static void remap_data(int index, ft_data *ftd){
msg_tag *mtag;
char *temp;
int i;
double dtime = start_timing();
temp = (char *)malloc(sites_on_node*ftd->size);
if(temp==NULL){
printf("remap_data: No room\n");
terminate(1);
}
mtag = start_gather_field(ftd->data, ftd->size, index, EVENANDODD,
gen_pt[0]);
wait_gather(mtag);
/* First copy gathered data to temporary */
for(i = 0; i < sites_on_node; i++)
memcpy(temp + ftd->size*i, gen_pt[0][i], ftd->size);
cleanup_gather(mtag);
/* Then copy temp back to field */
memcpy((char *)ftd->data, temp, sites_on_node*ftd->size);
free(temp);
print_timing(dtime, "REMAP FFTW remap");
}
void
dslash_fn_dir(su3_vector *src, su3_vector *dest, int parity,
fn_links_t *fn, int dir, int fb,
Real wtfat, Real wtlong)
{
register int i ;
site *s;
msg_tag *tag[2];
su3_matrix *fat = get_fatlinks(fn);
su3_matrix *lng = get_lnglinks(fn);
su3_vector tmp;
int do_long = (lng != NULL) && (wtlong != 0.);
char myname[] = "fn_shift";
if(fat == NULL) {
printf("%s(%d): fat or lng member is null\n", myname, this_node);
terminate(1);
}
if(fb > 0){
/* Shift from forward direction */
tag[0] = start_gather_field( src, sizeof(su3_vector), dir,
parity, gen_pt[0] );
if(do_long)
tag[1] = start_gather_field( src, sizeof(su3_vector), DIR3(dir),
parity, gen_pt[1] );
wait_gather(tag[0]);
if(do_long)
wait_gather(tag[1]);
FORSOMEPARITYDOMAIN(i,s,parity)
{
mult_su3_mat_vec( fat+4*i+dir, (su3_vector *)gen_pt[0][i], &tmp );
scalar_mult_add_su3_vector( dest+i, &tmp, wtfat, dest+i ) ;
if(do_long){
mult_su3_mat_vec( lng+4*i+dir, (su3_vector *)gen_pt[1][i], &tmp );
scalar_mult_add_su3_vector( dest+i, &tmp, wtlong, dest+i ) ;
}
} END_LOOP
/* Apply the symmetric shift operator in direction "dir" *
* This is the explicit version *
* Covariant shifts are used *
* The KS phases MUST BE in the links */
static void
sym_shift_field(int dir, su3_vector *src, su3_vector *dest)
{
register int i ;
register site *s ;
msg_tag *tag[2];
su3_vector *tvec = create_v_field();
tag[0] = start_gather_field( src, sizeof(su3_vector), dir, EVENANDODD, gen_pt[0] );
/* With ONE_SIDED_SHIFT defined, the shift is asymmetric */
#ifndef ONE_SIDED_SHIFT
FORALLSITES(i,s)
{
mult_adj_su3_mat_vec( &(s->link[dir]), src+i, tvec+i ) ;
}
// then gathered to x
FORALLSITES(i, s)
mult_su3_an((su3_matrix*)F_PT(s,lnk2), (su3_matrix*)F_PT(s,lnk1),
tempmat1 + i);
wait_gather(tag0);
wait_gather(tag1);
// Finish lower staple
