This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: PRE in gcse.c
- To: law at cygnus dot com
- Subject: Re: PRE in gcse.c
- From: "David S. Miller" <davem at dm dot cobaltmicro dot com>
- Date: Sat, 7 Nov 1998 05:43:10 -0800
- CC: alla at lysator dot liu dot se, egcs at cygnus dot com
- References: <30159.910411039@upchuck>
Date: Fri, 06 Nov 1998 20:57:19 -0700
From: Jeffrey A Law <law@cygnus.com>
In message <199811061539.QAA07175@c125.ryd.student.liu.se>you write:
> In the source i did find a comment that said:
> Eventually it would be nice to replace cse.c/gcse.c with SSA (static
> single
> assignment) based GVN (global value numbering). L. T. Simpson's paper
> (Rice University) on value numbering is a useful reference for this.
SSA is further out, but we are putting the pieces in place.
I missed the original message, but I wish to comment on it.
>From what I've read, and understand, GVN/LVN will catch cases that
all forms of CSE cannot. However, CSE catches many catches GVN/LVN
cannot. So I guess what I am asking is if this paper proposes a new
form of GVN which obsoletes CSE?
I ask this because I have a rough draft of an LVN pass for gcc I am
currently hacking on. I figured I'd do the local version now since
the SSA stuff will probably not be in for some time now. (no, no
early patches, it doesn't even compile yet, just jotting down the
basic structure right now).
Later,
David S. Miller
davem@dm.cobaltmicro.com