void Hopping_Matrix(const int ieo, spinor * const l, spinor * const k){
int ix;
su3 * restrict ALIGN U;
spinor * restrict ALIGN s;
halfspinor * restrict * phi ALIGN;
halfspinor32 * restrict * phi32 ALIGN;
/* We have 32 registers available */
_declare_hregs();
#ifdef _KOJAK_INST
#pragma pomp inst begin(hoppingmatrix)
#endif
#pragma disjoint(*s, *U)
#ifdef _GAUGE_COPY
if(g_update_gauge_copy) {
update_backward_gauge(g_gauge_field);
}
#endif
__alignx(16, l);
__alignx(16, k);
if(g_sloppy_precision == 1 && g_sloppy_precision_flag == 1) {
__alignx(16, HalfSpinor32);
/* We will run through the source vector now */
/* instead of the solution vector */
s = k;
_prefetch_spinor(s);
/* s contains the source vector */
if(ieo == 0) {
U = g_gauge_field_copy[0][0];
}
else {
U = g_gauge_field_copy[1][0];
}
phi32 = NBPointer32[ieo];
_prefetch_su3(U);
/**************** loop over all lattice sites ******************/
ix=0;
for(int i = 0; i < (VOLUME)/2; i++){
/*********************** direction +0 ************************/
_hop_t_p_pre32();
s++;
U++;
ix++;
/*********************** direction -0 ************************/
_hop_t_m_pre32();
ix++;
/*********************** direction +1 ************************/
_hop_x_p_pre32();
ix++;
U++;
/*********************** direction -1 ************************/
_hop_x_m_pre32();
ix++;
/*********************** direction +2 ************************/
_hop_y_p_pre32();
ix++;
U++;
/*********************** direction -2 ************************/
_hop_y_m_pre32();
ix++;
/*********************** direction +3 ************************/
_hop_z_p_pre32();
ix++;
U++;
/*********************** direction -3 ************************/
_hop_z_m_pre32();
ix++;
/************************ end of loop ************************/
}
# if (defined TM_USE_MPI && !defined _NO_COMM)
xchange_halffield32();
# endif
s = l;
phi32 = NBPointer32[2 + ieo];
if(ieo == 0) {
U = g_gauge_field_copy[1][0];
}
else {
U = g_gauge_field_copy[0][0];
}
//_prefetch_halfspinor(phi32[0]);
_prefetch_su3(U);
/* Now we sum up and expand to a full spinor */
ix = 0;
/* _prefetch_spinor_for_store(s); */
for(int i = 0; i < (VOLUME)/2; i++){
/* This causes a lot of trouble, do we understand this? */
/* _prefetch_spinor_for_store(s); */
//_prefetch_halfspinor(phi32[ix+1]);
/*********************** direction +0 ************************/
_hop_t_p_post32();
ix++;
/*********************** direction -0 ************************/
_hop_t_m_post32();
U++;
ix++;
/*********************** direction +1 ************************/
_hop_x_p_post32();
ix++;
/*********************** direction -1 ************************/
_hop_x_m_post32();
U++;
ix++;
/*********************** direction +2 ************************/
_hop_y_p_post32();
ix++;
/*********************** direction -2 ************************/
_hop_y_m_post32();
U++;
ix++;
/*********************** direction +3 ************************/
_hop_z_p_post32();
ix++;
/*********************** direction -3 ************************/
_hop_z_m_post32();
U++;
ix++;
s++;
}
}
else {
__alignx(16, HalfSpinor);
/* We will run through the source vector now */
/* instead of the solution vector */
s = k;
_prefetch_spinor(s);
/* s contains the source vector */
if(ieo == 0) {
U = g_gauge_field_copy[0][0];
}
else {
U = g_gauge_field_copy[1][0];
