_input(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.n");
usage();
}
/* initialize MPI parameters */
mpi_init(argc, argv);
#ifdef MPI
if((status = (int*)calloc(g_nproc, sizeof(int))) == (int*)NULL) {
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(7);
}
#endif
/* initialize fftw */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
#ifdef MPI
plan_p = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_BACKWARD, FFTW_MEASURE);
fftwnd_mpi_local_sizes(plan_p, &T, &Tstart, &l_LX_at, &l_LXstart_at, &FFTW_LOC_VOLUME);
#else
plan_p = fftwnd_create_plan(4, dims, FFTW_BACKWARD, FFTW_MEASURE | FFTW_IN_PLACE);
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
#endif
fprintf(stdout, "# [%2d] fftw parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(2);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(1);
}
geometry();
/* read the gauge field */
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
xchange_gauge();
/* measure the plaquette */
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value: %25.16e\n", plaq);
/* get the source location coordinates */
sx0 = g_source_location / (LX*LY*LZ );
sx1 = ( g_source_location % (LX*LY*LZ) ) / (LY*LZ);
sx2 = ( g_source_location % (LY*LZ) ) / LZ;
sx3 = ( g_source_location % LZ );
/* read the data for lnuy */
sprintf(filename, "cvc_lnuy_X.%.4d", Nconf);
ofs = fopen(filename, "r");
fprintf(stdout, "reading cvc lnuy from file %s\n", filename);
for(mu=0; mu<4; mu++) {
fscanf(ofs, "%lf%lf", cvc_lnuy+2*mu, cvc_lnuy+2*mu+1);
}
fclose(ofs);
/* allocate memory for the spinor fields */
no_fields = 2;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/****************************************
* allocate memory for the contractions
****************************************/
disc = (double*)calloc(8*VOLUME, sizeof(double));
if( disc == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(3);
}
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
disc2 = (double*)calloc(8*VOLUME, sizeof(double));
if( disc2 == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc2\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(3);
}
for(ix=0; ix<8*VOLUME; ix++) disc2[ix] = 0.;
work = (double*)calloc(48*VOLUME, sizeof(double));
if( work == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for work\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(3);
}
/****************************************
* prepare Fourier transformation arrays
****************************************/
in = (fftw_complex*)malloc(FFTW_LOC_VOLUME*sizeof(fftw_complex));
if(in==(fftw_complex*)NULL) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(4);
}
if(g_resume==1) { /* read current disc from file */
sprintf(filename, ".outcvc_current.%.4d", Nconf);
c = read_contraction(disc, &count, filename, 8);
#ifdef MPI
MPI_Gather(&c, 1, MPI_INT, status, 1, MPI_INT, 0, g_cart_grid);
if(g_cart_id==0) {
/* check the entries in status */
for(i=0; i<g_nproc; i++)
if(status[i]!=0) { status[0] = 1; break; }
}
MPI_Bcast(status, 1, MPI_INT, 0, g_cart_grid);
if(status[0]==1) {
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
count = 0;
}
#else
if(c != 0) {
fprintf(stdout, "could not read current disc; start new\n");
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
count = 0;
}
#endif
if(g_cart_id==0) fprintf(stdout, "starting with count = %d\n", count);
} /* of g_resume == 1 */
if(do_gt==1) {
/***********************************
* initialize gauge transformation
***********************************/
init_gauge_trafo(&gauge_trafo,1.0);
fprintf(stdout, "applying gauge trafo to gauge field\n");
apply_gt_gauge(gauge_trafo);
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "plaquette plaq = %25.16e\n", plaq);
}
unit_trace[0] = 0.;
unit_trace[1] = 0.;
D_trace[0] = 0.;
D_trace[1] = 0.;
/****************************************
* start loop on source id.s
****************************************/
for(sid=g_sourceid; sid<=g_sourceid2; sid++) {
/****************************************
* read the new propagator
****************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/****************************************
* check: write source before D-appl.
****************************************/
/*
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d", filename_prefix, Nconf, sid);
read_lime_spinor(g_spinor_field[0], filename, 0);
}
for(ix=0; ix<12*VOLUME; ix++) {
fprintf(stdout, "source: %6d%25.16e%25.16e\n", ix, g_spinor_field[0][2*ix], g_spinor_field[0][2*ix+1]);
}
*/
if(format==0) {
sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, Nconf, sid);
/* sprintf(filename, "%s.%.4d.%.2d", filename_prefix, Nconf, sid); */
if(read_lime_spinor(g_spinor_field[1], filename, 0) != 0) break;
}
else if(format==1) {
sprintf(filename, "%s.%.4d.%.5d.inverted", filename_prefix, Nconf, sid);
if(read_cmi(g_spinor_field[1], filename) != 0) break;
}
xchange_field(g_spinor_field[1]);
if(do_gt==1) {
fprintf(stdout, "applying gt on propagators\n");
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_cm_ti_fv(spinor1, gauge_trafo+18*ix, g_spinor_field[1]+_GSI(ix));
_fv_eq_fv(g_spinor_field[1]+_GSI(ix), spinor1);
}
}
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "time to read prop.: %e seconds\n", retime-ratime);
count++;
/****************************************
* calculate the source: apply Q_phi_tbc
****************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
Q_phi_tbc(g_spinor_field[0], g_spinor_field[1]);
xchange_field(g_spinor_field[0]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "time to calculate source: %e seconds\n", retime-ratime);
/****************************************
* check: write source after D-appl.
****************************************/
/*
for(ix=0; ix<12*VOLUME; ix++) {
fprintf(stdout, "D_source: %6d%25.16e%25.16e\n", ix, g_spinor_field[0][2*ix], g_spinor_field[0][2*ix+1]);
}
*/
/****************************************
* add new contractions to (existing) disc
****************************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
for(mu=0; mu<4; mu++) { /* loop on Lorentz index of the current */
iix = _GWI(mu,0,VOLUME);
for(ix=0; ix<VOLUME; ix++) { /* loop on lattice sites */
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix, mu)], &co_phase_up[mu]);
/* first contribution */
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(ix)], spinor2);
disc[iix ] -= 0.5 * w.re;
disc[iix+1] -= 0.5 * w.im;
/* second contribution */
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(g_iup[ix][mu])], spinor2);
disc2[iix ] -= 0.5 * w.re;
disc2[iix+1] -= 0.5 * w.im;
iix += 2;
} /* of ix */
} /* of mu */
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "[%2d] contractions for CVC in %e seconds\n", g_cart_id, retime-ratime);
/***************************************************
* check: convergence of trace of unit matrix
***************************************************/
_co_eq_fv_dag_ti_fv(&w, g_spinor_field[0]+_GSI(g_source_location), g_spinor_field[0]+_GSI(g_source_location));
unit_trace[0] += w.re;
unit_trace[1] += w.im;
fprintf(stdout, "unit_trace: %4d%25.16e%25.16e\n", count, w.re, w.im);
_co_eq_fv_dag_ti_fv(&w, g_spinor_field[0]+_GSI(g_source_location), g_spinor_field[0]+_GSI(g_iup[g_source_location][0]));
