This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug c++/15837] Optimization (of templatized functions?)
- From: "adruab at voria dot com" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 6 Jun 2004 06:22:22 -0000
- Subject: [Bug c++/15837] Optimization (of templatized functions?)
- References: <20040605201151.15837.adruab@voria.com>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Additional Comments From adruab at voria dot com 2004-06-06 06:22 -------
Ok, I understand that. However, if it is undefined why does gcc not give an
error? (sorry I missed your last post as I was adding the attachment).
Also, the pen class was originally designed by someone else in such a way to
allow it to work with non standard compiler packing structures (although that
was not accomplished as you'd also need a byte stride for the x dimension to do
that effectively). I.E. you could refer to a member of a struct which was
defined to align on a 2 byte boundary and still be able to jump from one element
to the other. Perhaps this is an unnecessary feature, and I suppose it could be
modified to work on a strict element basis rather than byte alignment. However,
Is any way to accomplish this besides the method used?
The only thing I can think of to satisfy this constraint is to be more explicit
(i.e. rather than ((char *&)p)+= stride; do p = (value_type*)((char*)p +
stride); ). Would that not work as well? I'll try that and see....
Thanks for the help though. Please excuse me if I seem a bit irritable, this is
the weirdest thing I have ever run into :).
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=15837