static void
path_prod(QDP_ColorMatrix *u[], QDP_ColorMatrix *m, int path[], int len,
int subl, QDP_Subset subset[], int (*neighsubl)(int subl, int dir))
{
QDP_ShiftDir fb;
QDP_ColorMatrix *p=NULL, *s=NULL;
QDP_Lattice *lat = QDP_get_lattice_M(m);
int nd = QDP_ndim_L(lat);
int sn = 0;
for(int i=0; i<len; i++) {
int dir = abs(path[i])-1;
// if the path moves in the + dir then we shift from the backward dir
fb = path[i]<0 ? QDP_forward : QDP_backward;
if(fb==QDP_backward) { // path is moving in + dir
if(i==0) {
QDP_M_eq_Ma(tm[sn], u[dir], subset[subl]);
} else {
QDP_M_eq_Ma_times_M(tm[sn], u[dir], p, subset[subl]);
QDP_discard_M(p);
}
subl = neighsubl(subl, path[i]);
s = sm[sn][nd+dir];
QDP_discard_M(s);
QDP_M_eq_sM(s, tm[sn], QDP_neighbor_L(lat)[dir], fb, eosub(subl));
//p = t1; t1 = t2; t2 = p;
sn = 1-sn;
p = s;
} else {
if(i==0) {
subl = neighsubl(subl, path[i]);
QDP_M_eq_M(tm[1-sn], u[dir], subset[subl]);
} else {
QDP_M_eq_M(tm[sn], p, subset[subl]);
QDP_discard_M(p);
subl = neighsubl(subl, path[i]);
s = sm[sn][dir];
QDP_discard_M(s);
QDP_M_eq_sM(s, tm[sn], QDP_neighbor_L(lat)[dir], fb, eosub(subl));
QDP_M_eq_M_times_M(tm[1-sn], u[dir], s, subset[subl]);
QDP_discard_M(s);
}
p = tm[1-sn];
}
}
QDP_M_eq_M(m, p, subset[subl]);
QDP_discard_M(p);
QDP_discard_M(s);
}
/* Computes the staple :
mu
+-------+
nu | |
| |
X X
Where the mu link can be any su3_matrix. The result is saved in staple.
if staple==NULL then the result is not saved.
It also adds the computed staple to the fatlink[mu] with weight coef.
*/
static void
compute_gen_staple(QDP_ColorMatrix *staple, int mu, int nu,
QDP_ColorMatrix *link, double dcoef,
QDP_ColorMatrix *gauge[], QDP_ColorMatrix *fl[])
{
QLA_Real coef = dcoef;
QDP_ColorMatrix *ts0, *ts1;
QDP_ColorMatrix *tmat1, *tmat2;
QDP_ColorMatrix *tempmat;
ts0 = QDP_create_M();
ts1 = QDP_create_M();
tmat1 = QDP_create_M();
tmat2 = QDP_create_M();
tempmat = QDP_create_M();
/* Upper staple */
QDP_M_eq_sM(ts0, link, QDP_neighbor[nu], QDP_forward, QDP_all);
QDP_M_eq_sM(ts1, gauge[nu], QDP_neighbor[mu], QDP_forward, QDP_all);
if(staple!=NULL) { /* Save the staple */
QDP_M_eq_M_times_Ma(tmat1, ts0, ts1, QDP_all);
QDP_M_eq_M_times_M(staple, gauge[nu], tmat1, QDP_all);
} else { /* No need to save the staple. Add it to the fatlinks */
QDP_M_eq_M_times_Ma(tmat1, ts0, ts1, QDP_all);
QDP_M_eq_M_times_M(tmat2, gauge[nu], tmat1, QDP_all);
QDP_M_peq_r_times_M(fl[mu], &coef, tmat2, QDP_all);
}
/* lower staple */
QDP_M_eq_sM(ts0, gauge[nu], QDP_neighbor[mu], QDP_forward, QDP_all);
QDP_M_eq_Ma_times_M(tmat1, gauge[nu], link, QDP_all);
