Let's go with 15M; it's possible that other polys will behave differently, and the "best" curve is rather flat from 13M to 20M. Looks like this lowers our rels_wanted by nearly 20%!!
Thanks very much for exploring this parameter! I've explored best starting-Q only for C100-C120 sized jobs, and best was small indeed (under 100k). I'll try boosting starting-Q when I resume C140-sized work later this month.
|