Re: 25% memory requirement regression since yesterday

[Mark Mitchell <mark at codesourcery dot com>]

Richard Guenther wrote:

> On Mon, 12 Jul 2004, Mark Mitchell wrote:
>
>  
>
> > Richard Guenther wrote:
> >
> >    
> >
> > > The tramp3d tracker caught a memory requirement regression for C++ on
> > > mainline inbetween (excluding)
> > >
> > > 2004-07-11  Mark Mitchell  <mark@codesourcery.com>
> > >
> > >       * cp-tree.h (saved_scope): Make old_bindings a vector.
> > >       (unuse_fields): Remove.
> > >       * name-lookup.h (cxx_saved_binding): Define it.
> > >       * class.c (pushclass): Don't use unuse_fields.
> > >       * name-lookup.c (cxx_saved_binding_make): Remove.
> > >       (store_binding): Add new bindings to a vector, using an
> > >       accumulator style, rather than adding them to a list.
> > >       (store_bindings): Adjust accordingly.
> > >       (store_class_bindings): Likewise.
> > >       (push_to_top_level): Likewise.
> > >       [...]
> > >
> > > also compile-time regressed somewhat.  Have a look at
> > > http://www.tat.physik.uni-tuebingen.de/~rguenth/gcc/monitor-summary.html
> > > for both detailled graphs and the testcase.
> > >
> > >
> > >      
> > I'll look into that.  Do you have data about compile-times, or just
> > about memory usage?
> >    
>
> I have both, but for compile time, this is only 3m52s -> 3m56s regression.
> Tomorrow we will see if this was just noise.
I am seeing a small memory *reduction* with that patch; from 385M to 
380M on an IA32 GNU/Linux machine, although I did not measure with your 
script.  I will do that now.

Can you confirm that is my patch that caused this problem?

--
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com