This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: CFG branch merge 5 - remove PROP_ALLOW_CFG_CHANGES
> On Thu, Feb 28, 2002 at 09:31:14PM +0100, Jan Hubicka wrote:
> > Would this be OK if I mplement the loop at cfg_cleanup time?
>
> I don't thinnk I know quite what you're suggesting.
> Show me the patch.
>
> > I just felt that the optimization is bit expensive given the benefit
> > it brings. What kind of problems it brought to pa except the pesimized
> > code?
>
> My memory is apparently faulty. I thought it was submitted
> to address some problem other than performance, but no.
>
> http://gcc.gnu.org/ml/gcc-patches/2001-07/msg01832.html
>
> Although I do seem to recall problems in, e.g. if_convert
> dealing with empty basic blocks, I can't find anything
> more specific.
I see. If we want to make csecleanup to enter cfg into state w/o
obvious mistakes, we need to loop. What I am thinking about is modifying
the update_life_info/update_life_info_in_dirty_blocks to return whether
dead code removal happent and then make cfg_cleanup loop in
try_optimize_cfg and update_life_info_in_dirty_blocks until both returns
false.
Honza
>
>
> r~