https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87047
--- Comment #8 from Alexander Monakov <amonakov at gcc dot gnu.org> --- (In reply to Jakub Jelinek from comment #7) > Treating the missing else cost as 0 cost isn't right either, it is ok to > accept some small cost, otherwise we wouldn't do the if conversion ever with > missing else_bb. > The thing is that it needs to be small enough. > Do we need some new target magic constant for this, or do we just use cost > of a few (say 4-5) moves or something similar? Well, original_costs is already initialized with COSTS_N_INSNS (2), accounting for test and branch that would be removed. So it's not too bad: in pr78120.c else_bb is in fact missing. (on x86 I think we shouldn't encourage if-conversion, by adding data dependencies it can lengthen critical paths)