FORALLSITES(i, s) {
mult_su3_nn(tempmat1 + i, (su3_matrix *)gen_pt[0][i], &tmat1);
su3mat_copy(&tmat1, tempmat1 + i);
}
// Gather staple from direction -dir2 to "home" site
tag2 = start_gather_field(tempmat1, sizeof(su3_matrix),
OPP_DIR(dir2), EVENANDODD, gen_pt[2]);
// Calculate upper staple, add it
FORALLSITES(i, s) {
mult_su3_nn((su3_matrix*)F_PT(s,lnk2), (su3_matrix *)gen_pt[1][i], &tmat1);
mult_su3_na(&tmat1, (su3_matrix *)gen_pt[0][i], &tmat2);
add_su3_matrix(stp + i, &tmat2, stp + i);
}
// Finally add the lower staple
wait_gather(tag2);
FORALLSITES(i, s)
add_su3_matrix(stp+i, (su3_matrix *)gen_pt[2][i], stp+i);
cleanup_gather(tag0);
cleanup_gather(tag1);
void hvy_pot(int do_det) {
register int i;
register site *s;
int t_dist, x_dist;
double wloop;
complex tc;
matrix tmat, tmat2;
msg_tag *mtag = NULL;
node0_printf("hvy_pot: MAX_T = %d, MAX_X = %d\n", MAX_T, MAX_X);
// Use staple to hold product of t_dist links at each point
for (t_dist = 1; t_dist <= MAX_T; t_dist++) {
if (t_dist == 1) {
FORALLSITES(i, s)
mat_copy(&(s->link[TUP]), &(staple[i]));
}
else {
mtag = start_gather_field(staple, sizeof(matrix),
goffset[TUP], EVENANDODD, gen_pt[0]);
// Be careful about overwriting staple;
// gen_pt may just point to it for on-node "gathers"
wait_gather(mtag);
FORALLSITES(i, s)
mult_nn(&(s->link[TUP]), (matrix *)gen_pt[0][i], &(tempmat2[i]));
cleanup_gather(mtag);
FORALLSITES(i, s)
mat_copy(&(tempmat2[i]), &(staple[i]));
}
// Copy staple to tempmat
// Will shoft at end of loop
FORALLSITES(i, s)
mat_copy(&(staple[i]), &(tempmat[i]));
for (x_dist = 0; x_dist <= MAX_X; x_dist++) {
// Evaluate potential at this separation
wloop = 0.0;
FORALLSITES(i, s) {
// Compute the actual Coulomb gauge Wilson loop product
mult_na(&(staple[i]), &(tempmat[i]), &tmat);
if (do_det == 1)
det_project(&tmat, &tmat2);
else
mat_copy(&tmat, &tmat2);
tc = trace(&tmat2);
wloop += tc.real;
}
g_doublesum(&wloop);
if (do_det == 1) { // Braces fix compiler error
node0_printf("D_LOOP ");
}
else
node0_printf("POT_LOOP ");
node0_printf("%d %d %.6g\n", x_dist, t_dist, wloop / volume);
// As we increment x, shift in x direction
shiftmat(tempmat, tempmat2, goffset[XUP]);
} // x_dist
} // t_dist
// -----------------------------------------------------------------
// Matrix--vector operation
// Applies either the operator (sign = 1) or its adjoint (sign = -1)
// Adjoint is simply overall negative sign...
