[... if __predict_false(ip_nmbclusters != nmbclusters) ...]

Hmm. OK, thanks for that idea. I will add a macro to do the test
inline, and add a function (ip_nmbcluster_recalc()?)  to recompute
all P-related parameters which are calculated from nmbclusters.

(I have be adding more parameters, fairly soon, and I'd want to use
the test in 2 or 3 places already.)


thanks again!