Table 3. Relative Timings (in s) for the Evaluation of All Molecular Gradients and NACs for Two Compounds: Dy–bbpenCl (64 Atoms) and Co(acac)2 (35 Atoms)a.
Basis Set | MCLR (step 1) | MCLR (step 2) | ALASKA | Total (original) | Total (two-step) | Speedup (%) |
---|---|---|---|---|---|---|
Dy-MB, 250 bf | 604 | 163 | 2575 | 734712 | 633445 | 13.8 |
Dy-VDZP, 478 bf | 5240 | 688 | 9405 | 3384522 | 2338250 | 30.9 |
Dy-VTZP, 850 bf | 56691 | 4927 | 72861 | 29938122 | 18037329 | 39.8 |
Co-MB, 116 bf | 17 | 8 | 242 | 14278 | 13800 | 3.3 |
Co-VDZP, 286 bf | 263 | 34 | 1120 | 76274 | 63942 | 16.2 |
Co-VTZP, 601 bf | 3751 | 236 | 6217 | 550590 | 361016 | 34.4 |
Dy–bbpenCl was computed with CAS(9,7), 21 roots S = 5/2 were optimized; Co(acac)2 was computed with CAS(7,5), 10 roots S = 3/2 were considered. The MCLR and ALASKA columns refer to the calculation of a single gradient or NAC vector, while the “Total” columns refer to the total time for all N(N + 1)/2 of them (N = 21, 10). The expected saving is approximately N(N + 1)/2 times the “step 1” column.