void fermion_op(vector *src, vector *dest, int sign) {
register int i;
register site *s;
int dir, a, b, c, d, par, L[NDIMS] = {nx, ny, nz, nt};
Real tr, halfG = 0.5 * G, m_ov_G, vev[DIMF][DIMF];
vector tvec, tvec_dir, tvec_opp;
msg_tag *tag[2 * NDIMS];
// Quick sanity check
if (sign != 1 && sign != -1) {
node0_printf("Error: incorrect sign in fermion_op: %d\n", sign);
terminate(1);
}
// Ignore site_mass if G = 0 to avoid dividing by zero
// Could be made more robust, but unlikely to matter
if (G == 0.0)
m_ov_G = 0.0;
else
m_ov_G = 2.0 * site_mass / G;
for (a = 0; a < DIMF; a++) {
for (b = 0; b < DIMF; b++)
vev[a][b] = 0.0;
}
vev[0][1] = m_ov_G;
vev[2][3] = m_ov_G;
vev[1][0] = -m_ov_G;
vev[3][2] = -m_ov_G;
// Start gathers for kinetic term
FORALLUPDIR(dir) {
if (L[dir] <= 1) // Will be skipped below
continue;
tag[dir] = start_gather_field(src, sizeof(vector), dir,
EVENANDODD, gen_pt[dir]);
tag[OPP_DIR(dir)] = start_gather_field(src, sizeof(vector), OPP_DIR(dir),
EVENANDODD, gen_pt[OPP_DIR(dir)]);
}
// Compute scalar term as gathers run
// Initialize dest = 0.5G * (sigma + 2m / G) * src
// Add SO(4)-breaking 'site mass' term with same structure as sigma
FORALLSITES(i, s) {
clearvec(&(dest[i]));
if (stagger == -1 || lattice[i].parity == EVEN)
par = 1;
else // Both stagger == 1 and lattice[i].parity == ODD
par = -1;
for (a = 0; a < DIMF; a++) {
for (b = a + 1; b < DIMF; b++) {
tr = s->sigma.e[as_index[a][b]] + par * vev[a][b];
for (c = 0; c < DIMF; c++) {
for (d = c + 1; d < DIMF; d++) {
tr += perm[a][b][c][d] * (s->sigma.e[as_index[c][d]]
+ par * vev[c][d]);
}
} // No half since not double-counting
dest[i].c[a] += tr * src[i].c[b];
dest[i].c[b] -= tr * src[i].c[a];
}
}
scalar_mult_vec(&(dest[i]), halfG, &(dest[i]));
}
/* Special dslash for use by congrad. Uses restart_gather_field() when
possible. Next to last argument is an array of message tags, to be set
if this is the first use, otherwise reused. If start=1,use
start_gather_field, otherwise use restart_gather_field.
The calling program must clean up the gathers and temps! */
void dslash_fn_field_special(su3_vector *src, su3_vector *dest,
int parity, msg_tag **tag, int start,
fn_links_t *fn){
register int i;
register site *s;
register int dir,otherparity=0;
register su3_matrix *fat4;
su3_matrix *t_fatlink;
#ifndef NO_LONG_LINKS
register su3_matrix *long4;
su3_matrix *t_longlink;
#endif
/* allocate temporary work space only if not already allocated */
if(temp_not_allocated)
{
for( dir=XUP; dir<=TUP; dir++ ){
temp[dir] =(su3_vector *)malloc(sites_on_node*sizeof(su3_vector));
temp[dir+4]=(su3_vector *)malloc(sites_on_node*sizeof(su3_vector));
}
temp[8]=(su3_vector *)malloc(sites_on_node*sizeof(su3_vector));
temp_not_allocated = 0 ;
}
/* load fatlinks and longlinks */
if(fn == NULL){
printf("dslash_fn_field_special: invalid fn links!\n");
terminate(1);