fprintf(stdout, "shift_trace: %4d%25.16e%25.16e\n", count, w.re, w.im);
/***************************************************
* check: convergence of trace D_u(source_location, source_location)
***************************************************/
Q_phi_tbc(g_spinor_field[1], g_spinor_field[0]);
_co_eq_fv_dag_ti_fv(&w, g_spinor_field[0]+_GSI(g_source_location), g_spinor_field[1]+_GSI(g_source_location));
D_trace[0] += w.re;
D_trace[1] += w.im;
/* fprintf(stdout, "D_trace: %4d%25.16e%25.16e\n", count, D_trace[0]/(double)count, D_trace[1]/(double)count); */
fprintf(stdout, "D_trace: %4d%25.16e%25.16e\n", count, w.re, w.im);
/***************************************************
* save results for count = multiple of Nsave
***************************************************/
if(count%Nsave == 0) {
if(g_cart_id == 0) fprintf(stdout, "save results for count = %d\n", count);
/* save the result in position space */
/* divide by number of propagators */
for(ix=0; ix<8*VOLUME; ix++) work[ix] = disc[ix] / (double)count;
sprintf(filename, "outcvc_Xm.%.4d.%.4d", Nconf, count);
write_contraction(work, NULL, filename, 4, 2, 0);
for(ix=0; ix<8*VOLUME; ix++) work[ix] = disc2[ix] / (double)count;
sprintf(filename, "outcvc_Xp.%.4d.%.4d", Nconf, count);
write_contraction(work, NULL, filename, 4, 2, 0);
for(ix=0; ix<8*VOLUME; ix++) work[ix] = (disc[ix] + disc2[ix]) / (double)count;
sprintf(filename, "outcvc_X.%.4d.%.4d", Nconf, count);
write_contraction(work, NULL, filename, 4, 2, 0);
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/****************************************
* Fourier transform data, copy to work
****************************************/
for(mu=0; mu<4; mu++) {
memcpy((void*)in, (void*)(work+_GWI(mu,0,VOLUME)), 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
memcpy((void*)(work+_GWI(mu,0,VOLUME)), (void*)in, 2*VOLUME*sizeof(double));
} /* of mu =0 ,..., 3*/
/* fnorm = 1. / ((double)count); */
fprintf(stdout, "fnorm = %e\n", fnorm);
for(mu=0; mu<4; mu++) {
for(nu=0; nu<4; nu++) {
cp1 = (complex*)(work+_GWI(mu,0,VOLUME));
cp2 = (complex*)(cvc_lnuy+2*nu);
cp3 = (complex*)(work+_GWI(4+4*mu+nu,0,VOLUME));
for(x0=0; x0<T; x0++) {
q[0] = (double)(x0+Tstart) / (double)T_global;
for(x1=0; x1<LX; x1++) {
q[1] = (double)(x1) / (double)LX;
for(x2=0; x2<LY; x2++) {
q[2] = (double)(x2) / (double)LY;
for(x3=0; x3<LZ; x3++) {
q[3] = (double)(x3) / (double)LZ;
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI * ( q[mu] - q[nu] - 2.*(sx0*q[0]+sx1*q[1]+sx2*q[2]+sx3*q[3])) );
w.im = sin( M_PI * ( q[mu] - q[nu] - 2.*(sx0*q[0]+sx1*q[1]+sx2*q[2]+sx3*q[3])) );
/* fprintf(stdout, "mu=%3d, nu=%3d, t=%3d, x=%3d, y=%3d, z=%3d, phase= %21.12e + %21.12ei\n", \
mu, nu, x0, x1, x2, x3, w.re, w.im); */
_co_eq_co_ti_co(&w1, cp1, cp2);
_co_eq_co_ti_co(cp3, &w1, &w);
/* _co_ti_eq_re(cp3, fnorm); */
cp1++; cp3++;
}
}
}
}
}
}
/* save the result in momentum space */
sprintf(filename, "outcvc_P.%.4d.%.4d", Nconf, count);
write_contraction(work+_GWI(4,0,VOLUME), NULL, filename, 16, 2, 0);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "time to cvc save results: %e seconds\n", retime-ratime);
} /* of count % Nsave == 0 */
} /* of loop on sid */
if(g_resume==1) {
/* write current disc to file */
sprintf(filename, ".outcvc_current.%.4d", Nconf);
write_contraction(disc, &count, filename, 4, 0, 0);
}
/**************************************
* free the allocated memory, finalize
**************************************/
free(g_gauge_field);
if(do_gt==1) free(gauge_trafo);
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field);
free_geometry();
fftw_free(in);
free(disc); free(disc2);
free(work);
#ifdef MPI
fftwnd_mpi_destroy_plan(plan_p);
free(status);
MPI_Finalize();
#else
fftwnd_destroy_plan(plan_p);
#endif
return(0);
}
int main(int argc, char **argv) {
int c, i, mu, K=16;
int count = 0;
int filename_set = 0;
int x0;
int estat; // exit status
unsigned long int ix, idx;
unsigned long int x1, VOL3, index_min;
int sid;
double *disc = (double*)NULL, *buffer=NULL, *buffer2=NULL;
int verbose = 0;
char filename[100];
double ratime, retime;
double plaq;
double spinor1[24];
double *gauge_field_f=NULL, *gauge_field_timeslice=NULL;
double v4norm = 0., vvnorm = 0.;
double *psi0 = NULL, *psi1 = NULL, *psi2 = NULL, *psi3 = NULL;
complex w;
FILE *ofs[4];
double addreal, addimag;
/* Initialise all the gamma matrix combinations
g5, g1, g2, g3, ig0g5, ig0gi, -1, -g5gi, g0 -g5g0gi */
int gindex[] = {5, 1, 2, 3, 6, 7, 8, 9, 4, 10, 11, 12, 0, 13, 14, 15};
double gsign[] = {-1., 1., 1., 1., -1., 1., 1., 1., 1., 1., 1., 1., 1., 1., -1., 1.};
#ifdef MPI
MPI_Status status;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vgf:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
// local time stamp
g_the_time = time;
if(g_cart_id == 0) {
fprintf(stdout, "\n# [disc] using global time stamp %s", ctime(&g_the_time));
}
/* set the default values */
if(filename_set==0) strcpy(filename, "cvc.input");
fprintf(stdout, "# reading input from file %s\n", filename);
read_input_parser(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.n");
usage();
}
/* initialize MPI parameters */
mpi_init(argc, argv);
#ifdef MPI
# if ! ( (defined PARALLELTX) || (defined PARALLELTXY) )
T = T_global / g_nproc;
Tstart = g_cart_id * T;
# endif
#else
T = T_global;
Tstart = 0;
#endif
VOL3 = LX*LY*LZ;
fprintf(stdout, "# [%2d] parameters:\n"\
"# [%2d] T_global = %3d\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] LX_global = %3d\n"\
"# [%2d] LX = %3d\n"\
"# [%2d] LXstart = %3d\n",
g_cart_id, g_cart_id, T_global, g_cart_id, T, g_cart_id, Tstart, g_cart_id, LX_global, g_cart_id, LX,
g_cart_id, LXstart);
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(1);
}
geometry();
/**********************************************
* read the gauge field
**********************************************/
// if(N_ape>0 || Nlong>0) {
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
xchange_gauge();
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "# measured plaquette value: %25.16e\n", plaq);
// } else {
// g_gauge_field = (double*)NULL;
// }
if(Nlong > 0) {
// N_ape = 1;
// alpha_ape = 0.4;
if(g_cart_id==0) fprintf(stdout, "# apply fuzzing of gauge field and propagators with parameters:\n"\
"# Nlong = %d\n# N_ape = %d\n# alpha_ape = %f\n", Nlong, N_ape, alpha_ape);
alloc_gauge_field(&gauge_field_f, VOLUMEPLUSRAND);
#if !( (defined PARALLELTX) || (defined PARALLELTXY) )
gauge_field_timeslice = (double*)calloc(72*VOL3, sizeof(double));
if( gauge_field_timeslice == (double*)NULL ) {
fprintf(stderr, "Error, could not allocate mem for gauge_field_timeslice\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(2);
}
for(x0=0; x0<T; x0++) {
memcpy((void*)gauge_field_timeslice, (void*)(g_gauge_field+_GGI(g_ipt[x0][0][0][0],0)), 72*VOL3*sizeof(double));
for(i=0; i<N_ape; i++) {
fprintf(stdout, "# [] APE smearing time slice %d step %d\n", x0, i);
APE_Smearing_Step_Timeslice(gauge_field_timeslice, alpha_ape);
}
if(Nlong > 0) {
fuzzed_links_Timeslice(gauge_field_f, gauge_field_timeslice, Nlong, x0);
} else {
memcpy(gauge_field_f+_GGI(g_ipt[x0][0][0][0], 0), gauge_field_timeslice, 72*VOL3*sizeof(double));
}
}
free(gauge_field_timeslice);
#else
for(i=0; i<N_ape; i++) {
APE_Smearing_Step(g_gauge_field, alpha_ape);
