This is the mail archive of the
mailing list for the GCC project.
Re: [tree-ssa] Removing redundant loads
- From: Diego Novillo <dnovillo at redhat dot com>
- To: Daniel Berlin <dberlin at dberlin dot org>
- Cc: Jeff Law <law at redhat dot com>, "gcc-patches at gcc dot gnu dot org" <gcc-patches at gcc dot gnu dot org>
- Date: 29 May 2003 17:21:08 -0400
- Subject: Re: [tree-ssa] Removing redundant loads
- Organization: Red Hat Canada
- References: <5AC53EC8-921A-11D7-A362-000A95A34564@dberlin.org>
On Thu, 2003-05-29 at 17:13, Daniel Berlin wrote:
> On Thursday, May 29, 2003, at 04:38 PM, firstname.lastname@example.org wrote:
> > I mentioned at the GCC summit, I had a nice one line patch which
> > allowed
> > tree-ssa to remove redundant loads at basically no cost.
> And this patch rapes the ability of load PRE to work.
Well, as with anything done in the renaming pass, these tricks will only
catch total redundancies. So, yes, it may hinder PRE. But don't
despair, nothing is written in stone here.
What we should do is make these transformations conditional as any other
transformation. Later on, when PRE is finished, we can start playing
the flags game with SPEC or what-have-you.
My idea is to put in -O2 all the options that help SPEC or any other
commonly used hunk of code. It may well happen that we end up using PRE
only. But for now, let's just get these things implemented. As the
number of optimizations grows, we can start combining them in various