QDP_M_eq_M_times_M(tempmat, tmat1, ts0, QDP_all);
QDP_M_eq_sM(ts0, tempmat, QDP_neighbor[nu], QDP_backward, QDP_all);
if(staple!=NULL) { /* Save the staple */
QDP_M_peq_M(staple, ts0, QDP_all);
QDP_M_peq_r_times_M(fl[mu], &coef, staple, QDP_all);
} else { /* No need to save the staple. Add it to the fatlinks */
QDP_M_peq_r_times_M(fl[mu], &coef, ts0, QDP_all);
}
QDP_destroy_M(ts0);
QDP_destroy_M(ts1);
QDP_destroy_M(tmat1);
QDP_destroy_M(tmat2);
QDP_destroy_M(tempmat);
} /* compute_gen_staple */
/* special case to transport a "connection" by one link, does both parities */
static void
link_transport_connection_qdp( QDP_ColorMatrix *dest, QDP_ColorMatrix *src,
QDP_ColorMatrix *gf[4], QDP_ColorMatrix *work,
QDP_ColorMatrix *st[8], int dir ){
if( GOES_FORWARDS(dir) ) {
QDP_M_eq_M(work, src, QDP_all);
QDP_M_eq_sM(st[dir], work, QDP_neighbor[dir], QDP_forward, QDP_all);
QDP_M_eq_M_times_M(dest, gf[dir], st[dir], QDP_all);
QDP_discard_M(st[dir]);
}
else { /* GOES_BACKWARDS(dir) */
QDP_M_eq_Ma_times_M(work, gf[OPP_DIR(dir)], src, QDP_all);
QDP_M_eq_sM(st[dir], work, QDP_neighbor[OPP_DIR(dir)],
QDP_backward,QDP_all);
QDP_M_eq_M(dest, st[dir], QDP_all);
QDP_discard_M(st[dir]);
}
} /* link_transport_connection_qdp */
// topdir = 1..nd
// sidedir = -nd..nd
// toplinknum,sidelinknum = 0..nin-1
void
QOP_staples(QOP_info_t *info, int nout, int nin,
QDP_ColorMatrix *out[], QDP_ColorMatrix *in[],
int nstaples[], int *topdir[], int *sidedir[],
int *toplinknum[], int *sidelinknum[], QLA_Real *coef[])
{
#define NC QDP_get_nc(in[0])
double dtime = QOP_time();
double nflops = 0;
int nd = QDP_ndim();
QDP_ColorMatrix *ftmps[nin][nd], *t1, *t2, *bt2[nd];
for(int i=0; i<nin; i++)
for(int j=0; j<nd; j++)
ftmps[i][j] = NULL;
for(int i=0; i<nd; i++) bt2[i] = NULL;
t1 = QDP_create_M();
t2 = QDP_create_M();
for(int io=0; io<nout; io++) {
//QOP_printf0("%i: ns: %i\n", io, nstaples[io]);
for(int s=0; s<nstaples[io]; s++) {
QLA_Real c = coef[io][s];
int tn = toplinknum[io][s];
int sdir = sidedir[io][s];
//QOP_printf0(" %i: sdir: %i c: %g\n", s, sdir, c);
if(sdir==0) {
if(c==1) {
QDP_M_peq_M(out[io], in[tn], QDP_all);
nflops += PEQM;
} else {
QDP_M_peq_r_times_M(out[io], &c, in[tn], QDP_all);
nflops += 2*PEQM;
}
} else if(sdir>0) {
int nu = sdir-1;
int mu = topdir[io][s]-1;
int sn = sidelinknum[io][s];
QDP_ColorMatrix *Umunu = getU(tn, mu, nu);
QDP_ColorMatrix *Unumu = getU(sn, nu, mu);
QDP_M_eq_M_times_M(t1, in[sn], Umunu, QDP_all);
if(c==1) {
QDP_M_peq_M_times_Ma(out[io], t1, Unumu, QDP_all);
nflops += EQMTM+PEQMTM;
} else {
QDP_M_eq_M_times_Ma(t2, t1, Unumu, QDP_all);
QDP_M_peq_r_times_M(out[io], &c, t2, QDP_all);
