void GTO2e_cart_or_sph(int (*intor)(), int (*cgto_in_shell)(), double *eri,
int *atm, int natm, int *bas, int nbas, double *env)
{
int *ao_loc = malloc(sizeof(int)*(nbas+1));
int i, j, ij;
int nao = 0;
for (i = 0; i < nbas; i++) {
ao_loc[i] = nao;
nao += (*cgto_in_shell)(i, bas);
}
ao_loc[nbas] = nao;
struct _VHFEnvs envs = {natm, nbas, atm, bas, env, nao,
ao_loc};
CINTOpt *cintopt;
cint2e_optimizer(&cintopt, atm, natm, bas, nbas, env);
CVHFOpt *vhfopt;
CVHFnr_optimizer(&vhfopt, atm, natm, bas, nbas, env);
vhfopt->fprescreen = CVHFnr_schwarz_cond;
#pragma omp parallel default(none) \
shared(intor, eri, nbas, envs, cintopt, vhfopt) \
private(ij, i, j)
#pragma omp for nowait schedule(dynamic, 2)
for (ij = 0; ij < nbas*(nbas+1)/2; ij++) {
i = (int)(sqrt(2*ij+.25) - .5 + 1e-7);
j = ij - (i*(i+1)/2);
store_ij(intor, eri, i, j, cintopt, vhfopt, &envs);
}
free(ao_loc);
CVHFdel_optimizer(&vhfopt);
CINTdel_optimizer(&cintopt);
}
void RInr_fill2c2e_sph(double *eri, int auxstart, int auxcount,
int *atm, int natm, int *bas, int nbas, double *env)
{
const int nbasnaux = auxstart + auxcount;
int *ao_loc = malloc(sizeof(int)*(nbasnaux+1));
CINTshells_spheric_offset(ao_loc, bas, nbasnaux);
ao_loc[nbasnaux] = ao_loc[nbasnaux-1] + CINTcgto_spheric(nbasnaux-1, bas);
const int naoaux = ao_loc[nbasnaux] - ao_loc[auxstart];
double *buf;
int ish, jsh, di, dj;
int i, j, i0, j0;
int shls[2];
CINTOpt *cintopt = NULL;
cint2c2e_sph_optimizer(&cintopt, atm, natm, bas, nbas, env);
#pragma omp parallel default(none) \
shared(eri, auxstart, auxcount, atm, natm, bas, nbas, env, \
ao_loc, cintopt) \
private(ish, jsh, di, dj, i, j, i0, j0, shls, buf)
#pragma omp for nowait schedule(dynamic)
for (ish = auxstart; ish < nbasnaux; ish++) {
for (jsh = auxstart; jsh <= ish; jsh++) {
di = ao_loc[ish+1] - ao_loc[ish];
dj = ao_loc[jsh+1] - ao_loc[jsh];
shls[0] = ish;
shls[1] = jsh;
buf = (double *)malloc(sizeof(double) * di * dj);
if (cint2c2e_sph(buf, shls, atm, natm, bas, nbas, env,
cintopt)) {
for (i0 = ao_loc[ish]-ao_loc[auxstart], i = 0; i < di; i++, i0++) {
for (j0 = ao_loc[jsh]-ao_loc[auxstart], j = 0; j < dj; j++, j0++) {
eri[i0*naoaux+j0] = buf[j*di+i];
} }
} else {
for (i0 = ao_loc[ish]-ao_loc[auxstart];
i0 < ao_loc[ish+1]-ao_loc[auxstart]; i0++) {
for (j0 = ao_loc[jsh]-ao_loc[auxstart];
j0 < ao_loc[jsh+1]-ao_loc[auxstart]; j0++) {
eri[i0*naoaux+j0] = 0;
} }
}
free(buf);
} }
for (i = 0; i < naoaux; i++) {
for (j = 0; j < i; j++) {
eri[j*naoaux+i] = eri[i*naoaux+j];
}
}
free(ao_loc);
CINTdel_optimizer(&cintopt);
}
void run_all(int *atm, int natm, int *bas, int nbas, double *env)
{
int i, j, k, l, ij, kl;
int di, dj, dk, dl;
int kl_max;
int shls[4];
double *buf;
int *ishls = malloc(sizeof(int)*nbas*nbas);
int *jshls = malloc(sizeof(int)*nbas*nbas);
for (i = 0, ij = 0; i < nbas; i++) {
for (j = 0; j <= i; j++, ij++) {
ishls[ij] = i;
jshls[ij] = j;
}
}
