http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55478
--- Comment #3 from Jan Hubicka <hubicka at gcc dot gnu.org> 2012-11-27 10:55:04 UTC --- Created attachment 28792 --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28792 patch This is patch I am going to test. It copies open64' notion of heavy calls - i.e. the calls that are not hot, but inlining is still importnat because callee takes considerable time. We had few examples where this is profitable thing, so I guess it is good idea to implement this. The heaviness should be propagated by ipa-profile that is the bit I did not implement for simplicity, so we now consider pretty much every nonleaf function or function with loop as possibly heavy.