}
#ifndef NO_LONG_LINKS
t_longlink = get_lnglinks(fn);
#endif
t_fatlink = get_fatlinks(fn);
switch(parity)
{
case EVEN: otherparity=ODD; break;
case ODD: otherparity=EVEN; break;
case EVENANDODD: otherparity=EVENANDODD; break;
}
/* Start gathers from positive directions */
/* And start the 3-step gather too */
for( dir=XUP; dir<=TUP; dir++ ){
if(start==1)
{
tag[dir] = start_gather_field( src, sizeof(su3_vector),
dir, parity,gen_pt[dir] );
#ifndef NO_LONG_LINKS
tag[DIR3(dir)] = start_gather_field(src, sizeof(su3_vector),
DIR3(dir),parity,
gen_pt[DIR3(dir)] );
#endif
}
else
{
restart_gather_field( src, sizeof(su3_vector),
dir, parity,gen_pt[dir], tag[dir]);
#ifndef NO_LONG_LINKS
restart_gather_field(src, sizeof(su3_vector), DIR3(dir), parity,
gen_pt[DIR3(dir)], tag[DIR3(dir)]);
#endif
}
}
/* Multiply by adjoint matrix at other sites */
/* Use fat link for single link transport */
FORSOMEPARITYDOMAIN_OMP( i, s, otherparity, private(fat4,long4) ){
//NOPRE if( i < loopend-FETCH_UP ){
//NOPRE fat4 = &(t_fatlink[4*(i+FETCH_UP)]);
//NOPRE prefetch_V(&(src[i+FETCH_UP]));
//NOPRE prefetch_4MVVVV(
//NOPRE fat4,
//NOPRE &(temp[0][i+FETCH_UP]),
//NOPRE &(temp[1][i+FETCH_UP]),
//NOPRE &(temp[2][i+FETCH_UP]),
//NOPRE &(temp[3][i+FETCH_UP]) );
#ifndef NO_LONG_LINKS
//NOPRE long4 = &(t_longlink[4*(i+FETCH_UP)]);
//NOPRE prefetch_4MVVVV(
//NOPRE long4,
//NOPRE &(temp[4][i+FETCH_UP]),
//NOPRE &(temp[5][i+FETCH_UP]),
//NOPRE &(temp[6][i+FETCH_UP]),
//NOPRE &(temp[7][i+FETCH_UP]) );
#endif
//NOPRE }
fat4 = &(t_fatlink[4*i]);
mult_adj_su3_mat_4vec( fat4, &(src[i]), &(temp[0][i]),
&(temp[1][i]), &(temp[2][i]), &(temp[3][i]) );
#ifndef NO_LONG_LINKS
/* multiply by 3-link matrices too */
long4 = &(t_longlink[4*i]);
mult_adj_su3_mat_4vec( long4, &(src[i]),&(temp[4][i]),
&(temp[5][i]), &(temp[6][i]), &(temp[7][i]) );
#endif
} END_LOOP_OMP
/* Start gathers from negative directions */
for( dir=XUP; dir <= TUP; dir++){
if (start==1) tag[OPP_DIR(dir)] = start_gather_field( temp[dir],
sizeof(su3_vector), OPP_DIR( dir), parity, gen_pt[OPP_DIR(dir)] );
else restart_gather_field( temp[dir], sizeof(su3_vector),
OPP_DIR( dir), parity, gen_pt[OPP_DIR(dir)], tag[OPP_DIR(dir)] );
}
/* Start 3-neighbour gathers from negative directions */
for( dir=X3UP; dir <= T3UP; dir++){
if (start==1) tag[OPP_3_DIR(dir)]=start_gather_field(
temp[INDEX_3RD(dir)+4], sizeof(su3_vector),
OPP_3_DIR( dir), parity, gen_pt[OPP_3_DIR(dir)] );
else restart_gather_field(temp[INDEX_3RD(dir)+4],
sizeof(su3_vector), OPP_3_DIR( dir),parity,
gen_pt[OPP_3_DIR(dir)], tag[OPP_3_DIR(dir)] );
}
/* Wait gathers from positive directions, multiply by matrix and
accumulate */
/* wait for the 3-neighbours from positive directions, multiply */
for(dir=XUP; dir<=TUP; dir++){
wait_gather(tag[dir]);
#ifndef NO_LONG_LINKS
wait_gather(tag[DIR3(dir)]);
#endif
}
