Re: C++ compile-time performance regressions / PR 3083

[Mark Mitchell <mark at codesourcery dot com>]

--On Wednesday, October 23, 2002 05:53:20 PM +0200 Gerald Pfeifer 
<pfeifer@dbai.tuwien.ac.at> wrote:

> General note: We have a significant performance problem with optimizing
> C++ compilations.  I did some new tests using the example from PR 3083,
> and here are the timings (i386-unknown-freebsd4.6):
>
>   /sw/gcc-3.0.4/bin/g++    45.45 user
>   /sw/gcc-3.1.1/bin/g++    57.79 user  +27%
>   /sw/gcc-3.2.1/bin/g++    59.30 user  +30%
>   /sw/gcc-current/bin/g++  81.18 user  +78%
>
> Could someone please have a look at this?
>
>
> Mark, I agree that this should not block the release of GCC 3.2.1,
> but I really, really would like to reopen PR 3083 as "high priority"
> for GCC 3.3 (removing you as "responsible").
Make a new PR, and mark it high priority with the "[mainline regression]"
flag.  See if you can look at the profile and see where the time is
going on the mainline that it's not on the branch.  (Make sure you
--disable-checking of course.)

--
Mark Mitchell                mark@codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com