https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113533
--- Comment #17 from Oleg Endo <olegendo at gcc dot gnu.org> --- (In reply to Jeffrey A. Law from comment #16) > Note that Jakub recently twiddled fwprop to throttle it back a little. As a > result the regressions seen in this BZ are no longer an issue. I'm going to > remove the regression marker. > > Roger/Oleg, if y'all are going to continue on the costing work, we may want > to keep this open. Otherwise I wouldn't lose any sleep if we just closed it. Thanks for the update on this, Jeff. Even if the regression disappeared, the cost calculations for memory accesses are clearly wrong in that spot. So I'd like to at least fix the core issues. Just haven't had time to test it more thoroughly for any side-effects yet.