FORSOMEPARITYDOMAIN_OMP(i,s,parity, private(fat4,long4) ){
//NOPRE if( i < loopend-FETCH_UP ){
//NOPRE fat4 = &(t_fatlink[4*(i+FETCH_UP)]);
//NOPRE prefetch_4MVVVV(
//NOPRE fat4,
//NOPRE (su3_vector *)gen_pt[XUP][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[YUP][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[ZUP][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[TUP][i+FETCH_UP] );
//NOPRE prefetch_VVVV(
//NOPRE (su3_vector *)gen_pt[XDOWN][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[YDOWN][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[ZDOWN][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[TDOWN][i+FETCH_UP] );
#ifndef NO_LONG_LINKS
//NOPRE long4 = &(t_longlink[4*(i+FETCH_UP)]);
//NOPRE prefetch_4MVVVV(
//NOPRE long4,
//NOPRE (su3_vector *)gen_pt[X3UP][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[Y3UP][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[Z3UP][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[T3UP][i+FETCH_UP] );
//NOPRE prefetch_VVVV(
//NOPRE (su3_vector *)gen_pt[X3DOWN][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[Y3DOWN][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[Z3DOWN][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[T3DOWN][i+FETCH_UP] );
#endif
//NOPRE }
fat4 = &(t_fatlink[4*i]);
mult_su3_mat_vec_sum_4dir( fat4,
(su3_vector *)gen_pt[XUP][i], (su3_vector *)gen_pt[YUP][i],
(su3_vector *)gen_pt[ZUP][i], (su3_vector *)gen_pt[TUP][i],
&(dest[i]) );
#ifndef NO_LONG_LINKS
long4 = &(t_longlink[4*i]);
mult_su3_mat_vec_sum_4dir( long4,
(su3_vector *)gen_pt[X3UP][i], (su3_vector *)gen_pt[Y3UP][i],
(su3_vector *)gen_pt[Z3UP][i], (su3_vector *)gen_pt[T3UP][i],
&(temp[8][i]));
#endif
} END_LOOP_OMP
/* Wait gathers from negative directions, accumulate (negative) */
/* and the same for the negative 3-rd neighbours */
for(dir=XUP; dir<=TUP; dir++){
wait_gather(tag[OPP_DIR(dir)]);
}
for(dir=X3UP; dir<=T3UP; dir++){
wait_gather(tag[OPP_3_DIR(dir)]);
}
FORSOMEPARITYDOMAIN_OMP(i,s,parity, ){
//NOPRE if( i < loopend-FETCH_UP ){
//NOPRE prefetch_VVVVV(
//NOPRE &(dest[i+FETCH_UP]),
//NOPRE (su3_vector *)gen_pt[XDOWN][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[YDOWN][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[ZDOWN][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[TDOWN][i+FETCH_UP] );
//NOPRE prefetch_VVVVV(
//NOPRE &(temp[8][i+FETCH_UP]),
//NOPRE (su3_vector *)gen_pt[X3DOWN][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[Y3DOWN][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[Z3DOWN][i+FETCH_UP],
//NOPRE (su3_vector *)gen_pt[T3DOWN][i+FETCH_UP] );
//NOPRE }
sub_four_su3_vecs( &(dest[i]),
(su3_vector *)(gen_pt[XDOWN][i]),
(su3_vector *)(gen_pt[YDOWN][i]),
(su3_vector *)(gen_pt[ZDOWN][i]),
(su3_vector *)(gen_pt[TDOWN][i]) );
sub_four_su3_vecs( &(temp[8][i]),
(su3_vector *)(gen_pt[X3DOWN][i]),
