This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [tree-ssa] Merging with bnw-simple
- From: Daniel Berlin <dberlin at dberlin dot org>
- To: Diego Novillo <dnovillo at redhat dot com>
- Cc: gcc at gcc dot gnu dot org, Jason Merrill <jason at redhat dot com>,"Frank Ch. Eigler" <fche at redhat dot com>,Steven Bosscher <s dot bosscher at student dot tudelft dot nl>
- Date: Tue, 24 Sep 2002 13:49:38 -0400
- Subject: Re: [tree-ssa] Merging with bnw-simple
On Tuesday, September 24, 2002, at 10:02 AM, Diego Novillo wrote:
Now that we have a working version of the re-designed tree IR, we
have to think about merging it with tree-ssa. I'd like opinions
on how we should proceed:
1- Merge the changes into tree-ssa-branch directly. This will
likely break all the current tree optimizers and mudflap.
I forget again, can you point me to the message describing the IR?
I'm trying to determine how significant the changes are from the
perspective of "Do things look basically the same".
IE do we still have the nice properties of SIMPLE, like a = 5 + b + c
being simplified
down into T.1 = 5 + b; a = T.1 + c; (or whatever).
If so, I'm all for just merging it now.
I'm working on making SSA-PRE update the SSA representation, so i've
got something to do on the current branch, in it's current state, while
everything is fixed (Or I could help fix if you like), and just update,
since I would imagine if you keep SSA working, SSA-PRE won't break, as
the only property of SIMPLE it depends on is the fact that we have
guarantees on what can appear in a binary operation, etc.
If these types of restrictions have loosened significantly, well, it's
gonna be painful.
--Dan