[vta, graphite, PR41888, PR41886] get graphite to work with VTA

[Alexandre Oliva]

On Nov  8, 2009, Sebastian Pop <sebpop@gmail.com> wrote:

> Hi Alexandre,
> Thanks for fixing this.  I have some comments about the patch.  You
> could hoist out of the sese_bad_liveouts_use function this condition:

> +      || bb_in_sese_p (bb, region))

> and put it as an initial filter in sese_build_liveouts:

>   if (MAY_HAVE_DEBUG_INSNS)
>     FOR_EACH_BB (bb)
>       if (!bb_in_sese_p (bb, region))
>         sese_reset_debug_liveouts_bb (region, liveouts, bb);

Indeed.  This is just as possible for the original code that I copied
and simplified to detect variables used only in deug stmts.  I'd rather
not make the two sets of functions diverge unnecessarily with this
patch.

How about I check this patch in, and let someone more familiar with
graphite undertake the implementation of the proposed optimization in
both sets of functions?

> That makes me wonder if you could walk the stmts of the REGION looking
> for uses escaping outside the REGION, instead of walking all the BBs
> of the function searching for the definitions that fall into the
> REGION.

That sounds like a good plan to me.

-- 
Alexandre Oliva, freedom fighter    http://FSFLA.org/~lxoliva/
You must be the change you wish to see in the world. -- Gandhi
Be Free! -- http://FSFLA.org/   FSF Latin America board member
Free Software Evangelist      Red Hat Brazil Compiler Engineer