nflops += 2*EQMTM+2*PEQM;
}
} else {
int nu = -sdir-1;
int mu = topdir[io][s]-1;
int sn = sidelinknum[io][s];
QDP_ColorMatrix *Unumu = getU(sn, nu, mu);
QDP_M_eq_M_times_M(t1, in[tn], Unumu, QDP_all);
QDP_M_eq_Ma_times_M(t2, in[sn], t1, QDP_all);
QDP_ColorMatrix *tb = shiftb(t2, nu);
if(c==1) {
QDP_M_peq_M(out[io], tb, QDP_all);
nflops += 2*EQMTM+PEQM;
} else {
QDP_M_peq_r_times_M(out[io], &c, tb, QDP_all);
nflops += 2*EQMTM+2*PEQM;
}
QDP_discard_M(tb);
}
}
}
for(int i=0; i<nin; i++)
for(int j=0; j<nd; j++)
if(ftmps[i][j]!=NULL) QDP_destroy_M(ftmps[i][j]);
for(int i=0; i<nd; i++) if(bt2[i]!=NULL) QDP_destroy_M(bt2[i]);
QDP_destroy_M(t1);
QDP_destroy_M(t2);
info->final_sec = QOP_time() - dtime;
info->final_flop = nflops*QDP_sites_on_node;
info->status = QOP_SUCCESS;
#undef NC
}
// topdir = 1..nd
// sidedir = -nd..nd
// toplinknum,sidelinknum = 0..nin-1
void
QOP_staples_deriv(QOP_info_t *info, int nout, int nin,
QDP_ColorMatrix *deriv[], QDP_ColorMatrix *chain[],
QDP_ColorMatrix *in[],
int nstaples[], int *topdir[], int *sidedir[],
int *toplinknum[], int *sidelinknum[], QLA_Real *coef[])
{
#define NC QDP_get_nc(in[0])
double dtime = QOP_time();
double nflops = 0;
int nd = QDP_ndim();
QDP_ColorMatrix *ftmps[nin][nd], *t1, *t2, *t3, *t4, *tc, *bt2[nd], *bt3[nd], *ctmps[nd];
int ctn[nd];
for(int i=0; i<nin; i++)
for(int j=0; j<nd; j++)
ftmps[i][j] = NULL;
for(int i=0; i<nd; i++) bt2[i] = bt3[i] = ctmps[i] = NULL;
t1 = QDP_create_M();
t2 = QDP_create_M();
t3 = QDP_create_M();
t4 = QDP_create_M();
tc = QDP_create_M();
// process in reverse in case calculated staples used as input for others
for(int io=nout-1; io>=0; io--) {
for(int i=0; i<nd; i++) {
if(ctmps[i]) QDP_discard_M(ctmps[i]);
ctn[i] = 0;
}
QDP_M_eq_M(tc, chain[io], QDP_all);
for(int s=0; s<nstaples[io]; s++) {
QLA_Real c = coef[io][s];
int tn = toplinknum[io][s];
int sdir = sidedir[io][s];
//QOP_printf0("io: %i s: %i sdir: %i tn: %i c: %g\n", io, s, sdir, tn, c);
if(sdir==0) {
if(c==1) {
QDP_M_peq_M(deriv[tn], tc, QDP_all);
nflops += PEQM;
} else {
QDP_M_peq_r_times_M(deriv[tn], &c, tc, QDP_all);
nflops += 2*PEQM;
}
} else if(sdir>0) {
int nu = sdir-1;
int mu = topdir[io][s]-1;
int sn = sidelinknum[io][s];
//QOP_printf0(" mu: %i nu: %i sn: %i\n", mu, nu, sn);
QDP_ColorMatrix *Umunu = getU(tn, mu, nu);
QDP_ColorMatrix *Unumu = getU(sn, nu, mu);
QDP_M_eq_M_times_M(t1, in[sn], Umunu, QDP_all);
QDP_M_eq_Ma_times_M(t2, tc, t1, QDP_all);
QDP_ColorMatrix *tb2 = shiftb(t2, mu);
QDP_M_eq_M_times_M(t1, tc, Unumu, QDP_all);
QDP_M_eq_Ma_times_M(t3, in[sn], t1, QDP_all);
