This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] PR 32475 stack setup code being deleted.
On 7/4/07, Paolo Bonzini <paolo.bonzini@lu.unisi.ch> wrote:
> I'd like to avoid munging the live bitmaps for various reasons:
> 1) munging the live vector doesn't give us the correct reg chain.
The question is whether we want the refs to appear in the reg chain? If
you have a good reason why we do, your patch is ok. Otherwise, I'm not
yet convinced.
Paolo
I don't see any reason not to keep the chain correct.
Also, note that this will create less refs than Kenny's proposed patch,
as there are presumably less insns that update SP
than MEM_LOAD/MEM_STORE that references hard frame pointer.
Also, munging the live bitmap has to happen everywhere
that manipulates the bitmap - the users of live/lr/urec
has to pay attention to sp whenever they update the bitmap themselves.
That's a lot of places and hence a lot more fragile.
This patch is the smallest and the most self-contained change I can think of.
However, if there's a better proposal, I'm all ears.
--
#pragma ident "Seongbae Park, compiler, http://seongbae.blogspot.com";