This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH 1/4] Add flag -ftree-loop-if-convert-memory-writes.
- From: Sebastian Pop <sebpop at gmail dot com>
- To: Richard Guenther <rguenther at suse dot de>
- Cc: gcc-patches at gcc dot gnu dot org, matz at suse dot de
- Date: Fri, 13 Aug 2010 17:10:49 -0500
- Subject: Re: [PATCH 1/4] Add flag -ftree-loop-if-convert-memory-writes.
- References: <1278625285-12667-1-git-send-email-sebpop@gmail.com> <1278625285-12667-2-git-send-email-sebpop@gmail.com> <alpine.LNX.2.00.1008131039380.25856@zhemvz.fhfr.qr>
On Fri, Aug 13, 2010 at 03:54, Richard Guenther <rguenther@suse.de> wrote:
> So you create this series of stmts in place of the conditional store,
> which means I don't see why materializing the condition before
> the cond-expr of the dominator does not work.
>
Yes, this could work. Inserting the predicate computations anywhere
before the predicated BB is fine. The reason why we have to insert
the predicates before or at the beginning of the BB in the case of mem
writes if-conversion is that we need the predicates to build the
conditional move statements.
For scalar if-conversion, we need the predicates as close as possible
to the phi nodes that merge the scalar values from different branches
in order to have low register pressure.
Sebastian