(su3_vector *)(gen_pt[Y3DOWN][i]),
(su3_vector *)(gen_pt[Z3DOWN][i]),
(su3_vector *)(gen_pt[T3DOWN][i]) );
/* Now need to add these things together */
add_su3_vector(&(dest[i]), &(temp[8][i]),&(dest[i]));
} END_LOOP_OMP
}
c_scalar_mult_sum_mat(&(Lambda[j]), &(in[iter]), &(plaq_src[5][i]));
iter++;
// 0, 3 --> 2
c_scalar_mult_sum_mat(&(Lambda[j]), &(in[iter]), &(src[i].Fplaq[2]));
iter++;
// 1, 2 --> 4
c_scalar_mult_sum_mat(&(Lambda[j]), &(in[iter]), &(plaq_src[4][i]));
iter++;
}
}
// Gather plaq_src[7] (2, 3) from x - 0 - 1 (gather path 23),
// plaq_src[5] (1, 3) from x - 0 - 2 (gather path 31)
// and plaq_src[4] (1, 2) from x - 0 - 3 (gather path 35)
mtag0 = start_gather_field(plaq_src[7], sizeof(matrix),
23, EVENANDODD, gen_pt[0]);
mtag1 = start_gather_field(plaq_src[5], sizeof(matrix),
31, EVENANDODD, gen_pt[1]);
mtag2 = start_gather_field(plaq_src[4], sizeof(matrix),
35, EVENANDODD, gen_pt[2]);
wait_gather(mtag0);
wait_gather(mtag1);
wait_gather(mtag2);
FORALLSITES(i, s) {
mat_copy((matrix *)(gen_pt[0][i]), &(src[i].Fplaq[7])); // 2, 3
mat_copy((matrix *)(gen_pt[1][i]), &(src[i].Fplaq[5])); // 1, 3
mat_copy((matrix *)(gen_pt[2][i]), &(src[i].Fplaq[4])); // 1, 2
}
cleanup_gather(mtag0);
cleanup_gather(mtag1);
void d_plaquette_field_hist(su3_matrix **U_field,
int Npowers, int *Nhist, double **hist,
double **hist_bounds,
double *ss_plaq, double *st_plaq) {
/* su3mat is scratch space of size su3_matrix */
su3_matrix *su3mat;
register int i,dir1,dir2;
register int ipower,ihist;
register site *s;
register su3_matrix *m1,*m4;
double *plaq_power, *step_hist;
su3_matrix mtmp;
double ss_sum,st_sum;
double rtrace, rtrace3;
msg_tag *mtag0,*mtag1;
ss_sum = st_sum = 0.0;
#ifdef HISQ_DUMP_PLAQ_INTO_FILE
FILE *fp;
char plaq_file_name[300];
#endif /* HISQ_DUMP_PLAQ_INTO_FILE */
su3mat = (su3_matrix *)malloc(sizeof(su3_matrix)*sites_on_node);
if(su3mat == NULL)
{
printf("plaquette: can't malloc su3mat\n");
fflush(stdout); terminate(1);
}
/* zero out the histogram */
for(ipower=0;ipower<Npowers;ipower++) {
for(ihist=0;ihist<Nhist[ipower];ihist++) {
hist[ipower][ihist]=0.0;
}
}
/* array with powers of (3-plaquette) */
plaq_power=(double*)malloc(sizeof(double)*Npowers);
/* array with step sizes */
step_hist=(double*)malloc(sizeof(double)*Npowers);
for(ipower=0;ipower<Npowers;ipower++) {
step_hist[ipower]=
(hist_bounds[ipower][1]-hist_bounds[ipower][0])/Nhist[ipower];
}
#ifdef HISQ_DUMP_PLAQ_INTO_FILE
sprintf( plaq_file_name, "plaq_W_node%04d.dat", this_node );
fp = fopen( plaq_file_name, "wt" );
#endif /* HISQ_DUMP_PLAQ_INTO_FILE */
for(dir1=YUP;dir1<=TUP;dir1++){
for(dir2=XUP;dir2<dir1;dir2++){
mtag0 = start_gather_field( U_field[dir2], sizeof(su3_matrix),
dir1, EVENANDODD, gen_pt[0] );
mtag1 = start_gather_field( U_field[dir1], sizeof(su3_matrix),
dir2, EVENANDODD, gen_pt[1] );
FORALLSITES(i,s){
m1 = &(U_field[dir1][i]);
