/* vv[nao,nao] = conj(ao1[i,nao]) * ao2[i,nao] */
void VXCzdot_ao_ao(double complex *vv, double complex *ao1, double complex *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 complex) * nao * nao);
#pragma omp parallel default(none) \
shared(vv, ao1, ao2, nao, ngrids, nbas, hermi, \
non0table, shls_slice, ao_loc)
{
int ip, ib;
double complex *v_priv = calloc(nao*nao+2, sizeof(double complex));
#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] += conj(v_priv[ip]);
}
}
free(v_priv);
}
if (hermi != 0) {
NPzhermi_triu(nao, vv, hermi);
}
}
void NPzunpack_tril(int n, double complex *tril, double complex *mat,
int hermi)
{
size_t i, j, ij;
for (ij = 0, i = 0; i < n; i++) {
for (j = 0; j <= i; j++, ij++) {
mat[i*n+j] = tril[ij];
}
}
if (hermi) {
NPzhermi_triu(n, mat, hermi);
}
}
