Re: [PING][PATCH][REVISED] Fix PR middle-end/PR28690, modify swap_commutative_operands_p

["H.J. Lu" <hjl at lucon dot org>]

On Sun, Jul 15, 2007 at 11:20:13PM -0500, Peter Bergner wrote:

> > However, this patch causes a performance regresion on x86. It fails
> > gcc.target/i386/387-11.c. Before the patch:
> 
> I'm a little confused here.  What source code base did you apply the
> patch you posted to?  Plain mainline?  Mainline with my submitted (and
> approved but not yet committed) patch?  Mainline plus some other
> modified patch you came up with?  It's hard to evaluate what could
> be going wrong without knowing exactly what you're testing.
> 
> You never mentioned any testsuite failures before, so I assume that
> testcase passes using mainline + my patch?  I didn't hit any testsuite
> failures while testing on x86_64.
> 

I checked my "make check" results. I saw

FAIL: gcc.target/i386/387-11.c scan-assembler-not fchs

on gcc with your patch:

http://gcc.gnu.org/ml/gcc-patches/2007-06/msg01493.html

Can you show your x86-64 assembly output of gcc.target/i386/387-11.c?


H.J.