[Bug tree-optimization/85964] [8/9 Regression] compile time hog w/ -O3 -ftracer -fno-guess-branch-probability

rguenth at gcc dot gnu.org gcc-bugzilla@gcc.gnu.org
Wed May 30 14:58:00 GMT 2018


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85964

--- Comment #11 from Richard Biener <rguenth at gcc dot gnu.org> ---
The CFG cleanup hog is gone on trunk now - an artificial testcase triggering it
would need to present us with a large number of independently discoverable
branches like maybe

 if (foo)
   {
     tem = 0;
     if (tem)
       { sth }
     else
       { sth }
   }
 ... repeat above ...

but eventually a more elaborate CFG is required to make fixing up dominators
iteratively blow up.  The above would trigger in the CFG cleanup after the
very first CCP pass then.

Otherwise I'll put the RFC pach on hold because it comes at a cost (or
rather I didn't do any measurements on regular code).


More information about the Gcc-bugs mailing list