Closed mmlanger closed 8 years ago
If I'm not mistaken, the algorithm actually needs 1/facmin. One could save another division if instead of facmin directly the inverse 1/facmin is computed and saved in the facmin_table
.
calc_h_opt uses facmin and 1/facmin, so both have to be known, we would have to table-ize both... The stepper doesn't seem to be micro optimized in that way and I don't think it would make much sense since the general algorithm is more computation heavy.
Ah ok, right. I overlooked the appearance of facmin. Indeed, if both facmin and 1/facmin are used there is no need for further optimization here.
Thanks a lot for this improvement!
Implements the change suggested in Issue #185.