}
phi = NBPointer[ieo];
_prefetch_su3(U);
/**************** loop over all lattice sites ******************/
ix=0;
for(int i = 0; i < (VOLUME)/2; i++){
/*********************** direction +0 ************************/
_hop_t_p_pre();
s++;
U++;
ix++;
/*********************** direction -0 ************************/
_hop_t_m_pre();
ix++;
/*********************** direction +1 ************************/
_hop_x_p_pre();
ix++;
U++;
/*********************** direction -1 ************************/
_hop_x_m_pre();
ix++;
/*********************** direction +2 ************************/
_hop_y_p_pre();
ix++;
U++;
/*********************** direction -2 ************************/
_hop_y_m_pre();
ix++;
/*********************** direction +3 ************************/
_hop_z_p_pre();
ix++;
U++;
/*********************** direction -3 ************************/
_hop_z_m_pre();
ix++;
/************************ end of loop ************************/
}
# if (defined TM_USE_MPI && !defined _NO_COMM)
xchange_halffield();
# endif
s = l;
phi = NBPointer[2 + ieo];
//_prefetch_halfspinor(phi[0]);
if(ieo == 0) {
U = g_gauge_field_copy[1][0];
}
else {
U = g_gauge_field_copy[0][0];
}
_prefetch_su3(U);
/* Now we sum up and expand to a full spinor */
ix = 0;
/* _prefetch_spinor_for_store(s); */
for(int i = 0; i < (VOLUME)/2; i++){
/* This causes a lot of trouble, do we understand this? */
/* _prefetch_spinor_for_store(s); */
//_prefetch_halfspinor(phi[ix+1]);
/*********************** direction +0 ************************/
_hop_t_p_post();
ix++;
/*********************** direction -0 ************************/
_hop_t_m_post();
U++;
ix++;
/*********************** direction +1 ************************/
_hop_x_p_post();
ix++;
/*********************** direction -1 ************************/
_hop_x_m_post();
U++;
ix++;
/*********************** direction +2 ************************/
_hop_y_p_post();
ix++;
/*********************** direction -2 ************************/
_hop_y_m_post();
U++;
ix++;
/*********************** direction +3 ************************/
_hop_z_p_post();
ix++;
/*********************** direction -3 ************************/
_hop_z_m_post();
U++;
ix++;
s++;
}
}
#ifdef _KOJAK_INST
#pragma pomp inst end(hoppingmatrix)
#endif
}
/* for ieo=0, k resides on odd sites and l on even sites */
void Hopping_Matrix(const int ieo, spinor * const l, spinor * const k){
int i,ix;
su3 * restrict U ALIGN;
spinor * restrict s ALIGN;
spinor rs;
static su3_vector psi, chi, psi2, chi2;
halfspinor * restrict * phi ALIGN;
halfspinor32 * restrict * phi32 ALIGN;
#ifdef _KOJAK_INST
#pragma pomp inst begin(hoppingmatrix)
#endif
#ifdef XLC
#pragma disjoint(*l, *k, *U, *s)
#endif
#ifdef _GAUGE_COPY
if(g_update_gauge_copy) {
update_backward_gauge();
}
#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);
}
s = k;
if(ieo == 0) {
U = g_gauge_field_copy[0][0];
}
else {
U = g_gauge_field_copy[1][0];
}
if(g_sloppy_precision == 1 && g_sloppy_precision_flag == 1) {
phi32 = NBPointer32[ieo];
/**************** loop over all lattice sites ****************/
ix=0;
for(i = 0; i < (VOLUME)/2; i++){
_vector_assign(rs.s0, (*s).s0);
_vector_assign(rs.s1, (*s).s1);
_vector_assign(rs.s2, (*s).s2);
_vector_assign(rs.s3, (*s).s3);
s++;
/*********************** direction +0 ************************/