QDP_ColorMatrix *tb3 = shiftb(t3, nu);
if(c==1) {
QDP_M_peq_M_times_Ma(deriv[sn], t1, Umunu, QDP_all);
QDP_M_peq_M(deriv[sn], tb2, QDP_all);
QDP_M_peq_M(deriv[tn], tb3, QDP_all);
nflops += 4*EQMTM+PEQMTM+2*PEQM;
} else {
QDP_M_eq_M_times_Ma(t4, t1, Umunu, QDP_all);
QDP_M_peq_r_times_M(deriv[sn], &c, t4, QDP_all);
QDP_M_peq_r_times_M(deriv[sn], &c, tb2, QDP_all);
QDP_M_peq_r_times_M(deriv[tn], &c, tb3, QDP_all);
nflops += 5*EQMTM+6*PEQM;
}
QDP_discard_M(tb2);
QDP_discard_M(tb3);
} else {
int nu = -sdir-1;
int mu = topdir[io][s]-1;
int sn = sidelinknum[io][s];
QDP_ColorMatrix *Cmunu = getC(nu);
QDP_ColorMatrix *Unumu = getU(sn, nu, mu);
QDP_M_eq_M_times_M(t1, in[sn], Cmunu, QDP_all);
QDP_M_eq_Ma_times_M(t2, in[tn], t1, QDP_all);
QDP_ColorMatrix *tb2 = shiftb(t2, mu);
QDP_M_eq_M_times_M(t3, in[tn], Unumu, QDP_all);
if(c==1) {
QDP_M_peq_M_times_Ma(deriv[tn], t1, Unumu, QDP_all);
QDP_M_peq_M_times_Ma(deriv[sn], t3, Cmunu, QDP_all);
QDP_M_peq_M(deriv[sn], tb2, QDP_all);
nflops += 3*EQMTM+2*PEQMTM+PEQM;
} else {
QDP_M_eq_M_times_Ma(t4, t1, Unumu, QDP_all);
QDP_M_peq_r_times_M(deriv[tn], &c, t4, QDP_all);
QDP_M_eq_M_times_Ma(t4, t3, Cmunu, QDP_all);
QDP_M_peq_r_times_M(deriv[sn], &c, t4, QDP_all);
QDP_M_peq_r_times_M(deriv[sn], &c, tb2, QDP_all);
nflops += 5*EQMTM+6*PEQM;
}
QDP_discard_M(tb2);
}
}
}
for(int i=0; i<nin; i++)
for(int j=0; j<nd; j++)
if(ftmps[i][j]!=NULL) QDP_destroy_M(ftmps[i][j]);
for(int i=0; i<nd; i++) {
if(bt2[i]!=NULL) QDP_destroy_M(bt2[i]);
if(bt3[i]!=NULL) QDP_destroy_M(bt3[i]);
if(ctmps[i]!=NULL) QDP_destroy_M(ctmps[i]);
}
QDP_destroy_M(t1);
QDP_destroy_M(t2);
QDP_destroy_M(t3);
QDP_destroy_M(t4);
QDP_destroy_M(tc);
info->final_sec = QOP_time() - dtime;
info->final_flop = nflops*QDP_sites_on_node;
info->status = QOP_SUCCESS;
#undef NC
}
void
QOPPC(symanzik_1loop_gauge_force1) (QOP_info_t *info, QOP_GaugeField *gauge,
QOP_Force *force, QOP_gauge_coeffs_t *coeffs, REAL eps)
{
REAL Plaq, Rect, Pgm ;
QDP_ColorMatrix *tempmom_qdp[4];
QDP_ColorMatrix *Amu[6]; // products of 2 links Unu(x)*Umu(x+nu)
QDP_ColorMatrix *tmpmat;
QDP_ColorMatrix *tmpmat1;
QDP_ColorMatrix *tmpmat2;
QDP_ColorMatrix *staples;
QDP_ColorMatrix *tmpmat3;
QDP_ColorMatrix *tmpmat4;
int i, k;
int mu, nu, sig;
double dtime;
//REAL eb3 = -eps*beta/3.0;
REAL eb3 = -eps/3.0;
int j[3][2] = {{1,2},
{0,2},
{0,1}};
// QOP_printf0("beta: %e, eb3: %e\n", beta, eb3);
dtime = -QOP_time();
for(mu=0; mu<4; mu++) {
tempmom_qdp[mu] = QDP_create_M();
QDP_M_eq_zero(tempmom_qdp[mu], QDP_all);
}
tmpmat = QDP_create_M();
for(i=0; i<QOP_common.ndim; i++) {
fblink[i] = gauge->links[i];
