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: ifcvt/crossjump patch: Fix PR 42496, 21803


On Wed, Jul 28, 2010 at 00:18, Eric Botcazou <ebotcazou@adacore.com> wrote:
>> That's a non-argument, and false IMO. ?Not every optimization can be
>> represented cleanly as a set of CFG manipulations, and this one can't.
>
> I wrote "this kind of transformations", not "all transformations".
>
> What about the algorithm sketched by Paolo?

That's what Bernd referred to when he said, "I think we can also
discard the suggestion of simulating the effect of a single
reorder_insns call with a series of complex CFG transformations, as
that seems entirely pointless."

I actually agree with him.  I don't think it is _that_ complex
(particularly because my sketch did more than a single reorder_insns),
but I agree it is pointless.  It is faking that head merging is a pure
CFG transformation when in fact it isn't.

Paolo


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