m4 = &(U_field[dir2][i]);
mult_su3_an(m4,m1,&su3mat[i]);
}
wait_gather(mtag0);
wait_gather(mtag1);
FORALLSITES(i,s){
mult_su3_nn( &su3mat[i], (su3_matrix *)(gen_pt[0][i]),
&mtmp);
if(dir1==TUP ) {
rtrace = (double)
realtrace_su3((su3_matrix *)(gen_pt[1][i]),&mtmp);
st_sum += rtrace;
}
else {
rtrace = (double)
realtrace_su3((su3_matrix *)(gen_pt[1][i]),&mtmp);
ss_sum += rtrace;
}
// printf("Plaq i=%d, dir1=%d, dir2=%d: %f %f\n",
// i,dir1,dir2,rtrace_s,rtrace_t);
#ifdef HISQ_DUMP_PLAQ_INTO_FILE
fprintf( fp, "%18.12g\n", rtrace );
#endif /* HISQ_DUMP_PLAQ_INTO_FILE */
/* powers of (3-plaquette) */
rtrace3=3.0-rtrace;
plaq_power[0]=rtrace3;
for(ipower=1;ipower<Npowers;ipower++) {
plaq_power[ipower]=plaq_power[ipower-1]*rtrace3;
}
/* find histogram entry */
for(ipower=0;ipower<Npowers;ipower++) {
if( (plaq_power[ipower]>hist_bounds[ipower][0])
&& (plaq_power[ipower]<hist_bounds[ipower][1]) ) {
ihist=(int)(
(plaq_power[ipower]-hist_bounds[ipower][0])/
step_hist[ipower] );
hist[ipower][ihist]+=1.0;
}
}
}
cleanup_gather(mtag0);
cleanup_gather(mtag1);
}
void sym_shift(int dir, field_offset src,field_offset dest)
{
register int i ;
register site *s ;
msg_tag *tag[2];
su3_vector *tvec;
tvec = (su3_vector *)malloc( sites_on_node*sizeof(su3_vector) );
tag[0] = start_gather_site( src, sizeof(su3_vector), dir, EVENANDODD ,gen_pt[0] );
FORALLSITES(i,s)
{
mult_adj_su3_mat_vec( &(s->link[dir]), (su3_vector *)F_PT(s,src),
&(tvec[i]) ) ;
}
tag[1] = start_gather_field(tvec, sizeof(su3_vector), OPP_DIR(dir),
EVENANDODD ,gen_pt[1] );
wait_gather(tag[0]);
FORALLSITES(i,s)
{
mult_su3_mat_vec( &(s->link[dir]), (su3_vector *)gen_pt[0][i],
(su3_vector *)F_PT(s,dest) ) ;
}
wait_gather(tag[1]);
FORALLSITES(i,s)
{
add_su3_vector( (su3_vector *)F_PT(s,dest), (su3_vector *)gen_pt[1][i],
(su3_vector *)F_PT(s,dest) ) ;
}
/* Now devide by 2 eq. (4.2b) of Golderman's Meson paper*/
FORALLSITES(i,s)
{
}
/**cleanup_gather(mtag0); Use same gather in next plaquette**/
cleanup_gather(mtag1);
/* Plaquette in -dir0 +dir1 direction */
/**mtag0 = start_gather_site( LINK_OFFSET(dir0),
sizeof(su3_matrix), dir1, EVENANDODD, gen_pt[0] );
wait_gather(mtag0); Already gathered above**/
FORALLSITES(i,s){
mult_su3_an( &LINK(dir1),
&LINK(dir0), &tmat1 );
mult_su3_an( (su3_matrix *)(gen_pt[0][i]), &tmat1, &temp1[i] );
}
mtag1 = start_gather_field( temp1, sizeof(su3_matrix),
OPP_DIR(dir0), EVENANDODD, gen_pt[1] );
wait_gather(mtag1);
FORALLSITES(i,s){
mult_su3_nn( &LINK(dir1), (su3_matrix *)(gen_pt[1][i]), &tmat1 );
su3_adjoint( &tmat1, &tmat2 );
add_su3_matrix( &FIELD_STRENGTH(component), &tmat1,
&FIELD_STRENGTH(component) );
sub_su3_matrix( &FIELD_STRENGTH(component), &tmat2,
&FIELD_STRENGTH(component) );
}
cleanup_gather(mtag0);
cleanup_gather(mtag1);
/* Plaquette in -dir0 -dir1 direction */
mtag0 = start_gather_site( LINK_OFFSET(dir0), sizeof(su3_matrix),