fblink[OPP_DIR(i)] = QDP_create_M();
QDP_M_eq_sM(tmpmat, fblink[i], QDP_neighbor[i], QDP_backward, QDP_all);
QDP_M_eq_Ma(fblink[OPP_DIR(i)], tmpmat, QDP_all);
}
for(i=0; i<6; i++) {
Amu[i] = QDP_create_M();
}
staples = QDP_create_M();
tmpmat1 = QDP_create_M();
tmpmat2 = QDP_create_M();
tmpmat3 = QDP_create_M();
tmpmat4 = QDP_create_M();
Plaq = coeffs->plaquette;
Rect = coeffs->rectangle;
Pgm = coeffs->parallelogram;
//Construct 3-staples and rectangles
for(mu=0; mu<4; mu++) {
i=0;
for(nu=0; nu<4; nu++) {
if(nu!=mu){
// tmpmat1 = Umu(x+nu)
QDP_M_eq_sM(tmpmat1, fblink[mu], QDP_neighbor[nu], QDP_forward, QDP_all);
QDP_M_eq_M_times_M(Amu[i], fblink[nu], tmpmat1, QDP_all);
//tmpmat2 = Umu(x-nu)
QDP_M_eq_sM(tmpmat2, fblink[mu], QDP_neighbor[nu], QDP_backward, QDP_all);
QDP_M_eq_M_times_M(Amu[i+3], fblink[OPP_DIR(nu)], tmpmat2, QDP_all);
//tmpmat = U_{nu}(x+mu)
QDP_M_eq_sM(tmpmat, fblink[nu], QDP_neighbor[mu], QDP_forward, QDP_all);
QDP_M_eq_M_times_Ma(staples, Amu[i], tmpmat, QDP_all);
QDP_M_peq_r_times_M(tempmom_qdp[mu], &Plaq, staples, QDP_all);
//tmpmat = U_{-nu}(x+mu)
QDP_M_eq_sM(tmpmat, fblink[OPP_DIR(nu)], QDP_neighbor[mu], QDP_forward, QDP_all);
QDP_M_eq_Ma_times_M(tmpmat3, fblink[OPP_DIR(nu)], staples, QDP_all);
QDP_M_eq_M_times_M(tmpmat4, tmpmat3, tmpmat, QDP_all);
QDP_M_eq_sM(tmpmat, tmpmat4, QDP_neighbor[nu], QDP_forward, QDP_all);
QDP_M_peq_r_times_M(tempmom_qdp[mu], &Rect, tmpmat, QDP_all);
QDP_M_eq_Ma_times_M(tmpmat4, tmpmat2, tmpmat3, QDP_all);
QDP_M_eq_sM(tmpmat, tmpmat4, QDP_neighbor[nu], QDP_forward, QDP_all);
QDP_M_eq_sM(tmpmat3, tmpmat, QDP_neighbor[mu], QDP_backward, QDP_all);
QDP_M_peq_r_times_M(tempmom_qdp[nu], &Rect, tmpmat3, QDP_all);
//tmpmat = U_{-nu}(x+mu)
QDP_M_eq_sM(tmpmat, fblink[OPP_DIR(nu)], QDP_neighbor[mu], QDP_forward, QDP_all);
QDP_M_eq_M_times_Ma(tmpmat3, tmpmat2, tmpmat, QDP_all);
QDP_M_eq_M_times_Ma(tmpmat, tmpmat3, staples, QDP_all);
QDP_M_eq_sM(tmpmat3, tmpmat, QDP_neighbor[nu], QDP_forward, QDP_all);
QDP_M_peq_r_times_M(tempmom_qdp[nu], &Rect, tmpmat3, QDP_all);
//tmpmat = U_{-nu}(x+mu)
QDP_M_eq_sM(tmpmat, fblink[OPP_DIR(nu)], QDP_neighbor[mu], QDP_forward, QDP_all);
QDP_M_eq_M_times_Ma(staples, Amu[i+3], tmpmat, QDP_all);
QDP_M_peq_r_times_M(tempmom_qdp[mu], &Plaq, staples, QDP_all);
QDP_M_eq_Ma_times_M(tmpmat3, fblink[nu], staples, QDP_all);
QDP_M_eq_sM(tmpmat, fblink[nu], QDP_neighbor[mu], QDP_forward, QDP_all);
QDP_M_eq_M_times_M(tmpmat4, tmpmat3, tmpmat, QDP_all);
QDP_M_eq_sM(tmpmat, tmpmat4, QDP_neighbor[nu], QDP_backward, QDP_all);
QDP_M_peq_r_times_M(tempmom_qdp[mu], &Rect, tmpmat, QDP_all);
