This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/32921] [4.3/4.4 Regression] Revision 126326 causes 12% slowdown
- From: "rguenth at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 2 May 2008 10:28:59 -0000
- Subject: [Bug tree-optimization/32921] [4.3/4.4 Regression] Revision 126326 causes 12% slowdown
- References: <bug-32921-682@http.gcc.gnu.org/bugzilla/>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Comment #48 from rguenth at gcc dot gnu dot org 2008-05-02 10:28 -------
Btw, on x86_64 leslie3d performance is now above that from before r126326.
The differences you mention can be seen on x86_64 as well, but they are
not related to aliasing or partitioning but due to differences in what
IVOPTs produces.
The first difference at all on the tree level is with mergephi2 that is able
to remove a single forwarder BB if --param max-aliased-vops=10000 is _not_
specified. First real changes happen when DOM is able to CSE some loads
with --param max-aliased-vops=10000 but not without:
- D.1747_808 = D.1747_763;
- D.1748_809 = D.1748_764;
- D.1749_811 = D.1749_766;
- D.1750_812 = D.1749_766 * D.1644_799;
+ D.1747_808 = pav.data;
+ D.1748_809 = (real(kind=8)[0:] *) D.1747_808;
+ D.1749_811 = pav.dim[0].stride;
+ D.1750_812 = D.1644_799 * D.1749_811;
and as PRE itself is not alias-oracle aware as well further missed
optimizations
occur.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32921