> On Sep 19, 2017, at 6:20 PM, Alexander Monakov <amona...@ispras.ru> wrote: > >> I'd like to keep read/write processing balanced. In the above "read" >> analysis >> has greater weight than "write" analysis. Also, autopref_rank_data() should >> not be called if !rtx_equal_p (data1->base, data2->base). > > I'm afraid this doesn't work. Consider you have insns A, B, C such that all > are relevant (so that irr1, irr2 stay 0), A and C have the same base and > compare such that C < A, B has a different base and therefore according > to this subcomparator you can have A == B == C < A and with following > tiebreakers A < B < C < A. > > In other words, if your comparison tries to subdivide input set into > equivalence > classes (here you want to introduce equivalence classes based on 'base'), then > you should introduce ordering between classes (i.e. for two items with a > different 'base' say if one is greater than the other).
Right. Let me meditate for couple of days on this :-) . -- Maxim Kuvyrkov www.linaro.org
