This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/16962] loop unrolling with x86-64 asm not efficient
- From: "falk at debian dot org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: 24 Aug 2004 21:06:51 -0000
- Subject: [Bug tree-optimization/16962] loop unrolling with x86-64 asm not efficient
- References: <20040810134244.16962.tomstdenis@iahu.ca>
- Reply-to: gcc-bugzilla at gcc dot gnu dot org
------- Additional Comments From falk at debian dot org 2004-08-24 21:06 -------
Looking more closely, my Alpha example actually generates reasonable code
now. As to your assembly: I can't really read i386 assembly, but it seems it
modifies something without telling gcc, so it is invalid. To improve
performance, you should avoid (%x) and use a "m" constraint, and not clobber
hard regs but use "=r" on temp vars. Something like:
asm("..." : "=m"(*out) :"r"(x), "m"(*y));
With this, I get entirely reasonable code on the lno-branch, so closing.
--
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |RESOLVED
Resolution| |INVALID
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16962