https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55157
--- Comment #9 from Aldy Hernandez <aldyh at gcc dot gnu.org> --- (In reply to Aldy Hernandez from comment #7) > (In reply to Andrew Macleod from comment #6) > > (In reply to Aldy Hernandez from comment #4) > > 3) It also seems to me that you then only need to add the zero/union iff the > > trailing bit has zeros. ie, if the are no trailing zeros, then just set the > > lb to 0, and calculate the UB based on the clz. > > That actually made it slightly worse. I did some more testing. Your suggestions actually improved the original code from 3.3% to 2.5%, but I added: + if (!wi::neg_p (mask, TYPE_SIGN (type ())) && prec > 1) instead of: + if (TYPE_UNSIGNED (type ()) && prec > 1) because we can still perform the optimization for signed positive numbers. And that brought the slowdown back to 3.6%. But ISTM that we're gonna do it, might as well do it for signed numbers as well. Hmmmm... I think we really should do our best to represent masks as ranges as it makes other optimizations possible across the ranger ecosystem. But I'm not ecstatic about the 3.6% drop (even though overall compilation is unaffected).