xchange_gauge_field_timeslice(g_gauge_field);
}
if ( Nlong > 0 ) {
if(g_cart_id==0) fprintf(stdout, "\n# [hdisc] fuzzing gauge field ...\n");
fuzzed_links2(gauge_field_f, g_gauge_field, Nlong);
} else {
memcpy(gauge_field_f, g_gauge_field, 72*VOLUMEPLUSRAND*sizeof(double));
}
xchange_gauge_field(gauge_field_f);
read_lime_gauge_field_doubleprec(filename);
xchange_gauge();
#endif
/*
for(ix=0; ix<VOLUME; ix++) {
for(mu=0; mu<4; mu++) {
for(i=0; i<9; i++) {
fprintf(stdout, "%6d%3d%3d%25.16e%25.16e%25.16e%25.16e\n", ix, mu, i,
gauge_field_f[_GGI(ix,mu)+2*i], gauge_field_f[_GGI(ix,mu)+2*i+1],
g_gauge_field[_GGI(ix,mu)+2*i], g_gauge_field[_GGI(ix,mu)+2*i+1]);
}}
}
*/
} // of if Nlong > 0
/* allocate memory for the spinor fields */
no_fields = 8;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUME+RAND);
/* allocate memory for the contractions */
/*
#ifdef PARALLELTX
if(g_xs_id==0) {idx = 4 * 4 * K * T_global * 2;}
else {idx = 4 * 4 * K * T * 2;}
#else
if(g_cart_id==0) {idx = 4 * 4 * K * T_global * 2;}
else {idx = 4 * 4 * K* T * 2;}
#endif
*/
disc = (double*)calloc(32*K*T, sizeof(double));
if( disc==(double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(3);
}
buffer = (double*)calloc(32*K*T, sizeof(double));
if( buffer==(double*)NULL ) {
fprintf(stderr, "could not allocate memory for buffer\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(4);
}
buffer2 = (double*)calloc(32*K*T_global, sizeof(double));
if( buffer2==(double*)NULL ) {
fprintf(stderr, "could not allocate memory for buffer2\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(5);
}
if(g_cart_id==0) {
sprintf(filename, "hdisc-ss.k0v4.%.4d", Nconf);
ofs[0] = fopen(filename, "w");
sprintf(filename, "hdisc-sc.k0v4.%.4d", Nconf);
ofs[1] = fopen(filename, "w");
sprintf(filename, "hdisc-cs.k0v4.%.4d", Nconf);
ofs[2] = fopen(filename, "w");
sprintf(filename, "hdisc-cc.k0v4.%.4d", Nconf);
ofs[3] = fopen(filename, "w");
if(ofs[0]==(FILE*)NULL || ofs[1]==(FILE*)NULL || ofs[2]==(FILE*)NULL || ofs[3]==(FILE*)NULL) {
fprintf(stderr, "Error, could not open files for writing.\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(6);
}
}
/*****************************************
* HPE coefficients
*****************************************/
/*
if(format==1) {
*/
addimag = 2*g_kappa*g_musigma/sqrt(1 + 4*g_kappa*g_kappa*(g_musigma*g_musigma-g_mudelta*g_mudelta)) *
LX*LY*LZ*3*4*2.*g_kappa*g_kappa*4;
// addreal = (1.+2*g_kappa*g_mudelta)/sqrt(1 + 4*g_kappa*g_kappa*(g_musigma*g_musigma-g_mudelta*g_mudelta)) *
// LX*LY*LZ*3*4*2.*g_kappa*g_kappa*4;
addreal = (1.- 2*g_kappa*g_mudelta)/sqrt(1 + 4*g_kappa*g_kappa*(g_musigma*g_musigma-g_mudelta*g_mudelta)) *
LX*LY*LZ*3*4*2.*g_kappa*g_kappa*4;
v4norm = 1. / ( 8. * g_kappa * g_kappa );
vvnorm = 1. / ( 8. * g_kappa * g_kappa );
/*
} else {
addimag = 2*g_kappa*g_musigma/sqrt(1 + 4*g_kappa*g_kappa*(g_musigma*g_musigma-g_mudelta*g_mudelta)) *
LX*LY*LZ*3*4*2.*g_kappa*2;
addreal = (1.+2*g_kappa*g_mudelta)/sqrt(1 + 4*g_kappa*g_kappa*(g_musigma*g_musigma-g_mudelta*g_mudelta)) *
LX*LY*LZ*3*4*2.*g_kappa*2;
v4norm = 1. / ( 4. * g_kappa );
vvnorm = 1. / ( 4. * g_kappa );
}
*/
if(g_cart_id==0) fprintf(stdout, "# addimag = %25.16e;\t addreal = %25.16e\n"\
"# v4norm = %25.16e;\t vvnorm = %25.16e\n", addimag, addreal, v4norm, vvnorm);
/******************************************
* start loop on source id.s
******************************************/
count = -1;
for(sid=g_sourceid; sid<=g_sourceid2; sid+=g_sourceid_step) {
for(ix=0; ix<32*K*T; ix++) disc[ix] = 0.;
for(ix=0; ix<32*K*T; ix++) buffer[ix] = 0.;
for(ix=0; ix<32*K*T_global; ix++) buffer2[ix] = 0.;
/* read the new propagator */
sprintf(filename, "%s.%.4d.%.5d.hinverted", filename_prefix, Nconf, sid);
/* sprintf(filename, "%s.%.4d.%.2d.inverted", filename_prefix, Nconf, sid); */
estat = read_lime_spinor(g_spinor_field[2], filename, 0);
if( estat != 0 ) {
fprintf(stderr, "[%2d] Error, could not read from file %s at position 0\n", g_cart_id, filename);
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(7);
}
estat = read_lime_spinor(g_spinor_field[3], filename, 1);
if( estat != 0 ) {
fprintf(stderr, "[%2d] Error, could not read from file %s at position 1\n", g_cart_id, filename);
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(7);
}
count++;
xchange_field(g_spinor_field[2]);
xchange_field(g_spinor_field[3]);
/* calculate the source: apply Q_phi_tbc */
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
Q_h_phi(g_spinor_field[0], g_spinor_field[1], g_spinor_field[2], g_spinor_field[3]);
xchange_field(g_spinor_field[0]);
xchange_field(g_spinor_field[1]);
// print the sources
/*
for(ix=0; ix<VOLUME; ix++) {
for(mu=0; mu<12; mu++) {
fprintf(stdout, "%6d%3d%25.16e%25.16e%25.16e%25.16e\n", ix, mu,
g_spinor_field[0][_GSI(ix)+2*mu], g_spinor_field[0][_GSI(ix)+2*mu+1],
g_spinor_field[1][_GSI(ix)+2*mu], g_spinor_field[1][_GSI(ix)+2*mu+1]);
}
}
*/
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "\n# [hdisc] time for applying Q_tm_h: %e seconds\n", retime-ratime);
/* apply gamma5_BdagH4_gamma5 */
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
gamma5_B_h_dagH4_gamma5(g_spinor_field[4], g_spinor_field[5], g_spinor_field[0], g_spinor_field[1], g_spinor_field[6], g_spinor_field[7]);
/*
for(ix=0; ix<VOLUME; ix++) {
for(mu=0; mu<12; mu++) {
fprintf(stdout, "%6d%3d%25.16e%25.16e%25.16e%25.16e\n", ix, mu,
g_spinor_field[4][_GSI(ix)+2*mu], g_spinor_field[4][_GSI(ix)+2*mu+1],
g_spinor_field[5][_GSI(ix)+2*mu], g_spinor_field[5][_GSI(ix)+2*mu+1]);
}
}
*/
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time for applying noise reduction: %e seconds\n", retime-ratime);
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(Nlong>0) {
if(g_cart_id==0) fprintf(stdout, "# fuzzing propagator with Nlong = %d\n", Nlong);
memcpy((void*)g_spinor_field[6], (void*)g_spinor_field[2], 24*(VOLUME+RAND)*sizeof(double));
/* xchange_field_timeslice(g_spinor_field[6]); */
Fuzz_prop3(gauge_field_f, g_spinor_field[6], g_spinor_field[0], Nlong);
xchange_field_timeslice(g_spinor_field[6]);
memcpy((void*)g_spinor_field[7], (void*)g_spinor_field[3], 24*(VOLUME+RAND)*sizeof(double));
/* xchange_field_timeslice(g_spinor_field[7]); */
Fuzz_prop3(gauge_field_f, g_spinor_field[7], g_spinor_field[1], Nlong);
xchange_field_timeslice(g_spinor_field[7]);
} else {
for(ix=0;ix<VOLUME;ix++) { _fv_eq_zero(g_spinor_field[6]+_GSI(ix)); }
for(ix=0;ix<VOLUME;ix++) { _fv_eq_zero(g_spinor_field[7]+_GSI(ix)); }
}
/*
for(ix=0; ix<VOLUME; ix++) {
for(mu=0; mu<12; mu++) {
fprintf(stdout, "%6d%3d%25.16e%25.16e%25.16e%25.16e\n", ix, mu,
g_spinor_field[6][_GSI(ix)+2*mu], g_spinor_field[6][_GSI(ix)+2*mu+1],
g_spinor_field[7][_GSI(ix)+2*mu], g_spinor_field[7][_GSI(ix)+2*mu+1]);
}
}
*/
// recalculate the sources --- they are changed in Fuzz_prop3
Q_h_phi(g_spinor_field[0], g_spinor_field[1], g_spinor_field[2], g_spinor_field[3]);
xchange_field(g_spinor_field[0]);
xchange_field(g_spinor_field[1]);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time for fuzzing: %e seconds\n", retime-ratime);
/********************************
* add new contractions to disc
********************************/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
for(c=0; c<4; c++)
{