_vector_add(psi, rs.s0, rs.s2);
_su3_multiply(chi,(*U),psi);
_complex_times_vector((*phi32[ix]).s0, ka0, chi);
_vector_add(psi, rs.s1, rs.s3);
_su3_multiply(chi,(*U),psi);
_complex_times_vector((*phi32[ix]).s1, ka0, chi);
U++;
ix++;
/*********************** direction -0 ************************/
_vector_sub((*phi32[ix]).s0, rs.s0, rs.s2);
_vector_sub((*phi32[ix]).s1, rs.s1, rs.s3);
ix++;
/*********************** direction +1 ************************/
_vector_i_add(psi, rs.s0, rs.s3);
_su3_multiply(chi, (*U), psi);
_complex_times_vector((*phi32[ix]).s0, ka1, chi);
_vector_i_add(psi, rs.s1, rs.s2);
_su3_multiply(chi, (*U), psi);
_complex_times_vector((*phi32[ix]).s1, ka1, chi);
U++;
ix++;
/*********************** direction -1 ************************/
_vector_i_sub((*phi32[ix]).s0, rs.s0, rs.s3);
_vector_i_sub((*phi32[ix]).s1, rs.s1, rs.s2);
ix++;
/*********************** direction +2 ************************/
_vector_add(psi, rs.s0, rs.s3);
_su3_multiply(chi,(*U),psi);
_complex_times_vector((*phi32[ix]).s0, ka2, chi);
_vector_sub(psi, rs.s1, rs.s2);
_su3_multiply(chi,(*U),psi);
_complex_times_vector((*phi32[ix]).s1, ka2, chi);
U++;
ix++;
/*********************** direction -2 ************************/
_vector_sub((*phi32[ix]).s0, rs.s0, rs.s3);
_vector_add((*phi32[ix]).s1, rs.s1, rs.s2);
ix++;
/*********************** direction +3 ************************/
_vector_i_add(psi, rs.s0, rs.s2);
_su3_multiply(chi, (*U), psi);
_complex_times_vector((*phi32[ix]).s0, ka3, chi);
_vector_i_sub(psi, rs.s1, rs.s3);
_su3_multiply(chi,(*U),psi);
_complex_times_vector((*phi32[ix]).s1, ka3, chi);
U++;
ix++;
/*********************** direction -3 ************************/
_vector_i_sub((*phi32[ix]).s0, rs.s0, rs.s2);
_vector_i_add((*phi32[ix]).s1, rs.s1, rs.s3);
ix++;
/************************ end of loop ************************/
}
# if (defined MPI && !defined _NO_COMM)
xchange_halffield32();
# endif
s = l;
phi32 = NBPointer32[2 + ieo];
if(ieo == 0) {
U = g_gauge_field_copy[1][0];
}
else {
U = g_gauge_field_copy[0][0];
}
ix = 0;
for(i = 0; i < (VOLUME)/2; i++){
/*********************** direction +0 ************************/
_vector_assign(rs.s0, (*phi32[ix]).s0);
_vector_assign(rs.s2, (*phi32[ix]).s0);
_vector_assign(rs.s1, (*phi32[ix]).s1);
_vector_assign(rs.s3, (*phi32[ix]).s1);
ix++;
/*********************** direction -0 ************************/
_vector_assign(psi, (*phi32[ix]).s0);
_su3_inverse_multiply(chi,(*U), psi);
_complexcjg_times_vector(psi,ka0,chi);
_vector_add_assign(rs.s0, psi);
_vector_sub_assign(rs.s2, psi);
_vector_assign(psi, (*phi32[ix]).s1);
_su3_inverse_multiply(chi,(*U), psi);
_complexcjg_times_vector(psi,ka0,chi);
_vector_add_assign(rs.s1, psi);
_vector_sub_assign(rs.s3, psi);
ix++;
U++;
/*********************** direction +1 ************************/
_vector_add_assign(rs.s0, (*phi32[ix]).s0);
_vector_i_sub_assign(rs.s3, (*phi32[ix]).s0);
_vector_add_assign(rs.s1, (*phi32[ix]).s1);
_vector_i_sub_assign(rs.s2, (*phi32[ix]).s1);
ix++;
/*********************** direction -1 ************************/
_vector_assign(psi, (*phi32[ix]).s0);
_su3_inverse_multiply(chi,(*U), psi);
_complexcjg_times_vector(psi,ka1,chi);
_vector_add_assign(rs.s0, psi);
_vector_i_add_assign(rs.s3, psi);
_vector_assign(psi, (*phi32[ix]).s1);
