/* vv[nao,nao] = ao1[i,nao] * ao2[i,nao] */
void VXCdot_ao_ao(double *vv, double *ao1, double *ao2,
int nao, int ngrids, int blksize, char *non0table,
int *atm, int natm, int *bas, int nbas, double *env)
{
const int nblk = (ngrids+blksize-1) / blksize;
int ip, ib;
double *v_priv;
memset(vv, 0, sizeof(double) * nao * nao);
#pragma omp parallel default(none) \
shared(vv, ao1, ao2, nao, ngrids, blksize, non0table, \
atm, natm, bas, nbas, env) \
private(ip, ib, v_priv)
{
v_priv = malloc(sizeof(double) * nao * nao);
memset(v_priv, 0, sizeof(double) * nao * nao);
#pragma omp for nowait schedule(static)
for (ib = 0; ib < nblk; ib++) {
ip = ib * blksize;
dot_ao_ao(v_priv, ao1+ip*nao, ao2+ip*nao,
nao, MIN(ngrids-ip, blksize), non0table+ib*nbas,
atm, natm, bas, nbas, env);
}
#pragma omp critical
{
for (ip = 0; ip < nao*nao; ip++) {
vv[ip] += v_priv[ip];
}
}
free(v_priv);
}
}
/* vv[nao,nao] = ao1[i,nao] * ao2[i,nao] */
void VXCdot_ao_ao(double *vv, double *ao1, double *ao2,
int nao, int ngrids, int nbas, int hermi,
unsigned char *non0table, int *shls_slice, int *ao_loc)
{
const int nblk = (ngrids+BLKSIZE-1) / BLKSIZE;
memset(vv, 0, sizeof(double) * nao * nao);
#pragma omp parallel
{
int ip, ib;
double *v_priv = calloc(nao*nao+2, sizeof(double));
#pragma omp for nowait schedule(static)
for (ib = 0; ib < nblk; ib++) {
ip = ib * BLKSIZE;
dot_ao_ao(v_priv, ao1+ip, ao2+ip,
nao, ngrids, MIN(ngrids-ip, BLKSIZE), hermi,
non0table+ib*nbas, shls_slice, ao_loc);
}
#pragma omp critical
{
for (ip = 0; ip < nao*nao; ip++) {
vv[ip] += v_priv[ip];
}
}
free(v_priv);
}
if (hermi != 0) {
NPdsymm_triu(nao, vv, hermi);
}
}