if(c==0) { psi0 = g_spinor_field[2]; psi1 = g_spinor_field[4]; psi2 = g_spinor_field[6]; psi3 = g_spinor_field[0]; }
if(c==1) { psi0 = g_spinor_field[2]; psi1 = g_spinor_field[5]; psi2 = g_spinor_field[6]; psi3 = g_spinor_field[1]; }
if(c==2) { psi0 = g_spinor_field[3]; psi1 = g_spinor_field[4]; psi2 = g_spinor_field[7]; psi3 = g_spinor_field[0]; }
if(c==3) { psi0 = g_spinor_field[3]; psi1 = g_spinor_field[5]; psi2 = g_spinor_field[7]; psi3 = g_spinor_field[1]; }
/*
for(ix=0; ix<VOLUME; ix++) {
for(mu=0; mu<12; mu++) {
fprintf(stdout, "%6d%3d%16.7e%16.7e%16.7e%16.7e%16.7e%16.7e%16.7e%16.7e\n", ix, mu,
psi0[_GSI(ix)+2*mu], psi0[_GSI(ix)+2*mu+1], psi1[_GSI(ix)+2*mu], psi1[_GSI(ix)+2*mu+1],
psi2[_GSI(ix)+2*mu], psi2[_GSI(ix)+2*mu+1], psi3[_GSI(ix)+2*mu], psi3[_GSI(ix)+2*mu+1]);
}
}
*/
for(x0=0; x0<T; x0++) {
for(mu=0; mu<16; mu++) {
index_min = x0 * K + mu + c * 4 * K * T;
for(x1=0; x1<VOL3; x1++) {
ix = x0*VOL3 + x1;
idx = _GSI( ix );
_fv_eq_gamma_ti_fv(spinor1, mu, psi0+idx);
_co_eq_fv_dag_ti_fv(&w, psi1+idx, spinor1);
disc[2*( index_min) ] += w.re;
disc[2*( index_min)+1] += w.im;
if(Nlong>0) {
_fv_eq_gamma_ti_fv(spinor1, mu, psi2+idx);
_co_eq_fv_dag_ti_fv(&w, psi1+idx, spinor1);
disc[2*( K*T + index_min) ] += w.re;
disc[2*( K*T + index_min)+1] += w.im;
}
_fv_eq_gamma_ti_fv(spinor1, mu, psi0+idx);
_co_eq_fv_dag_ti_fv(&w, psi3+idx, spinor1);
disc[2*(2*K*T + index_min) ] += w.re;
disc[2*(2*K*T + index_min)+1] += w.im;
if(Nlong>0) {
_fv_eq_gamma_ti_fv(spinor1, mu, psi2+idx);
_co_eq_fv_dag_ti_fv(&w, psi3+idx, spinor1);
disc[2*(3*K*T + index_min) ] += w.re;
disc[2*(3*K*T + index_min)+1] += w.im;
}
}
}
} // of loop on x0
for(x0=0; x0<T; x0++) {
disc[2*( x0*K+4 + 4*c*K*T) ] += addreal;
disc[2*( x0*K+5 + 4*c*K*T)+1] -= addimag;
if(Nlong>0) {
disc[2*(K*T + x0*K+4 + 4*c*K*T) ] += addreal;
disc[2*(K*T + x0*K+5 + 4*c*K*T)+1] -= addimag;
}
}
} // of c=0,...,4
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "# time for contracting: %e seconds\n", retime-ratime);
#ifdef MPI
/* collect results to disc */
#if (defined PARALLELTX) || (defined PARALLELTXY)
MPI_Allreduce(disc, buffer, 32*K*T, MPI_DOUBLE, MPI_SUM, g_ts_comm);
MPI_Allgather(buffer, 32*K*T, MPI_DOUBLE, buffer2, 32*K*T, MPI_DOUBLE, g_xs_comm);
# else
MPI_Gather(disc, 32*K*T, MPI_DOUBLE, buffer2, 32*K*T, MPI_DOUBLE, 0, g_cart_grid);
# endif
#else
memcpy((void*)buffer2, (void*)disc, 32*K*T_global*sizeof(double));
#endif
/* write current disc to file */
if(g_cart_id==0) {
for(c=0; c<4; c++) {
if(sid==g_sourceid) fprintf(ofs[c], "#%6d%3d%3d%3d%3d\t%f\t%f\t%f\t%f\n", Nconf, T_global, LX_global, LY_global, LZ,
g_kappa, g_mu, g_musigma, g_mudelta);
for(x0=0; x0<T_global; x0++) {
for(mu=0; mu<16; mu++) {
idx = gindex[mu];
ix = K*(x0%T) + idx + 16*K*T*(x0/T) + c*4*K*T;
fprintf(ofs[c], "%6d%3d%4d%4d%25.16e%25.16e%25.16e%25.16e%25.16e%25.16e%25.16e%25.16e\n",
Nconf, mu, x0, sid,
gsign[mu]*buffer2[2*( ix)]*v4norm, gsign[mu]*buffer2[2*( ix)+1]*v4norm,
gsign[mu]*buffer2[2*( K*T+ix)]*v4norm, gsign[mu]*buffer2[2*( K*T+ix)+1]*v4norm,
gsign[mu]*buffer2[2*(2*K*T+ix)]*vvnorm, gsign[mu]*buffer2[2*(2*K*T+ix)+1]*vvnorm,
gsign[mu]*buffer2[2*(3*K*T+ix)]*vvnorm, gsign[mu]*buffer2[2*(3*K*T+ix)+1]*vvnorm);
}
}
}
}
if(g_cart_id==0) fprintf(stdout, "# finished all sid %d\n", sid);
} /* of loop on sid */
if(g_cart_id==0) { fclose(ofs[0]); fclose(ofs[1]); fclose(ofs[2]); fclose(ofs[3]); }
/* free the allocated memory, finalize */
free(g_gauge_field);
if(no_fields>0) {
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field);
}
free_geometry();
free(disc);
free(buffer);
free(buffer2);
if(Nlong>0) free(gauge_field_f);
if(g_cart_id == 0) {
g_the_time = time(NULL);
fprintf(stdout, "\n# [disc] %s# [disc] end of run\n", ctime(&g_the_time));
fprintf(stderr, "\n# [disc] %s# [disc] end of run\n", ctime(&g_the_time));
}
#ifdef MPI
MPI_Finalize();
#endif
return(0);
}
int main(int argc, char **argv) {
const int n_c = 3; // number of colors
int c, i, j, mu, nu, ir, is, ia, imunu;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int source_location, have_source_flag = 0;
int x0, x1, x2, x3, ix;
int sx0, sx1, sx2, sx3;
int isimag[4];
int gperm[5][4], gperm2[4][4];
int check_position_space_WI=0;
int num_threads = 1, nthreads=-1, threadid=-1;
int exitstatus;
int write_ascii=0;
int mms = 0, mass_id = -1;
int outfile_prefix_set = 0;
int source_proc_coords[4], source_proc_id = -1;
int ud_single_file = 0;
double gperm_sign[5][4], gperm2_sign[4][4];
double *conn = NULL;
double *conn2 = NULL;
double contact_term[8];
double *work=NULL;
int verbose = 0;
int do_gt = 0, status;
char filename[100], contype[400], outfile_prefix[400];
double ratime, retime;
double plaq;
double spinor1[24], spinor2[24], U_[18];
double *gauge_trafo=(double*)NULL;
double *phi=NULL, *chi=NULL;
complex w;
double Usourcebuff[72], *Usource[4];
FILE *ofs;
#ifdef MPI
int *status;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "swah?vgf:t:m:o:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'g':
do_gt = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'w':
check_position_space_WI = 1;
fprintf(stdout, "\n# [avc_exact2_lowmem_xspace] will check Ward identity in position space\n");
break;
case 't':
num_threads = atoi(optarg);
fprintf(stdout, "\n# [avc_exact2_lowmem_xspace] will use %d threads in spacetime loops\n", num_threads);
break;
case 'a':
write_ascii = 1;
fprintf(stdout, "\n# [avc_exact2_lowmem_xspace] will write data in ASCII format too\n");
break;
case 'm':
mms = 1;
mass_id = atoi(optarg);
fprintf(stdout, "\n# [avc_exact2_lowmem_xspace] will read propagators in MMS format with mass id %d\n", mass_id);
break;
case 'o':
strcpy(outfile_prefix, optarg);
fprintf(stdout, "\n# [avc_exact2_lowmem_xspace] will use prefix %s for output filenames\n", outfile_prefix);
outfile_prefix_set = 1;
break;
case 's':
ud_single_file = 1;
fprintf(stdout, "\n# [avc_exact2_lowmem_xspace] will read up and down propagator from same file\n");
break;
case 'h':
case '?':
default:
usage();
break;
}
}
if(g_cart_id==0) {
g_the_time = time(NULL);
fprintf(stdout, "\n# [avc_exact2_lowmem_xspace] using global time stamp %s", ctime(&g_the_time));
}
/*********************************
* set number of openmp threads
*********************************/
#ifdef OPENMP
omp_set_num_threads(num_threads);
#endif
/* set the default values */
if(filename_set==0) strcpy(filename, "cvc.input");
fprintf(stdout, "# Reading input from file %s\n", filename);
read_input_parser(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stderr, "\n[avc_exact2_lowmem_xspace] T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stderr, "\n[avc_exact2_lowmem_xspace] kappa should be > 0.n");
usage();
}
/* initialize MPI parameters */
mpi_init(argc, argv);
#ifdef MPI
if((status = (int*)calloc(g_nproc, sizeof(int))) == (int*)NULL) {
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(7);
}
#endif
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
#ifndef MPI
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
#endif
fprintf(stdout, "# [%2d] parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(2);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(1);
}
geometry();
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
if(!(strcmp(gaugefilename_prefix,"identity")==0)) {
