This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch] Predictive commoning, updated
- From: "Diego Novillo" <dnovillo at google dot com>
- To: "Zdenek Dvorak" <rakdver at kam dot mff dot cuni dot cz>
- Cc: gcc-patches at gcc dot gnu dot org, richard dot guenther at gmail dot com
- Date: Wed, 23 May 2007 11:34:46 -0700
- Subject: Re: [patch] Predictive commoning, updated
- References: <20070214222919.GA29894@atrey.karlin.mff.cuni.cz> <462E3485.9090809@redhat.com> <20070504115009.GA2215@kam.mff.cuni.cz>
On 5/4/07, Zdenek Dvorak <rakdver@kam.mff.cuni.cz> wrote:
regarding the array indexed by SSA versions -- this does not seem like a good choice,
since only a few ssa names are usually inserted to the table, so
allocating and deallocating it would consume too much time. Pointer
maps were not available when the patch was written; I have rewritten the
patch to use them (although it does not simplify it significantly).
OK.
> Ouch. Why not just store BSIs in the chains?
Since bsi_for_stmt will be a constant-time operation once gimple tupples
are implemented, I do not think it is worth complicating the patch with
this.
Well, we are still a ways away from having tuples integrated. But if
this code is not hot enough, then I don't mind keeping this
implementation.
I don't have other concerns with the patch. You are now only waiting
for Sebastian's review, correct? If not, then I'm OK with checking
the patch in.