QDP_M_eq_Ma_times_M(tmpmat, tmpmat3, tmpmat1, QDP_all);
QDP_M_eq_sM(tmpmat4, tmpmat, QDP_neighbor[mu], QDP_backward, QDP_all);
QDP_M_peq_r_times_M(tempmom_qdp[nu], &Rect, tmpmat4, QDP_all);
QDP_M_eq_sM(tmpmat, fblink[nu], QDP_neighbor[mu], QDP_forward, QDP_all);
QDP_M_eq_M_times_M(tmpmat3, staples, tmpmat, QDP_all);
QDP_M_eq_M_times_Ma(tmpmat4, tmpmat3, tmpmat1, QDP_all);
QDP_M_peq_r_times_M(tempmom_qdp[nu], &Rect, tmpmat4, QDP_all);
i++;
}
}
// Construct the pgm staples and add them to force
QDP_M_eq_zero(staples, QDP_all);
i=0;
for(nu=0; nu<4; nu++){
if(nu!=mu){
k=0;
for(sig=0; sig<4;sig ++){
if(sig!=mu && nu!=sig){
// the nu_sig_mu ... staple and 3 reflections
//tmpmat = Amu["sig"](x+nu)
QDP_M_eq_sM(tmpmat, Amu[j[i][k]], QDP_neighbor[nu], QDP_forward, QDP_all);
//tmpmat1 = Unu(x)*Amu["sig"](x+nu)
QDP_M_eq_M_times_M(tmpmat1, fblink[nu], tmpmat, QDP_all);
//tmpmat3 = Unu(x+mu+sig)
QDP_M_eq_sM(tmpmat, fblink[nu], QDP_neighbor[mu], QDP_forward, QDP_all);
QDP_M_eq_sM(tmpmat3, tmpmat, QDP_neighbor[sig], QDP_forward, QDP_all); // HERE?
//tmpmat2 = Unu(x)*Amu["sig"](x+nu)*adj(Unu(x+mu+sig))
QDP_M_eq_M_times_Ma(tmpmat2, tmpmat1, tmpmat3, QDP_all);
//tmpmat = Usig(x+mu)
QDP_M_eq_sM(tmpmat, fblink[sig], QDP_neighbor[mu], QDP_forward, QDP_all);
//tmpmat1 = Unu(x)*Amu["sig"](x+nu)*adj(Unu(x+mu+sig))*adj(Usig(x+mu))
QDP_M_eq_M_times_Ma(tmpmat1, tmpmat2, tmpmat, QDP_all);
QDP_M_peq_M(staples, tmpmat1, QDP_all);
//tmpmat = Amu["sig"](x-nu)
QDP_M_eq_sM(tmpmat, Amu[j[i][k]], QDP_neighbor[nu], QDP_backward, QDP_all);
//tmpmat1 = U_{-nu}(x)*Amu["sig"](x-nu)
QDP_M_eq_M_times_M(tmpmat1, fblink[OPP_DIR(nu)], tmpmat, QDP_all);
//tmpmat3 = U_{-nu}(x+mu+sig)
QDP_M_eq_sM(tmpmat, fblink[OPP_DIR(nu)], QDP_neighbor[mu], QDP_forward, QDP_all);
QDP_M_eq_sM(tmpmat3, tmpmat, QDP_neighbor[sig], QDP_forward, QDP_all); // HERE?
//tmpmat2 = U_{-nu}nu(x)*Amu["sig"](x-nu)*adj(Unu(x+mu+sig))
QDP_M_eq_M_times_Ma(tmpmat2, tmpmat1, tmpmat3, QDP_all);
//tmpmat = Usig(x+mu)
QDP_M_eq_sM(tmpmat, fblink[sig], QDP_neighbor[mu], QDP_forward, QDP_all);
//tmpmat1 = U_{-nu}(x)*Amu["sig"](x-nu)*adj(Unu(x+mu+sig))*adj(Usig(x+mu))
QDP_M_eq_M_times_Ma(tmpmat1, tmpmat2, tmpmat, QDP_all);
QDP_M_peq_M(staples, tmpmat1, QDP_all);
//tmpmat = Amu["-sig"](x-nu)
QDP_M_eq_sM(tmpmat, Amu[j[i][k]+3], QDP_neighbor[nu], QDP_backward, QDP_all);
//tmpmat1 = U_{-nu}(x)*Amu["-sig"](x-nu)
QDP_M_eq_M_times_M(tmpmat1, fblink[OPP_DIR(nu)], tmpmat, QDP_all);
//tmpmat = U_{-nu}(x+mu-sig)
QDP_M_eq_sM(tmpmat, fblink[OPP_DIR(nu)], QDP_neighbor[mu], QDP_forward, QDP_all);
QDP_M_eq_sM(tmpmat3, tmpmat, QDP_neighbor[sig], QDP_backward, QDP_all); // HERE?