/* read the gauge field */
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
} else {
/* initialize unit matrices */
if(g_cart_id==0) fprintf(stdout, "\n# [avc_exact] initializing unit matrices\n");
for(ix=0;ix<VOLUME;ix++) {
_cm_eq_id( g_gauge_field + _GGI(ix, 0) );
_cm_eq_id( g_gauge_field + _GGI(ix, 1) );
_cm_eq_id( g_gauge_field + _GGI(ix, 2) );
_cm_eq_id( g_gauge_field + _GGI(ix, 3) );
}
}
#ifdef MPI
xchange_gauge();
#endif
/* measure the plaquette */
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value: %25.16e\n", plaq);
/*
sprintf(filename, "gauge.%.2d", g_cart_id);
ofs = fopen(filename, "w");
for(x0=0;x0<T;x0++) {
for(x1=0;x1<LX;x1++) {
for(x2=0;x2<LY;x2++) {
for(x3=0;x3<LZ;x3++) {
ix = g_ipt[x0][x1][x2][x3];
for(mu=0;mu<4;mu++) {
for(i=0;i<9;i++) {
fprintf(ofs, "%8d%3d%3d%3d%3d%3d%3d%25.16e%25.16e\n", ix, x0+Tstart, x1+LXstart, x2+LYstart, x3, mu, i, g_gauge_field[_GGI(ix,mu)+2*i], g_gauge_field[_GGI(ix,mu)+2*i+1]);
}
}
}}}}
fclose(ofs);
if(g_cart_id==0) fprintf(stdout, "\nWarning: forced exit\n");
fflush(stdout);
fflush(stderr);
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 255);
MPI_Finalize();
#endif
exit(255);
*/
/* allocate memory for the spinor fields */
no_fields = 2;
if(mms) no_fields++;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
if(mms) {
work = g_spinor_field[no_fields-1];
}
/* allocate memory for the contractions */
conn = (double*)calloc(2 * 16 * VOLUME, sizeof(double));
if( conn==(double*)NULL ) {
fprintf(stderr, "could not allocate memory for contr. fields\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 3);
MPI_Finalize();
#endif
exit(3);
}
#ifdef OPENMP
#pragma omp parallel for
#endif
for(ix=0; ix<32*VOLUME; ix++) conn[ix] = 0.;
conn2 = (double*)calloc(2 * 16 * VOLUME, sizeof(double));
if( conn2 == NULL ) {
fprintf(stderr, "could not allocate memory for contr. fields\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 3);
MPI_Finalize();
#endif
exit(3);
}
#ifdef OPENMP
#pragma omp parallel for
#endif
for(ix=0; ix<32*VOLUME; ix++) conn2[ix] = 0.;
/***********************************************************
* determine source coordinates, find out, if source_location is in this process
***********************************************************/
#if (defined PARALLELTX) || (defined PARALLELTXY)
sx0 = g_source_location / (LX_global*LY_global*LZ);
sx1 = (g_source_location%(LX_global*LY_global*LZ)) / (LY_global*LZ);
sx2 = (g_source_location%(LY_global*LZ)) / LZ;
sx3 = (g_source_location%LZ);
source_proc_coords[0] = sx0 / T;
source_proc_coords[1] = sx1 / LX;
source_proc_coords[2] = sx2 / LY;
source_proc_coords[3] = 0;
MPI_Cart_rank(g_cart_grid, source_proc_coords, &source_proc_id);
have_source_flag = (int)(g_cart_id == source_proc_id);
if(have_source_flag==1) {
fprintf(stdout, "\n# process %2d has source location\n", source_proc_id);
fprintf(stdout, "\n# global source coordinates: (%3d,%3d,%3d,%3d)\n", sx0, sx1, sx2, sx3);
fprintf(stdout, "\n# source proc coordinates: (%3d,%3d,%3d,%3d)\n", source_proc_coords[0],
source_proc_coords[1], source_proc_coords[2], source_proc_coords[3]);
}
sx0 = sx0 % T;
sx1 = sx1 % LX;
sx2 = sx2 % LY;
sx3 = sx3 % LZ;
# else
have_source_flag = (int)(g_source_location/(LX*LY*LZ)>=Tstart && g_source_location/(LX*LY*LZ)<(Tstart+T));
if(have_source_flag==1) fprintf(stdout, "process %2d has source location\n", g_cart_id);
sx0 = g_source_location/(LX*LY*LZ)-Tstart;
sx1 = (g_source_location%(LX*LY*LZ)) / (LY*LZ);
sx2 = (g_source_location%(LY*LZ)) / LZ;
sx3 = (g_source_location%LZ);
#endif
if(have_source_flag==1) {
fprintf(stdout, "local source coordinates: (%3d,%3d,%3d,%3d)\n", sx0, sx1, sx2, sx3);
source_location = g_ipt[sx0][sx1][sx2][sx3];
}
#ifdef MPI
# if (defined PARALLELTX) || (defined PARALLELTXY)
have_source_flag = source_proc_id;
MPI_Bcast(Usourcebuff, 72, MPI_DOUBLE, have_source_flag, g_cart_grid);
# else
MPI_Gather(&have_source_flag, 1, MPI_INT, status, 1, MPI_INT, 0, g_cart_grid);
if(g_cart_id==0) {
for(mu=0; mu<g_nproc; mu++) fprintf(stdout, "status[%1d]=%d\n", mu,status[mu]);
}
if(g_cart_id==0) {
for(have_source_flag=0; status[have_source_flag]!=1; have_source_flag++);
fprintf(stdout, "have_source_flag= %d\n", have_source_flag);
}
MPI_Bcast(&have_source_flag, 1, MPI_INT, 0, g_cart_grid);
# endif
fprintf(stdout, "[%2d] have_source_flag = %d\n", g_cart_id, have_source_flag);
#else
have_source_flag = 0;
#endif
/*
if(g_cart_id==0) fprintf(stdout, "\nWarning: forced exit\n");
fflush(stdout);
fflush(stderr);
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 255);
MPI_Finalize();
#endif
exit(255);
*/
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
/***********************************************************
* initialize the Gamma matrices
***********************************************************/
// gamma_5:
gperm[4][0] = gamma_permutation[5][ 0] / 6;
gperm[4][1] = gamma_permutation[5][ 6] / 6;
gperm[4][2] = gamma_permutation[5][12] / 6;
gperm[4][3] = gamma_permutation[5][18] / 6;
gperm_sign[4][0] = gamma_sign[5][ 0];
gperm_sign[4][1] = gamma_sign[5][ 6];
gperm_sign[4][2] = gamma_sign[5][12];
gperm_sign[4][3] = gamma_sign[5][18];
// gamma_nu gamma_5
for(nu=0;nu<4;nu++) {
// permutation
gperm[nu][0] = gamma_permutation[6+nu][ 0] / 6;
gperm[nu][1] = gamma_permutation[6+nu][ 6] / 6;
gperm[nu][2] = gamma_permutation[6+nu][12] / 6;
gperm[nu][3] = gamma_permutation[6+nu][18] / 6;
// is imaginary ?
isimag[nu] = gamma_permutation[6+nu][0] % 2;
// (overall) sign
gperm_sign[nu][0] = gamma_sign[6+nu][ 0];
gperm_sign[nu][1] = gamma_sign[6+nu][ 6];
gperm_sign[nu][2] = gamma_sign[6+nu][12];
gperm_sign[nu][3] = gamma_sign[6+nu][18];
// write to stdout
if(g_cart_id == 0) {
fprintf(stdout, "# gamma_%d5 = (%f %d, %f %d, %f %d, %f %d)\n", nu,
gperm_sign[nu][0], gperm[nu][0], gperm_sign[nu][1], gperm[nu][1],
gperm_sign[nu][2], gperm[nu][2], gperm_sign[nu][3], gperm[nu][3]);
}
}
// gamma_nu
for(nu=0;nu<4;nu++) {
// permutation
gperm2[nu][0] = gamma_permutation[nu][ 0] / 6;
gperm2[nu][1] = gamma_permutation[nu][ 6] / 6;
gperm2[nu][2] = gamma_permutation[nu][12] / 6;
gperm2[nu][3] = gamma_permutation[nu][18] / 6;
// (overall) sign
gperm2_sign[nu][0] = gamma_sign[nu][ 0];
gperm2_sign[nu][1] = gamma_sign[nu][ 6];
gperm2_sign[nu][2] = gamma_sign[nu][12];
gperm2_sign[nu][3] = gamma_sign[nu][18];
// write to stdout
if(g_cart_id == 0) {
fprintf(stdout, "# gamma_%d = (%f %d, %f %d, %f %d, %f %d)\n", nu,
gperm2_sign[nu][0], gperm2[nu][0], gperm2_sign[nu][1], gperm2[nu][1],
gperm2_sign[nu][2], gperm2[nu][2], gperm2_sign[nu][3], gperm2[nu][3]);
}
}
/**********************************************************
**********************************************************
**
** first contribution
**
**********************************************************
**********************************************************/
/**********************************************
* loop on the Lorentz index nu at source
**********************************************/
for(ia=0; ia<n_c; ia++) {
for(nu=0; nu<4; nu++)
//for(nu=0; nu<4; nu++)
{
// fprintf(stdout, "\n# [avc_exact2_lowmem_xspace] 1st part, processing nu = %d ...\n", nu);
for(ir=0; ir<4; ir++) {
// read 1 up-type propagator color components for spinor index ir
if(!mms) {
