This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Testsuite Results for gcc-2.96 19990628
- To: ronis at onsager dot chem dot mcgill dot ca
- Subject: Re: Testsuite Results for gcc-2.96 19990628
- From: Joe Buck <jbuck at yamato dot synopsys dot com>
- Date: Wed, 30 Jun 99 8:29:01 PDT
- Cc: kunert at physik dot tu-dresden dot de, egcs at egcs dot cygnus dot com
> Thanks for the reply. I think you've got it. I build with lots of
> optimization flags turned on for the stage1 and stage2 steps, specifically:
>
> -O3 -march=pentium -ffast-math -funroll-loops -fomit-frame-pointer -fforce-mem
> -fforce-addr -malign-double -malign-jumps=2 -malign-loops=2 -malign-functions=2
>
> Of course, if the compiler generates bad code when building itself, then why
> should we trust it for anything else with optimization turned on. This is
> clearly a problem.
Rather, clearly we can't trust the compiler on your platform with the
choice of optimization flags you gave. Compiling the compiler with
the default flags is safest, because that's the well-tested path.
The rule in large software systems is "if it hasn't been tested, it
probably doesn't work".