//tmpmat2 = U_{-nu}nu(x)*Amu["-sig"](x-nu)*adj(Unu(x+mu-sig))
QDP_M_eq_M_times_Ma(tmpmat2, tmpmat1, tmpmat3, QDP_all);
//tmpmat = U_{-sig}(x+mu)
QDP_M_eq_sM(tmpmat, fblink[OPP_DIR(sig)], QDP_neighbor[mu], QDP_forward, QDP_all);
//tmpmat1 = U_{-nu}(x)*Amu["-sig"](x-nu)*adj(Unu(x+mu-sig))*adj(U_{-sig}(x+mu))
QDP_M_eq_M_times_Ma(tmpmat1, tmpmat2, tmpmat, QDP_all);
QDP_M_peq_M(staples, tmpmat1, QDP_all);
//tmpmat = Amu["-sig"](x+nu)
QDP_M_eq_sM(tmpmat, Amu[j[i][k]+3], QDP_neighbor[nu], QDP_forward, QDP_all);
//tmpmat1 = Unu(x)*Amu["-sig"](x+nu)
QDP_M_eq_M_times_M(tmpmat1, fblink[nu], tmpmat, QDP_all);
//tmpmat3 = Unu(x+mu-sig)
QDP_M_eq_sM(tmpmat, fblink[nu], QDP_neighbor[mu], QDP_forward, QDP_all);
QDP_M_eq_sM(tmpmat3, tmpmat, QDP_neighbor[sig], QDP_backward, QDP_all); // HERE?
//tmpmat2 = Unu(x)*Amu["-sig"](x+nu)*adj(Unu(x+mu-sig))
QDP_M_eq_M_times_Ma(tmpmat2, tmpmat1, tmpmat3, QDP_all);
//tmpmat = U_{-sig}(x+mu)
QDP_M_eq_sM(tmpmat, fblink[OPP_DIR(sig)], QDP_neighbor[mu], QDP_forward, QDP_all);
//tmpmat1 = Unu(x)*Amu["sig"](x+nu)*adj(Unu(x+mu+sig))*adj(Usig(x+mu))
QDP_M_eq_M_times_Ma(tmpmat1, tmpmat2, tmpmat, QDP_all);
QDP_M_peq_M(staples, tmpmat1, QDP_all);
k++;
}//close if sig!=nu ...
}//close sig loop
i++;
}// close if nu!=mu
}//close the pgm nu loop
QDP_M_peq_r_times_M(tempmom_qdp[mu], &Pgm, staples, QDP_all);
}// closes the mu loop
#ifdef CHKSUM
QLA_ColorMatrix qcm;
QLA_Complex det, chk;
QLA_c_eq_r(chk, 0);
#endif
for(mu=0; mu<4; mu++){
QDP_M_eq_M_times_Ma(tmpmat, fblink[mu], tempmom_qdp[mu], QDP_all); // HERE?
QDP_M_eq_r_times_M_plus_M( tempmom_qdp[mu], &eb3, tmpmat, force->force[mu], QDP_all);// HERE?