int ncgto = CINTtot_cgto_spheric(bas, nbas);
printf("\tshells = %d, total cGTO = %d, total pGTO = %d\n",
nbas, ncgto,
CINTtot_pgto_spheric(bas, nbas));
int pct;
long count;
double time0, time1 = 0;
double tt, tot;
tot = (double)ncgto*ncgto*ncgto*ncgto/8;
time0 = omp_get_wtime();
printf("\tcint2e_sph with optimizer: total num ERI = %.2e\n", tot);
CINTOpt *opt = NULL;
cint2e_sph_optimizer(&opt, atm, natm, bas, nbas, env);
pct = 0; count = 0;
#pragma omp parallel default(none) \
shared(atm, natm, bas, nbas, env, ishls, jshls, opt, time0, pct, count, stdout) \
private(di, dj, dk, dl, i, j, k, l, ij, kl, kl_max, shls, buf, time1)
#pragma omp for nowait schedule(dynamic, 2)
for (ij = 0; ij < nbas*(nbas+1)/2; ij++) {
i = ishls[ij];
j = jshls[ij];
di = CINTcgto_spheric(i, bas);
dj = CINTcgto_spheric(j, bas);
// when ksh==ish, there exists k<i, so it's possible kl>ij
kl_max = (i+1)*(i+2)/2;
for (kl = 0; kl < kl_max; kl++) {
k = ishls[kl];
l = jshls[kl];
dk = CINTcgto_spheric(k, bas);
dl = CINTcgto_spheric(l, bas);
shls[0] = i;
shls[1] = j;
shls[2] = k;
shls[3] = l;
buf = malloc(sizeof(double) * di*dj*dk*dl);
cint2e_sph(buf, shls, atm, natm, bas, nbas, env, opt);
free(buf);
}
count += kl_max;
if (100l*count/((long)nbas*nbas*(nbas+1)*(nbas+2)/8) > pct) {
pct++;
time1 = omp_get_wtime();
printf("\t%d%%, CPU time = %8.2f\r", pct, time1-time0);
fflush(stdout);
}
}
time1 = omp_get_wtime();
tt = time1-time0;
printf("\t100%%, CPU time = %8.2f, %8.4f Mflops\n", tt, tot/1e6/tt);
CINTdel_optimizer(&opt);
free(ishls);
free(jshls);
}
int main()
{
int natm = 2;
int nbas = 4;
// ATM_SLOTS = 6; BAS_SLOTS = 8;
int *atm = malloc(sizeof(int) * natm * ATM_SLOTS);
int *bas = malloc(sizeof(int) * nbas * BAS_SLOTS);
double *env = malloc(sizeof(double) * 10000);
int i, n, off;
off = PTR_ENV_START; // = 20
i = 0;
atm[CHARGE_OF + ATM_SLOTS * i] = 1;
atm[PTR_COORD + ATM_SLOTS * i] = off;
env[off + 0] = 0; // x (Bohr)
env[off + 1] = 0; // y (Bohr)
env[off + 2] =-.8; // z (Bohr)
i++;
off += 3;
atm[CHARGE_OF + ATM_SLOTS * i] = 1;
atm[PTR_COORD + ATM_SLOTS * i] = off;
env[off + 0] = 0;
env[off + 1] = 0;
env[off + 2] =.8; // (Bohr)
i++;
off += 3;
n = 0;
/* basis #0 with kappa > 0 => p_1/2 */
bas[ATOM_OF + BAS_SLOTS * n] = 0;
bas[ANG_OF + BAS_SLOTS * n] = 0;
bas[NPRIM_OF + BAS_SLOTS * n] = 3;
bas[NCTR_OF + BAS_SLOTS * n] = 2;
bas[KAPPA_OF + BAS_SLOTS * n] = 1;
bas[PTR_EXP + BAS_SLOTS * n] = off;
env[off + 0] = 6.;
env[off + 1] = 2.;
env[off + 2] = .8;
off += 3;
bas[PTR_COEFF+ BAS_SLOTS * n] = off;
env[off + 0] = .7 * CINTgto_norm(bas[ANG_OF+BAS_SLOTS*n], env[bas[PTR_EXP+BAS_SLOTS*n]+0]);
env[off + 1] = .6 * CINTgto_norm(bas[ANG_OF+BAS_SLOTS*n], env[bas[PTR_EXP+BAS_SLOTS*n]+1]);
env[off + 2] = .5 * CINTgto_norm(bas[ANG_OF+BAS_SLOTS*n], env[bas[PTR_EXP+BAS_SLOTS*n]+2]);
env[off + 3] = .4 * CINTgto_norm(bas[ANG_OF+BAS_SLOTS*n], env[bas[PTR_EXP+BAS_SLOTS*n]+0]);
