https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122663
--- Comment #11 from Andrew Pinski <pinskia at gcc dot gnu.org> --- (In reply to Andrew Pinski from comment #9) > (In reply to Richard Biener from comment #7) > > this shows that the way to discover the "unused leaf" does not handle cycles > > properly, as would simple_dce_from_worklist. We have > > > > m_9 = _7 | m_8; > > ... > > # m_8 = PHI <2(4), m_9(5)> > > > > here. > > yes simple_dce_from_worklist only handles the simpliest cycle case (the ssa > name used in its own phi definition which was added with > r14-712-g3e8f8ab5237002). Expanding it to handle the one statement case > might be fast enough still (I wonder if this is now missed in ccp1 after > r14-569-g21e2ef2dc25de3; I didn't see it at the time though). Filed pR 122678 for that. There are more cylces it does not handle than that; but I am not sure we want to cause a huge slow down on cycle detection :).