QDP_M_eq_antiherm_M(force->force[mu], tempmom_qdp[mu], QDP_all);// HERE
#ifdef CHKSUM
QDP_m_eq_sum_M(&qcm, force->force[mu], QDP_all);
QLA_C_eq_det_M(&det, &qcm);
QLA_c_peq_c(chk, det);
#endif
}
#ifdef CHKSUM
QOP_printf0("chksum: %g %g\n", QLA_real(chk), QLA_imag(chk));
#endif
//DESTROY various fields
QDP_destroy_M(tmpmat);
QDP_destroy_M(tmpmat1);
QDP_destroy_M(tmpmat2);
QDP_destroy_M(tmpmat3);
QDP_destroy_M(staples);
QDP_destroy_M(tmpmat4);
for(mu=0; mu<4; mu++){
QDP_destroy_M(tempmom_qdp[mu]);
}
for(i=0; i<6; i++) {
QDP_destroy_M(Amu[i]);
}
for(i=4; i<8; i++) {
QDP_destroy_M(fblink[i]);
}
dtime += QOP_time();
double nflop = 96720;
info->final_sec = dtime;
info->final_flop = nflop*QDP_sites_on_node;
info->status = QOP_SUCCESS;
//QOP_printf0("Time in slow g_force: %e\n", info->final_sec);
}
static void
get_staple_plaq(QDP_ColorMatrix *staple, int mu, QDP_ColorMatrix *u[],
QOP_gauge_coeffs_t *coeffs,
QDP_Subset subset, QDP_Subset osubset)
{
#define NC QDP_get_nc(staple)
QDP_Lattice *lat = QDP_get_lattice_M(staple);
int nd = QDP_ndim_L(lat);
QDP_Shift *neighbor = QDP_neighbor_L(lat);
QLA_Real plaq = coeffs->plaquette;
QLA_Real adpl = coeffs->adjoint_plaquette;
#if 1
QDP_ColorMatrix *temp1, *temp2, *temp3, *temp4, *temp5, *temp6;
//temp1 = QDP_create_M();
temp2 = QDP_create_M_L(lat);
//temp3 = QDP_create_M();
temp4 = QDP_create_M_L(lat);
//temp5 = QDP_create_M();
temp6 = QDP_create_M_L(lat);
QDP_Complex *tc = NULL;
if(adpl!=0) tc = QDP_create_C_L(lat);
/* staple += u[nu](x) u[mu](x+nu) u*[nu](x+mu)
* + u*[nu](x-nu) u[mu](x-nu) u[nu](x-nu+mu) */
for(int nu=0; nu<nd; nu++) {
if (nu == mu) continue;
temp1 = QDP_create_M_L(lat);
temp3 = QDP_create_M_L(lat);
temp5 = QDP_create_M_L(lat);
QDP_M_eq_sM(temp1, u[nu], neighbor[mu], QDP_forward, QDP_all_L(lat));
QDP_M_eq_Ma_times_M(temp2, u[nu], u[mu], osubset);
QDP_M_eq_sM(temp3, u[mu], neighbor[nu], QDP_forward, subset);
QDP_M_eq_M_times_M(temp4, temp2, temp1, osubset);
QDP_M_eq_sM(temp5, temp4, neighbor[nu], QDP_backward, subset);
QDP_M_eq_M_times_M(temp6, u[nu], temp3, subset);
//QDP_M_peq_M_times_Ma(staple, temp6, temp1, subset);
//QDP_M_peq_M(staple, temp5, subset);
if(adpl==0) {
QDP_M_peq_M_times_Ma(temp5, temp6, temp1, subset);
QDP_M_peq_r_times_M(staple, &plaq, temp5, subset);
} else {
QLA_Complex z;
QLA_c_eq_r(z, plaq/adpl);
QDP_C_eq_c(tc, &z, subset);
QDP_M_eq_M_times_Ma(temp2, temp6, temp1, subset);
QDP_C_peq_M_dot_M(tc, temp2, u[mu], subset);
QDP_C_eq_r_times_C(tc, &adpl, tc, subset);
QDP_M_peq_C_times_M(staple, tc, temp2, subset);
QDP_C_eq_c(tc, &z, subset);
QDP_C_peq_M_dot_M(tc, temp5, u[mu], subset);
QDP_C_eq_r_times_C(tc, &adpl, tc, subset);
QDP_M_peq_C_times_M(staple, tc, temp5, subset);
}
//QDP_discard_M(temp1);
//QDP_discard_M(temp3);
//QDP_discard_M(temp5);
QDP_destroy_M(temp1);
QDP_destroy_M(temp3);
QDP_destroy_M(temp5);
} /* closes nu loop */
if(adpl!=0) QDP_destroy_C(tc);
//QDP_destroy_M(temp1);
QDP_destroy_M(temp2);
//QDP_destroy_M(temp3);
QDP_destroy_M(temp4);
//QDP_destroy_M(temp5);
QDP_destroy_M(temp6);
#else
QDP_ColorMatrix *t = QDP_create_M_L(lat);
int nu, path[3];
QDP_Subset subs[2];
subs[0] = subset;
subs[1] = osubset;
for(nu=0; nu<nd; nu++) {
if (nu == mu) continue;
path[0] = 1+nu;
path[1] = -(1+mu);
path[2] = -(1+nu);
path_prod(u, t, path, 3, 1, subs, neighsubeo);
QDP_M_peq_M(staple, t, subset);
path[0] = -(1+nu);
path[1] = -(1+mu);
path[2] = 1+nu;
path_prod(u, t, path, 3, 1, subs, neighsubeo);
QDP_M_peq_M(staple, t, subset);
}
QDP_destroy_M(t);
#endif
}