get_filename(filename, 0, 3*ir+ia, 1);
exitstatus = read_lime_spinor(g_spinor_field[0], filename, 0);
if(exitstatus != 0) {
fprintf(stderr, "\n# [avc_exact2_lowmem_xspace] Error from read_lime_spinor\n");
exit(111);
}
xchange_field(g_spinor_field[0]);
} else {
sprintf(filename, "%s.%.4d.00.%.2d.cgmms.%.2d.inverted", filename_prefix, Nconf, 3*ir+ia, mass_id);
exitstatus = read_lime_spinor(work, filename, 0);
if(exitstatus != 0) {
fprintf(stderr, "\n# [avc_exact2_lowmem_xspace] Error from read_lime_spinor\n");
exit(111);
}
xchange_field(work);
Qf5(g_spinor_field[0], work, -g_mu);
xchange_field(g_spinor_field[0]);
}
// read 1 dn-type propagator color components for spinor index gamma_perm ( ir )
if(!mms) {
if(ud_single_file) {
get_filename(filename, 0, 3*gperm[nu][ir]+ia, 1);
exitstatus = read_lime_spinor(g_spinor_field[1], filename, 1);
} else {
get_filename(filename, 0, 3*gperm[nu][ir]+ia, -1);
exitstatus = read_lime_spinor(g_spinor_field[1], filename, 0);
}
if(exitstatus != 0) {
fprintf(stderr, "\n# [avc_exact2_lowmem_xspace] Error from read_lime_spinor\n");
exit(111);
}
xchange_field(g_spinor_field[1]);
} else {
sprintf(filename, "%s.%.4d.%.2d.%.2d.cgmms.%.2d.inverted", filename_prefix, Nconf, 4, 3*gperm[nu][ir]+ia, mass_id);
exitstatus = read_lime_spinor(work, filename, 0);
if(exitstatus != 0) {
fprintf(stderr, "\n# [avc_exact2_lowmem_xspace] Error from read_lime_spinor\n");
exit(111);
}
xchange_field(work);
Qf5(g_spinor_field[1], work, g_mu);
xchange_field(g_spinor_field[1]);
}
phi = g_spinor_field[0];
chi = g_spinor_field[1];
//fprintf(stdout, "\n# [nu5] spin index pair (%d, %d); col index %d\n", ir, gperm[nu][ir], ia);
// 1) gamma_nu gamma_5 x U
for(mu=0; mu<4; mu++)
//for(mu=0; mu<1; mu++)
{
imunu = 4*mu+nu;
#ifdef OPENMP
#pragma omp parallel for private(ix, spinor1, spinor2, U_, w) shared(imunu, ia, nu, mu)
#endif
for(ix=0; ix<VOLUME; ix++) {
/*
threadid = omp_get_thread_num();
nthreads = omp_get_num_threads();
fprintf(stdout, "[thread%d] number of threads = %d\n", threadid, nthreads);
*/
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix,mu)], &co_phase_up[mu]);
_fv_eq_cm_ti_fv(spinor1, U_, phi+_GSI(g_iup[ix][mu]));
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_fv_eq_gamma_ti_fv(spinor1, 5, spinor2);
_co_eq_fv_dag_ti_fv(&w, chi+_GSI(ix), spinor1);
if(!isimag[nu]) {
conn[_GWI(imunu,ix,VOLUME) ] += gperm_sign[nu][ir] * w.re;
conn[_GWI(imunu,ix,VOLUME)+1] += gperm_sign[nu][ir] * w.im;
} else {
conn[_GWI(imunu,ix,VOLUME) ] += gperm_sign[nu][ir] * w.im;
conn[_GWI(imunu,ix,VOLUME)+1] -= gperm_sign[nu][ir] * w.re;
}
} // of ix
#ifdef OPENMP
#pragma omp parallel for private(ix, spinor1, spinor2, U_, w) shared(imunu, ia, nu, mu)
#endif
for(ix=0; ix<VOLUME; ix++) {
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix,mu)], &co_phase_up[mu]);
_fv_eq_cm_dag_ti_fv(spinor1, U_, phi+_GSI(ix));
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_fv_eq_gamma_ti_fv(spinor1, 5, spinor2);
_co_eq_fv_dag_ti_fv(&w, chi+_GSI(g_iup[ix][mu]), spinor1);
if(!isimag[nu]) {
conn[_GWI(imunu,ix,VOLUME) ] += gperm_sign[nu][ir] * w.re;
conn[_GWI(imunu,ix,VOLUME)+1] += gperm_sign[nu][ir] * w.im;
} else {
conn[_GWI(imunu,ix,VOLUME) ] += gperm_sign[nu][ir] * w.im;
conn[_GWI(imunu,ix,VOLUME)+1] -= gperm_sign[nu][ir] * w.re;
}
} // of ix
// contribution to local-local correlator
#ifdef OPENMP
#pragma omp parallel for private(ix, spinor1, spinor2, U_, w) shared(imunu, ia, nu, mu)
#endif
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_gamma_ti_fv(spinor2, mu, phi+_GSI(ix) );
_fv_eq_gamma_ti_fv(spinor1, 5, spinor2);
_co_eq_fv_dag_ti_fv(&w, chi+_GSI(ix), spinor1);
if(!isimag[nu]) {
conn2[_GWI(imunu,ix,VOLUME) ] += gperm_sign[nu][ir] * w.re;
conn2[_GWI(imunu,ix,VOLUME)+1] += gperm_sign[nu][ir] * w.im;
} else {
conn2[_GWI(imunu,ix,VOLUME) ] += gperm_sign[nu][ir] * w.im;
conn2[_GWI(imunu,ix,VOLUME)+1] -= gperm_sign[nu][ir] * w.re;
}
} // of ix
} // of mu
} // of ir
} // of nu
} // of ia loop on colors
// normalisation of contractions
#ifdef OPENMP
#pragma omp parallel for
#endif
for(ix=0; ix<32*VOLUME; ix++) conn[ix] *= -0.5;
#ifdef OPENMP
#pragma omp parallel for
#endif
for(ix=0; ix<32*VOLUME; ix++) conn2[ix] *= -1.;
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "contractions in %e seconds\n", retime-ratime);
// save results
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(outfile_prefix_set) {
sprintf(filename, "%s/cvc_lvc_x.%.4d.t%.2dx%.2dy%.2dz%.2d", outfile_prefix, Nconf, sx0, sx1, sx2, sx3);
} else {
sprintf(filename, "cvc_lvc_x.%.4d.t%.2dx%.2dy%.2dz%.2d", Nconf, sx0, sx1, sx2, sx3);
}
sprintf(contype, "cvc - lvc in position space, all 16 components");
status = write_lime_contraction(conn, filename, 64, 16, contype, Nconf, 0);
if(status != 0) {
fprintf(stderr, "[] Error from write_lime_contractions, status was %d\n", status);
exit(16);
}
if(outfile_prefix_set) {
sprintf(filename, "%s/lvc_lvc_x.%.4d.t%.2dx%.2dy%.2dz%.2d", outfile_prefix, Nconf, sx0, sx1, sx2, sx3);
} else {
sprintf(filename, "lvc_lvc_x.%.4d.t%.2dx%.2dy%.2dz%.2d", Nconf, sx0, sx1, sx2, sx3);
}
sprintf(contype, "lvc - lvc in position space, all 16 components");
status = write_lime_contraction(conn2, filename, 64, 16, contype, Nconf, 0);
if(status != 0) {
fprintf(stderr, "[] Error from write_lime_contractions, status was %d\n", status);
exit(17);
}
#ifndef MPI
if(write_ascii) {
if(outfile_prefix_set) {
sprintf(filename, "%s/cvc_lvc_x.%.4d.ascii", outfile_prefix, Nconf);
} else {
sprintf(filename, "cvc_lvc_x.%.4d.ascii", Nconf);
}
write_contraction(conn, NULL, filename, 16, 2, 0);
if(outfile_prefix_set) {
sprintf(filename, "%s/lvc_lvc_x.%.4d.ascii", outfile_prefix, Nconf);
} else {
sprintf(filename, "lvc_lvc_x.%.4d.ascii", Nconf);
}
write_contraction(conn2, NULL, filename, 16, 2, 0);
}
#endif
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id==0) fprintf(stdout, "saved position space results in %e seconds\n", retime-ratime);
#ifndef MPI
// check the Ward identity in position space
if(check_position_space_WI) {
sprintf(filename, "WI_X.%.4d", Nconf);
ofs = fopen(filename,"w");
fprintf(stdout, "\n# [avc_exact2_lowmem_xspace] checking Ward identity in position space ...\n");
for(x0=0; x0<T; x0++) {
for(x1=0; x1<LX; x1++) {
for(x2=0; x2<LY; x2++) {
for(x3=0; x3<LZ; x3++) {
fprintf(ofs, "# t=%2d x=%2d y=%2d z=%2d\n", x0, x1, x2, x3);
ix=g_ipt[x0][x1][x2][x3];
for(nu=0; nu<4; nu++) {
w.re = conn[_GWI(4*0+nu,ix,VOLUME)] + conn[_GWI(4*1+nu,ix,VOLUME)]
+ conn[_GWI(4*2+nu,ix,VOLUME)] + conn[_GWI(4*3+nu,ix,VOLUME)]
- conn[_GWI(4*0+nu,g_idn[ix][0],VOLUME)] - conn[_GWI(4*1+nu,g_idn[ix][1],VOLUME)]
- conn[_GWI(4*2+nu,g_idn[ix][2],VOLUME)] - conn[_GWI(4*3+nu,g_idn[ix][3],VOLUME)];
w.im = conn[_GWI(4*0+nu,ix,VOLUME)+1] + conn[_GWI(4*1+nu,ix,VOLUME)+1]
+ conn[_GWI(4*2+nu,ix,VOLUME)+1] + conn[_GWI(4*3+nu,ix,VOLUME)+1]
- conn[_GWI(4*0+nu,g_idn[ix][0],VOLUME)+1] - conn[_GWI(4*1+nu,g_idn[ix][1],VOLUME)+1]
- conn[_GWI(4*2+nu,g_idn[ix][2],VOLUME)+1] - conn[_GWI(4*3+nu,g_idn[ix][3],VOLUME)+1];
fprintf(ofs, "\t%3d%25.16e%25.16e\n", nu, w.re, w.im);
}
}}}}
fclose(ofs);
}
#endif
/****************************************
* free the allocated memory, finalize
****************************************/
free(g_gauge_field);