env[off + 4] = .3 * CINTgto_norm(bas[ANG_OF+BAS_SLOTS*n], env[bas[PTR_EXP+BAS_SLOTS*n]+1]);
env[off + 5] = .2 * CINTgto_norm(bas[ANG_OF+BAS_SLOTS*n], env[bas[PTR_EXP+BAS_SLOTS*n]+2]);
off += 6;
n++;
/* basis #1 with kappa = 0 => p_1/2, p_3/2 */
bas[ATOM_OF + BAS_SLOTS * n] = 0;
bas[ANG_OF + BAS_SLOTS * n] = 1;
bas[NPRIM_OF + BAS_SLOTS * n] = 1;
bas[NCTR_OF + BAS_SLOTS * n] = 1;
bas[KAPPA_OF + BAS_SLOTS * n] = 0;
bas[PTR_EXP + BAS_SLOTS * n] = off;
env[off + 0] = .9;
off += 1;
bas[PTR_COEFF+ BAS_SLOTS * n] = off;
env[off + 0] = 1. * CINTgto_norm(bas[ANG_OF+BAS_SLOTS*n], env[bas[PTR_EXP+BAS_SLOTS*n]]);
off += 1;
n++;
/* basis #2 == basis #0 */
bas[ATOM_OF + BAS_SLOTS * n] = 1;
bas[ANG_OF + BAS_SLOTS * n] = bas[ANG_OF + BAS_SLOTS * 0];
bas[NPRIM_OF + BAS_SLOTS * n] = bas[NPRIM_OF + BAS_SLOTS * 0];
bas[NCTR_OF + BAS_SLOTS * n] = bas[NCTR_OF + BAS_SLOTS * 0];
bas[KAPPA_OF + BAS_SLOTS * n] = bas[KAPPA_OF + BAS_SLOTS * 0];
bas[PTR_EXP + BAS_SLOTS * n] = bas[PTR_EXP + BAS_SLOTS * 0];
bas[PTR_COEFF+ BAS_SLOTS * n] = bas[PTR_COEFF+ BAS_SLOTS * 0];
n++;
/* basis #3 == basis #1 */
bas[ATOM_OF + BAS_SLOTS * n] = 1;
bas[ANG_OF + BAS_SLOTS * n] = bas[ANG_OF + BAS_SLOTS * 1];
bas[NPRIM_OF + BAS_SLOTS * n] = bas[NPRIM_OF + BAS_SLOTS * 1];
bas[NCTR_OF + BAS_SLOTS * n] = bas[NCTR_OF + BAS_SLOTS * 1];
bas[KAPPA_OF + BAS_SLOTS * n] = bas[KAPPA_OF + BAS_SLOTS * 1];
bas[PTR_EXP + BAS_SLOTS * n] = bas[PTR_EXP + BAS_SLOTS * 1];
bas[PTR_COEFF+ BAS_SLOTS * n] = bas[PTR_COEFF+ BAS_SLOTS * 1];
n++;
/*
* call one-electron spinor integrals
*/
int j, k, l;
int di, dj, dk, dl;
int shls[4];
double complex *buf;
i = 0; shls[0] = i; di = CINTcgto_spinor(i, bas);
j = 1; shls[1] = j; dj = CINTcgto_spinor(j, bas);
buf = malloc(sizeof(double complex) * di * dj);
if (0 != cint1e_spnucsp(buf, shls, atm, natm, bas, nbas, env)) {
printf("This integral is not 0.\n");
} else {
printf("This integral is 0.\n");
}
free(buf);
/*
* call two-electron spinor integrals
*/
i = 0; shls[0] = i; di = CINTcgto_spinor(i, bas);
j = 1; shls[1] = j; dj = CINTcgto_spinor(j, bas);
k = 2; shls[2] = k; dk = CINTcgto_spinor(k, bas);
l = 2; shls[3] = l; dl = CINTcgto_spinor(l, bas);
buf = malloc(sizeof(double complex) * di * dj * dk * dl);
if (0 != cint2e_spsp1(buf, shls, atm, natm, bas, nbas, env, NULL)) {
printf("This integral is not 0.\n");
} else {
printf("This integral is 0.\n");
}
free(buf);
CINTOpt *opt = NULL;
cint2e_spsp1_optimizer(&opt, atm, natm, bas, nbas, env);
i = 0; shls[0] = i; di = CINTcgto_spinor(i, bas);
j = 1; shls[1] = j; dj = CINTcgto_spinor(j, bas);
k = 2; shls[2] = k; dk = CINTcgto_spinor(k, bas);
l = 2; shls[3] = l; dl = CINTcgto_spinor(l, bas);
buf = malloc(sizeof(double complex) * di * dj * dk * dl);
if (0 != cint2e_spsp1(buf, shls, atm, natm, bas, nbas, env, opt)) {
printf("This integral is not 0.\n");
} else {
printf("This integral is 0.\n");
}
free(buf);
CINTdel_optimizer(&opt);
free(atm);
free(bas);
free(env);
}
