This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/52571] vectorizer changes alignment of common symbols
- From: "mrs at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Thu, 10 May 2012 16:06:20 +0000
- Subject: [Bug tree-optimization/52571] vectorizer changes alignment of common symbols
- Auto-submitted: auto-generated
- References: <bug-52571-4@http.gcc.gnu.org/bugzilla/>
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=52571
--- Comment #14 from mrs at gcc dot gnu.org <mrs at gcc dot gnu.org> 2012-05-10 16:06:20 UTC ---
I'd disagree with that. I suspect 99% of them are not overridden, just like
99% of C functions are not weak and not overridden either. As for what should
be done, doesn't matter, as since the system places no constraint on selection
for the linker, there simply is nothing one can depend upon. I agree,
selecting the one with the maximal alignment would be nice, but, just be wary
this may cause one to try and make use of that property, and this just would
turn into a complex form of bug pushing or bug hiding.
Uncommoning things I think might well be slightly harder than trivial. A ctor
is allowed to dynamically load a shared library, that defines a symbol, maybe,
conditional on the environment. But with proper escape analysis, seems like
one can figure it all out, just be careful with the edge cases.