for(i=0; i<no_fields; i++) free(g_spinor_field[i]);
free(g_spinor_field);
free_geometry();
if(conn != NULL) free(conn);
if(conn2 != NULL) free(conn2);
#ifdef MPI
free(status);
MPI_Finalize();
#endif
if(g_cart_id==0) {
g_the_time = time(NULL);
fprintf(stdout, "\n# [cvc_lvc_exact2_lowmem_xspace] %s# [cvc_lvc_exact2_lowmem_xspace] end of run\n", ctime(&g_the_time));
fprintf(stderr, "\n# [cvc_lvc_exact2_lowmem_xspace] %s# [cvc_lvc_exact2_lowmem_xspace] end of run\n", ctime(&g_the_time));
}
return(0);
}
void gamma5_B_h_dagH4_gamma5 (double *xi_c, double *xi_s, double *phi_c, double *phi_s, double *work1, double *work2) {
int ix, i;
double spinor1[24];
double sign = 1.;
// multiply original source (phi) with gamma_5, save in xi
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_gamma_ti_fv(xi_c + _GSI(ix), 5, phi_c + _GSI(ix) );
}
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_gamma_ti_fv(xi_s + _GSI(ix), 5, phi_s + _GSI(ix) );
}
/************************************************************
* apply B^dagger H four times
* - status: source = xi from last step, dest = work
* - NOTE: sign = +1 for B^dagger
************************************************************/
// 1st application
xchange_field(xi_c);
Hopping(work2, xi_c);
xchange_field(xi_s);
Hopping(work1, xi_s);
B_h_phi(xi_c, xi_s, work2, work1, sign);
// 2nd application
xchange_field(xi_c);
Hopping(work1, xi_c);
xchange_field(xi_s);
Hopping(work2, xi_s);
B_h_phi(xi_c, xi_s, work1, work2, sign);
// 3rd application
xchange_field(xi_c);
Hopping(work1, xi_c);
xchange_field(xi_s);
Hopping(work2, xi_s);
B_h_phi(xi_c, xi_s, work1, work2, sign);
// 4th application
xchange_field(xi_c);
Hopping(work1, xi_c);
xchange_field(xi_s);
Hopping(work2, xi_s);
B_h_phi(xi_c, xi_s, work1, work2, sign);
/* final step: multiply with gamma_5 and */
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_gamma_ti_fv(spinor1, 5, xi_c + _GSI(ix));
_fv_eq_fv(xi_c + _GSI(ix), spinor1);
}
for(ix=0; ix<VOLUME; ix++) {
_fv_eq_gamma_ti_fv(spinor1, 5, xi_s + _GSI(ix));
_fv_eq_fv(xi_s + _GSI(ix), spinor1);
}
}
int main(int argc, char **argv) {
int c, i, mu, nu;
int count = 0;
int filename_set = 0;
int dims[4] = {0,0,0,0};
int l_LX_at, l_LXstart_at;
int x0, x1, x2, x3, ix;
int sid;
double *disc = (double*)NULL;
double *work = (double*)NULL;
double *disc_diag = (double*)NULL;
double phase[4];
int verbose = 0;
int do_gt = 0;
char filename[100];
double ratime, retime;
double plaq;
double spinor1[24], spinor2[24], U_[18];
complex w, w1, psi1[4], psi2[4];
FILE *ofs;
fftw_complex *in=(fftw_complex*)NULL;
#ifdef MPI
fftwnd_mpi_plan plan_p, plan_m;
int *status;
#else
fftwnd_plan plan_p, plan_m;
#endif
#ifdef MPI
MPI_Init(&argc, &argv);
#endif
while ((c = getopt(argc, argv, "h?vgf:")) != -1) {
switch (c) {
case 'v':
verbose = 1;
break;
case 'g':
do_gt = 1;
break;
case 'f':
strcpy(filename, optarg);
filename_set=1;
break;
case 'h':
case '?':
default:
usage();
break;
}
}
/* set the default values */
set_default_input_values();
if(filename_set==0) strcpy(filename, "cvc.input");
/* read the input file */
read_input(filename);
/* some checks on the input data */
if((T_global == 0) || (LX==0) || (LY==0) || (LZ==0)) {
if(g_proc_id==0) fprintf(stdout, "T and L's must be set\n");
usage();
}
if(g_kappa == 0.) {
if(g_proc_id==0) fprintf(stdout, "kappa should be > 0.n");
usage();
}
/* initialize MPI parameters */
mpi_init(argc, argv);
#ifdef MPI
if((status = (int*)calloc(g_nproc, sizeof(int))) == (int*)NULL) {
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(7);
}
#endif
/* initialize fftw */
dims[0]=T_global; dims[1]=LX; dims[2]=LY; dims[3]=LZ;
#ifdef MPI
plan_p = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_BACKWARD, FFTW_MEASURE);
plan_m = fftwnd_mpi_create_plan(g_cart_grid, 4, dims, FFTW_FORWARD, FFTW_MEASURE);
fftwnd_mpi_local_sizes(plan_p, &T, &Tstart, &l_LX_at, &l_LXstart_at, &FFTW_LOC_VOLUME);
#else
plan_p = fftwnd_create_plan(4, dims, FFTW_BACKWARD, FFTW_MEASURE | FFTW_IN_PLACE);
plan_m = fftwnd_create_plan(4, dims, FFTW_FORWARD, FFTW_MEASURE | FFTW_IN_PLACE);
T = T_global;
Tstart = 0;
l_LX_at = LX;
l_LXstart_at = 0;
FFTW_LOC_VOLUME = T*LX*LY*LZ;
#endif
fprintf(stdout, "# [%2d] fftw parameters:\n"\
"# [%2d] T = %3d\n"\
"# [%2d] Tstart = %3d\n"\
"# [%2d] l_LX_at = %3d\n"\
"# [%2d] l_LXstart_at = %3d\n"\
"# [%2d] FFTW_LOC_VOLUME = %3d\n",
g_cart_id, g_cart_id, T, g_cart_id, Tstart, g_cart_id, l_LX_at,
g_cart_id, l_LXstart_at, g_cart_id, FFTW_LOC_VOLUME);
#ifdef MPI
if(T==0) {
fprintf(stderr, "[%2d] local T is zero; exit\n", g_cart_id);
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
exit(2);
}
#endif
if(init_geometry() != 0) {
fprintf(stderr, "ERROR from init_geometry\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(1);
}
geometry();
/* read the gauge field */
alloc_gauge_field(&g_gauge_field, VOLUMEPLUSRAND);
sprintf(filename, "%s.%.4d", gaugefilename_prefix, Nconf);
if(g_cart_id==0) fprintf(stdout, "reading gauge field from file %s\n", filename);
read_lime_gauge_field_doubleprec(filename);
#ifdef MPI
xchange_gauge();
#endif
/* measure the plaquette */
plaquette(&plaq);
if(g_cart_id==0) fprintf(stdout, "measured plaquette value: %25.16e\n", plaq);
/* allocate memory for the spinor fields */
no_fields = 2;
g_spinor_field = (double**)calloc(no_fields, sizeof(double*));
for(i=0; i<no_fields; i++) alloc_spinor_field(&g_spinor_field[i], VOLUMEPLUSRAND);
/* allocate memory for the contractions */
disc = (double*)calloc(8*VOLUME, sizeof(double));
work = (double*)calloc(20*VOLUME, sizeof(double));
if( (disc==(double*)NULL) || (work==(double*)NULL) ) {
fprintf(stderr, "could not allocate memory for disc/work\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(3);
}
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
if(g_subtract == 1) {
/* allocate memory for disc_diag */
disc_diag = (double*)calloc(20*VOLUME, sizeof(double));
if( disc_diag == (double*)NULL ) {
fprintf(stderr, "could not allocate memory for disc_diag\n");
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(8);
}
for(ix=0; ix<20*VOLUME; ix++) disc_diag[ix] = 0.;
}
/* prepare Fourier transformation arrays */
in = (fftw_complex*)malloc(FFTW_LOC_VOLUME*sizeof(fftw_complex));
if(in==(fftw_complex*)NULL) {
#ifdef MPI
MPI_Abort(MPI_COMM_WORLD, 1);
MPI_Finalize();
#endif
exit(4);
}
if(g_resume==1) { /* read current disc from file */
sprintf(filename, ".outcvc_current.%.4d", Nconf);
c = read_contraction(disc, &count, filename, 4);
if( (g_subtract==1) && (c==0) ) {
sprintf(filename, ".outcvc_diag_current.%.4d", Nconf);
c = read_contraction(disc_diag, (int*)NULL, filename, 10);
}
#ifdef MPI
MPI_Gather(&c, 1, MPI_INT, status, 1, MPI_INT, 0, g_cart_grid);
if(g_cart_id==0) {
/* check the entries in status */
for(i=0; i<g_nproc; i++)
if(status[i]!=0) { status[0] = 1; break; }
}
MPI_Bcast(status, 1, MPI_INT, 0, g_cart_grid);
if(status[0]==1) {
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
count = 0;
}
#else
if(c != 0) {
fprintf(stdout, "could not read current disc; start new\n");
for(ix=0; ix<8*VOLUME; ix++) disc[ix] = 0.;
if(g_subtract==1) for(ix=0; ix<20*VOLUME; ix++) disc_diag[ix] = 0.;
count = 0;
}
#endif
if(g_cart_id==0) fprintf(stdout, "starting with count = %d\n", count);
} /* of g_resume == 1 */