_su3_inverse_multiply(chi,(*U), psi);
_complexcjg_times_vector(psi,ka1,chi);
_vector_add_assign(rs.s1, psi);
_vector_i_add_assign(rs.s2, psi);
U++;
ix++;
/*********************** direction +2 ************************/
_vector_add_assign(rs.s0, (*phi32[ix]).s0);
_vector_add_assign(rs.s3, (*phi32[ix]).s0);
_vector_add_assign(rs.s1, (*phi32[ix]).s1);
_vector_sub_assign(rs.s2, (*phi32[ix]).s1);
ix++;
/*********************** direction -2 ************************/
_vector_assign(psi, (*phi32[ix]).s0);
_su3_inverse_multiply(chi,(*U), psi);
_complexcjg_times_vector(psi,ka2,chi);
_vector_add_assign(rs.s0, psi);
_vector_sub_assign(rs.s3, psi);
_vector_assign(psi, (*phi32[ix]).s1);
_su3_inverse_multiply(chi, (*U), psi);
_complexcjg_times_vector(psi,ka2,chi);
_vector_add_assign(rs.s1, psi);
_vector_add_assign(rs.s2, psi);
U++;
ix++;
/*********************** direction +3 ************************/
_vector_add_assign(rs.s0, (*phi32[ix]).s0);
_vector_i_sub_assign(rs.s2, (*phi32[ix]).s0);
_vector_add_assign(rs.s1, (*phi32[ix]).s1);
_vector_i_add_assign(rs.s3, (*phi32[ix]).s1);
ix++;
/*********************** direction -3 ************************/
_vector_assign(psi, (*phi32[ix]).s0);
_su3_inverse_multiply(chi,(*U), psi);
_complexcjg_times_vector(psi,ka3,chi);
_vector_add((*s).s0, rs.s0, psi);
_vector_i_add((*s).s2, rs.s2, psi);
_vector_assign(psi, (*phi32[ix]).s1);
_su3_inverse_multiply(chi,(*U), psi);
_complexcjg_times_vector(psi,ka3,chi);
_vector_add((*s).s1, rs.s1, psi);
_vector_i_sub((*s).s3, rs.s3, psi);
U++;
ix++;
s++;
}
}
else {
phi = NBPointer[ieo];
/**************** loop over all lattice sites ****************/
ix=0;
/* #pragma ivdep*/
for(i = 0; i < (VOLUME)/2; i++){
_vector_assign(rs.s0, (*s).s0);
_vector_assign(rs.s1, (*s).s1);
_vector_assign(rs.s2, (*s).s2);
_vector_assign(rs.s3, (*s).s3);
s++;
/*********************** direction +0 ************************/
_vector_add(psi, rs.s0, rs.s2);
_vector_add(psi2, rs.s1, rs.s3);
_su3_multiply(chi,(*U),psi);
_su3_multiply(chi2,(*U),psi2);
_complex_times_vector((*phi[ix]).s0, ka0, chi);
_complex_times_vector((*phi[ix]).s1, ka0, chi2);
U++;
ix++;
/*********************** direction -0 ************************/
_vector_sub((*phi[ix]).s0, rs.s0, rs.s2);
_vector_sub((*phi[ix]).s1, rs.s1, rs.s3);
ix++;
/*********************** direction +1 ************************/
_vector_i_add(psi, rs.s0, rs.s3);
_vector_i_add(psi2, rs.s1, rs.s2);
_su3_multiply(chi, (*U), psi);
_su3_multiply(chi2, (*U), psi2);
_complex_times_vector((*phi[ix]).s0, ka1, chi);
_complex_times_vector((*phi[ix]).s1, ka1, chi2);
U++;
ix++;
/*********************** direction -1 ************************/
_vector_i_sub((*phi[ix]).s0, rs.s0, rs.s3);
_vector_i_sub((*phi[ix]).s1, rs.s1, rs.s2);
ix++;
/*********************** direction +2 ************************/
_vector_add(psi, rs.s0, rs.s3);
_vector_sub(psi2, rs.s1, rs.s2);
_su3_multiply(chi,(*U),psi);
_su3_multiply(chi2,(*U),psi2);
_complex_times_vector((*phi[ix]).s0, ka2, chi);
_complex_times_vector((*phi[ix]).s1, ka2, chi2);
U++;
ix++;
/*********************** direction -2 ************************/
_vector_sub((*phi[ix]).s0, rs.s0, rs.s3);
_vector_add((*phi[ix]).s1, rs.s1, rs.s2);
ix++;
/*********************** direction +3 ************************/
