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);
}
/*
* tot. contracted atomic spheric GTOs in a shell
*/
FINT cinttot_cgto_spheric_(const FINT *bas, const FINT *nbas)
{
return CINTtot_cgto_spheric(bas, *nbas);
}