This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Optimization: RedHat vs. GCC v3.0x?
- To: geoffk at redhat dot com
- Subject: Re: Optimization: RedHat vs. GCC v3.0x?
- From: Joe Buck <jbuck at synopsys dot COM>
- Date: Tue, 2 Oct 2001 09:09:35 -0700 (PDT)
- Cc: swsnyder at home dot com, gcc at gcc dot gnu dot org
[ comparisons of "2.96" with FSF releases ]
> >From that web page:
>
> * gcc 2.96 is actually more standards compliant than any other version
> of gcc released at the time Red Hat made this decision
> ...
> * gcc 2.96 generates better, more optimized code
>
> I think bero meant 'more optimized code than any other version of gcc
> released at the time Red Hat made this decision', which was around the
> middle of last year.
For C, the "better, more optimized code" claim was valid (compared with
2.95.2), especially on x86. For C++, it definitely was not; because
ADDRESSOF was broken in 2.96-RH, C++ code produced by the original 2.96-RH was
substantially *worse* than 2.95.2, and the Stepanov benchmark score was
atrocious. Pretty much any program using STL was substantially slower
on 2.96-RH. This bug was later fixed, but a very long period of time went
by before it was.