_vector_i_add(psi, rs.s0, rs.s2);
_vector_i_sub(psi2, rs.s1, rs.s3);
_su3_multiply(chi, (*U), psi);
_su3_multiply(chi2,(*U),psi2);
_complex_times_vector((*phi[ix]).s0, ka3, chi);
_complex_times_vector((*phi[ix]).s1, ka3, chi2);
U++;
ix++;
/*********************** direction -3 ************************/
_vector_i_sub((*phi[ix]).s0, rs.s0, rs.s2);
_vector_i_add((*phi[ix]).s1, rs.s1, rs.s3);
ix++;
/************************ end of loop ************************/
}
# if (defined MPI && !defined _NO_COMM)
xchange_halffield();
# endif
s = l;
phi = NBPointer[2 + ieo];
if(ieo == 0) {
U = g_gauge_field_copy[1][0];
}
else {
U = g_gauge_field_copy[0][0];
}
ix = 0;
/* #pragma ivdep */
for(i = 0; i < (VOLUME)/2; i++){
/*********************** direction +0 ************************/
_vector_assign(rs.s0, (*phi[ix]).s0);
_vector_assign(rs.s2, (*phi[ix]).s0);
_vector_assign(rs.s1, (*phi[ix]).s1);
_vector_assign(rs.s3, (*phi[ix]).s1);
ix++;
/*********************** direction -0 ************************/
_su3_inverse_multiply(chi,(*U),(*phi[ix]).s0);
_su3_inverse_multiply(chi2,(*U),(*phi[ix]).s1);
_complexcjg_times_vector(psi,ka0,chi);
_complexcjg_times_vector(psi2,ka0,chi2);
_vector_add_assign(rs.s0, psi);
_vector_sub_assign(rs.s2, psi);
_vector_add_assign(rs.s1, psi2);
_vector_sub_assign(rs.s3, psi2);
ix++;
U++;
/*********************** direction +1 ************************/
_vector_add_assign(rs.s0, (*phi[ix]).s0);
_vector_i_sub_assign(rs.s3, (*phi[ix]).s0);
_vector_add_assign(rs.s1, (*phi[ix]).s1);
_vector_i_sub_assign(rs.s2, (*phi[ix]).s1);
ix++;
/*********************** direction -1 ************************/
_su3_inverse_multiply(chi,(*U), (*phi[ix]).s0);
_su3_inverse_multiply(chi2, (*U), (*phi[ix]).s1);
_complexcjg_times_vector(psi,ka1,chi);
_complexcjg_times_vector(psi2,ka1,chi2);
_vector_add_assign(rs.s0, psi);
_vector_i_add_assign(rs.s3, psi);
_vector_add_assign(rs.s1, psi2);
_vector_i_add_assign(rs.s2, psi2);
U++;
ix++;
/*********************** direction +2 ************************/
_vector_add_assign(rs.s0, (*phi[ix]).s0);
_vector_add_assign(rs.s3, (*phi[ix]).s0);
_vector_add_assign(rs.s1, (*phi[ix]).s1);
_vector_sub_assign(rs.s2, (*phi[ix]).s1);
ix++;
/*********************** direction -2 ************************/
_su3_inverse_multiply(chi,(*U), (*phi[ix]).s0);
_su3_inverse_multiply(chi2, (*U), (*phi[ix]).s1);
_complexcjg_times_vector(psi,ka2,chi);
_complexcjg_times_vector(psi2,ka2,chi2);
_vector_add_assign(rs.s0, psi);
_vector_sub_assign(rs.s3, psi);
_vector_add_assign(rs.s1, psi2);
_vector_add_assign(rs.s2, psi2);
U++;
ix++;
/*********************** direction +3 ************************/
_vector_add_assign(rs.s0, (*phi[ix]).s0);
_vector_i_sub_assign(rs.s2, (*phi[ix]).s0);
_vector_add_assign(rs.s1, (*phi[ix]).s1);
_vector_i_add_assign(rs.s3, (*phi[ix]).s1);
ix++;
/*********************** direction -3 ************************/
_su3_inverse_multiply(chi,(*U), (*phi[ix]).s0);
_su3_inverse_multiply(chi2, (*U), (*phi[ix]).s1);
_complexcjg_times_vector(psi,ka3,chi);
_complexcjg_times_vector(psi2,ka3,chi2);
_vector_add((*s).s0, rs.s0, psi);
_vector_i_add((*s).s2, rs.s2, psi);
_vector_add((*s).s1, rs.s1, psi2);
_vector_i_sub((*s).s3, rs.s3, psi2);
U++;
ix++;
s++;
}
}
#ifdef _KOJAK_INST
#pragma pomp inst end(hoppingmatrix)
#endif
}
