This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: 3.0-pre vs 3.0 on Haney speed
- To: pcarlini at unitus dot it
- Subject: Re: 3.0-pre vs 3.0 on Haney speed
- From: Joe Buck <jbuck at synopsys dot COM>
- Date: Sat, 23 Jun 2001 15:32:00 -0700 (PDT)
- Cc: gcc at gcc dot gnu dot org
> Comparing Gcc-3.0 and Gcc-3.0 pre 20010611 on the Haney_speed tests (-O2
> -funroll-loops) I'm seeing a puzzling result (Linux,PII-400):
>
> gcc version 3.0
> ---------------
> Timing Complex matrix operations for n = 200
> Language Answer Check Time Iter Time
> O-O C++ (-1.06982e+07,-2.82516e+08): 18.07 0.7228
> Hand-coded C (-1.06982e+07,-2.82516e+08): 13.46 0.5384
>
> gcc version 3.0 20010611 (prerelease)
> -------------------------------------
> Timing Complex matrix operations for n = 200
> Language Answer Check Time Iter Time
> O-O C++ (-1.06982e+07,-2.82516e+08): 13 0.52
> Hand-coded C (-1.06982e+07,-2.82516e+08): 12.81 0.5124
>
> It looks like something "wrong" happened to C++ optimization during the
> last days before the release.
To make progress, it would help to see the assembly language for a
representative inner loop to see what's going wrong.