http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54402
--- Comment #17 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-12-06 17:00:35 UTC --- I don't see the #c0 testcase particularly slow, I'm afraid we won't get rid of all the slowdowns caused by the memory invalidation checking in var-tracking.c, that has been added for 4.8 for debug information accuracy (to avoid emitting wrong-debug, not just incomplete debug), but hope it can be speeded up at least a little bit. And as one comment in this PR and another one in another PR show, there is also a regression that LRA makes some functions that were previously without frame pointer now use forcefully frame pointer. It needs investigation whether that is just a bug somewhere, or intentional decision.
