This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch] tree-ssa-phiopt.c: Update a comment about the pass
- From: Jeffrey A Law <law at redhat dot com>
- To: Kazu Hirata <kazu at cs dot umass dot edu>
- Cc: gcc-patches at gcc dot gnu dot org, pinskia at physics dot uc dot edu
- Date: Wed, 20 Apr 2005 08:52:00 -0600
- Subject: Re: [patch] tree-ssa-phiopt.c: Update a comment about the pass
- Organization: Red Hat, Inc
- References: <20050419.204548.05624837.kazu@cs.umass.edu>
- Reply-to: law at redhat dot com
On Tue, 2005-04-19 at 20:45 -0400, Kazu Hirata wrote:
> Hi,
>
> Attached is a patch to update a comment about the pass to reflect the
> way PHI-OPT currently operates in.
>
> I've removed the following sentence
>
> bb1 will become unreachable and bb0 and bb2 will almost always be
> merged into a single block.
>
> because I don't think this is true at least without copy propagation,
> DCE, and CFG cleanup.
The statement is true -- the statement does not say when the blocks
will be merged, just that they will be merged (by later optimizations).
The right thing to do would have been to update the comment to reflect
that later optimizations will eventually result in bb0 and bb2 being
merged.
> Note that the current PHI-OPT does not clean up
> a degenerate PHI node of the form
>
> x = PHI <x' (bb0)>;
>
> that it may create as a result of transformations.
Why not? That's the whole point of PHI-OPT. If it's no longer
performing that optimization, then something as probably gone
wrong.
>
> Built cc1. Committed as obvious.
I would rather you not commit this kind of comment change as obvious
since it's far from obvious IMHO.
jeff