--- Comment #5 from rguenth at gcc dot gnu dot org 2007-10-23 14:26 ---
This is wrong alias info. We hoist
# VUSE
p_42 = result.pDirty;
out of the loop. It looks like we don't have VUSEs of a and MPT.47 here like
# VUSE
p_38 = p_35->pDirty;
because results address is only
--- Comment #4 from hjl at lucon dot org 2007-10-23 14:24 ---
Revision 127629
http://gcc.gnu.org/ml/gcc-cvs/2007-08/msg00523.html
introduced this regression.
--
hjl at lucon dot org changed:
What|Removed |Added
---
--- Comment #3 from hjl at lucon dot org 2007-10-23 13:54 ---
I started regression hunt. It is introduced between 127611 and 127769.
--
hjl at lucon dot org changed:
What|Removed |Added
--
--- Comment #2 from rguenth at gcc dot gnu dot org 2007-10-23 13:48 ---
Janis, can you hunt this? (in case it also fails for you that is)
--
rguenth at gcc dot gnu dot org changed:
What|Removed |Added
--
--- Comment #1 from rguenth at gcc dot gnu dot org 2007-10-23 13:16 ---
Also fails with -O --param max-aliased-vops=500 (the -O2 setting)
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33870