/* start loop on source id.s */
for(sid=g_sourceid; sid<=g_sourceid2; sid++) {
/* read the new propagator */
/* sprintf(filename, "%s.%.4d.%.2d", filename_prefix, Nconf, sid); */
sprintf(filename, "source.%.4d.%.2d.inverted", Nconf, sid);
if(format==0) {
if(read_lime_spinor(g_spinor_field[1], filename, 0) != 0) break;
}
else if(format==1) {
if(read_cmi(g_spinor_field[1], filename) != 0) break;
}
count++;
xchange_field(g_spinor_field[1]);
/* calculate the source: apply Q_phi_tbc */
Q_phi_tbc(g_spinor_field[0], g_spinor_field[1]);
xchange_field(g_spinor_field[0]);
/*
sprintf(filename, "%s.source.%.2d", filename, g_cart_id);
ofs = fopen(filename, "w");
printf_spinor_field(g_spinor_field[0], ofs);
fclose(ofs);
*/
/* add new contractions to (existing) disc */
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
for(ix=0; ix<VOLUME; ix++) { /* loop on lattice sites */
for(mu=0; mu<4; mu++) { /* loop on Lorentz index of the current */
_cm_eq_cm_ti_co(U_, &g_gauge_field[_GGI(ix, mu)], &co_phase_up[mu]);
/* first contribution */
_fv_eq_cm_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(g_iup[ix][mu])]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_mi_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(ix)], spinor2);
disc[_GJI(ix, mu) ] -= 0.25 * w.re;
disc[_GJI(ix, mu)+1] -= 0.25 * w.im;
if(g_subtract==1) {
work[_GWI(mu,ix,VOLUME) ] = -0.25 * w.re;
work[_GWI(mu,ix,VOLUME)+1] = -0.25 * w.im;
}
/* second contribution */
_fv_eq_cm_dag_ti_fv(spinor1, U_, &g_spinor_field[1][_GSI(ix)]);
_fv_eq_gamma_ti_fv(spinor2, mu, spinor1);
_fv_pl_eq_fv(spinor2, spinor1);
_co_eq_fv_dag_ti_fv(&w, &g_spinor_field[0][_GSI(g_iup[ix][mu])], spinor2);
disc[_GJI(ix, mu) ] -= 0.25 * w.re;
disc[_GJI(ix, mu)+1] -= 0.25 * w.im;
if(g_subtract==1) {
work[_GWI(mu,ix,VOLUME) ] -= 0.25 * w.re;
work[_GWI(mu,ix,VOLUME)+1] -= 0.25 * w.im;
}
}
}
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "[%2d] contractions in %e seconds\n", g_cart_id, retime-ratime);
if(g_subtract==1) {
/* add current contribution to disc_diag */
for(mu=0; mu<4; mu++) {
for(i=0; i<4; i++) phase[i] = (double)(i==mu);
memcpy((void*)in, (void*)&work[_GWI(mu,0,VOLUME)], 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
for(x0=0; x0<T; x0++) {
for(x1=0; x1<LX; x1++) {
for(x2=0; x2<LY; x2++) {
for(x3=0; x3<LZ; x3++) {
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
w.im = -sin( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
_co_eq_co_ti_co(&w1, &in[ix], &w);
work[_GWI(4+mu,ix,VOLUME) ] = w1.re;
work[_GWI(4+mu,ix,VOLUME)+1] = w1.im;
}
}
}
}
memcpy((void*)in, (void*)&work[_GWI(mu,0,VOLUME)], 2*VOLUME*sizeof(double));
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
for(x0=0; x0<T; x0++) {
for(x1=0; x1<LX; x1++) {
for(x2=0; x2<LY; x2++) {
for(x3=0; x3<LZ; x3++) {
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
w.im = sin( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
_co_eq_co_ti_co(&w1, &in[ix], &w);
work[_GWI(mu,ix,VOLUME) ] = w1.re;
work[_GWI(mu,ix,VOLUME)+1] = w1.im;
}
}
}
}
} /* of mu */
for(ix=0; ix<VOLUME; ix++) {
i=-1;
for(mu=0; mu<4; mu++) {
for(nu=mu; nu<4; nu++) {
i++;
_co_eq_co_ti_co(&w, (complex*)&work[_GWI(mu,ix,VOLUME)], (complex*)&work[_GWI(4+nu,ix,VOLUME)]);
disc_diag[_GWI(ix,i,10) ] += w.re;
disc_diag[_GWI(ix,i,10)+1] += w.im;
}
}
}
} /* of g_subtract == 1 */
/* save results for count = multiple of Nsave */
if(count%Nsave == 0) {
#ifdef MPI
ratime = MPI_Wtime();
#else
ratime = (double)clock() / CLOCKS_PER_SEC;
#endif
if(g_cart_id == 0) fprintf(stdout, "save results for count = %d\n", count);
/* save the result in position space */
sprintf(filename, "outcvc_X.%.4d.%.4d", Nconf, count);
write_contraction(disc, NULL, filename, 4, 1, 0);
/* Fourier transform data, copy to work */
for(mu=0; mu<4; mu++) {
for(i=0; i<4; i++) phase[i] = (double)(i==mu);
for(ix=0; ix<VOLUME; ix++) {
in[ix].re = disc[_GJI(ix,mu) ];
in[ix].im = disc[_GJI(ix,mu)+1];
}
#ifdef MPI
fftwnd_mpi(plan_m, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_m, in, NULL);
#endif
for(x0=0; x0<T; x0++) {
for(x1=0; x1<LX; x1++) {
for(x2=0; x2<LY; x2++) {
for(x3=0; x3<LZ; x3++) {
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
w.im = -sin( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
_co_eq_co_ti_co(&w1, &in[ix], &w);
work[_GWI(ix,4+mu,8) ] = w1.re / (double)count;
work[_GWI(ix,4+mu,8)+1] = w1.im / (double)count;
}
}
}
}
for(ix=0; ix<VOLUME; ix++) {
in[ix].re = disc[_GJI(ix, mu) ];
in[ix].im = disc[_GJI(ix, mu)+1];
}
#ifdef MPI
fftwnd_mpi(plan_p, 1, in, NULL, FFTW_NORMAL_ORDER);
#else
fftwnd_one(plan_p, in, NULL);
#endif
for(x0=0; x0<T; x0++) {
for(x1=0; x1<LX; x1++) {
for(x2=0; x2<LY; x2++) {
for(x3=0; x3<LZ; x3++) {
ix = g_ipt[x0][x1][x2][x3];
w.re = cos( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
w.im = sin( M_PI *
(phase[0]*(double)(Tstart+x0)/(double)T_global + phase[1]*(double)x1/(double)LX +
phase[2]*(double)x2/(double)LY + phase[3]*(double)x3/(double)LZ) );
_co_eq_co_ti_co(&w1, &in[ix], &w);
work[_GWI(ix,mu,8) ] = w1.re / (double)count;
work[_GWI(ix,mu,8)+1] = w1.im / (double)count;
}
}
}
}
} /* of mu =0 ,..., 3*/
/* save the result in momentum space */
sprintf(filename, "outcvc_P.%.4d.%.4d", Nconf, count);
write_contraction(work, NULL, filename, 8, 1, 0);
/* calculate the correlations 00, 01, 02, 03, 11, 12, ..., 23, 33 */
for(ix=VOLUME-1; ix>=0; ix--) {
/* copy current data to auxilliary vector */
memcpy((void*)psi1, (void*)&work[_GWI(ix,0,8)], 8*sizeof(double));
memcpy((void*)psi2, (void*)&work[_GWI(ix,4,8)], 8*sizeof(double));
i = -1;
for(mu=0; mu<4; mu++) {
for(nu=mu; nu<4; nu++) {
i++;
_co_eq_co_ti_co(&w,&psi1[mu],&psi2[nu]);
if(g_subtract !=1 ) {
work[_GWI(ix,i,10) ] = w.re / (double)(T_global*LX*LY*LZ);
work[_GWI(ix,i,10)+1] = w.im / (double)(T_global*LX*LY*LZ);
}
else {
work[_GWI(ix,i,10) ] =
( w.re - disc_diag[_GWI(ix,i,10) ]/(double)(count*count) ) /
(double)(T_global*LX*LY*LZ);
work[_GWI(ix,i,10)+1] =
( w.im - disc_diag[_GWI(ix,i,10)+1]/(double)(count*count) ) /
(double)(T_global*LX*LY*LZ);
}
}
}
}
/* save current results to file */
sprintf(filename, "outcvc_final.%.4d.%.4d", Nconf, count);
write_contraction(work, (int*)NULL, filename, 10, 1, 0);
#ifdef MPI
retime = MPI_Wtime();
#else
retime = (double)clock() / CLOCKS_PER_SEC;
#endif
fprintf(stdout, "[%2d] time to save results: %e seconds\n", g_cart_id, retime-ratime);
} /* of count % Nsave == 0 */
} /* of loop on sid */
/* write current disc to file */
sprintf(filename, ".outcvc_current.%.4d", Nconf);
write_contraction(disc, &count, filename, 4, 0, 0);
if(g_subtract == 1) {
/* write current disc_diag to file */
sprintf(filename, ".outcvc_diag_current.%.4d", Nconf);
write_contraction(disc_diag, (int*)NULL, filename, 10, 0, 0);
}
/* free the allocated memory, finalize */
free(g_gauge_field); g_gauge_field=(double*)NULL;
for(i=0; i<no_fields; i++) {
free(g_spinor_field[i]);
g_spinor_field[i] = (double*)NULL;
}
free(g_spinor_field); g_spinor_field=(double**)NULL;
free_geometry();
fftw_free(in);
free(disc);
free(work);
if(g_subtract==1) free(disc_diag);
#ifdef MPI
fftwnd_mpi_destroy_plan(plan_p);
fftwnd_mpi_destroy_plan(plan_m);
free(status);
MPI_Finalize();
#else
fftwnd_destroy_plan(plan_p);
fftwnd_destroy_plan(plan_m);
#endif
return(0);
}
