This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Put scope blocks on a diet


On Oct 12, 2007 7:13 AM, Jan Hubicka <hubicka@ucw.cz> wrote:
> > On Oct 11, 2007, Jan Hubicka <jh@suse.cz> wrote:
> >
> > >> +              /* Removing declarations before inlining is going to affect
> > >> +                 DECL_UID that in turn is going to affect hashtables and
> > >> +                 code generation.  */
> > >> +              || !cfun->after_inlining)
> >
> > > Actually I think your proposed patch (modulo the asserts) is identical
> > > to one I sent in August, or did I missed something trivial?
> >
> > The difference is that my patch sets "unused = false;" when we refrain
> > from removing an unused variable for the sake of debug info.
> >
> > Without it, we may remove a leaf block that still contains
> > declarations, and this causes differences in declaration numbers for
> > inlining, although I don't have a testcase that shows the problem
> > except in the vta branch.
>
> Uh, thanks, I didin't noticed that.
> Honza
>
> >
> > > I tought that one is already on your branch and just waiting for
> > > approval to mainline.
> >
> > Yes, but I'm now convinced it is wrong, for it will unintentionally
> > remove declarations of unused variables before inlining by removing
> > their containing blocks, and this may result in different code *and*
> > harm debug info.

With

2007-11-26  Alexandre Oliva  <aoliva@redhat.com>, Jan Hubicka  <jh@suse.cz>

        * tree-ssa-live.c (remove_unused_scope_block_p): Drop
        declarations and blocks only after inlining.  Check that
        non-empty blocks are not dropped.
        * tree-inline.c (expand_call_inline): Check that functions are
        not inlined too late.

there is a 50% increase in memory consumption and a 5% increase in
compile-time for tramp3d.

I don't think this is acceptable.  Please revert the patch given that we are
late in stage3.

Thanks,
Richard.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]