This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [3.4-bib] SSE moves patch
- From: Richard Henderson <rth at redhat dot com>
- To: Jan Hubicka <jh at suse dot cz>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Sun, 6 Oct 2002 13:30:31 -0700
- Subject: Re: [3.4-bib] SSE moves patch
- References: <20021006185408.GB21339@kam.mff.cuni.cz>
On Sun, Oct 06, 2002 at 08:54:08PM +0200, Jan Hubicka wrote:
> case 5:
> - if (TARGET_SSE2 && !TARGET_ATHLON)
> - return "pxor\t%0, %0";
> - else
> - return "xorps\t%0, %0";
> + return "xorps\t%0, %0";
This is wrong. On P4, pxor avoids the false dependency on
the previous contents of the register while xorp[sd] don't.
I'd be nice to have a pass that recognized when the stalls
would be possible and add the clears, rather than adding
them all the time. This would be useful for the integer
registers as well -- avoid the use of movzbl when not needed.
r~