This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [tree-ssa] Death to WFL nodes
- From: Daniel Berlin <dberlin at dberlin dot org>
- To: law at redhat dot com
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 14 Jan 2003 12:30:26 -0500
- Subject: Re: [tree-ssa] Death to WFL nodes
On Tuesday, January 14, 2003, at 12:21 PM, law@redhat.com wrote:
[ Resent as it appears my outgoing mail to gcc.gnu.org was being
dropped. ]
Wrapping nodes with EXPR_WITH_FILE_LOCATION nodes to carry file/line
information is rather expensive, ugly and simply doesn't work well.
I agree completely.
I can't count the number of times i've thought i've got some function
or pass working, and am ready to submit, only for bootstrap or
regtesting to be foiled by the function or pass being passed an
EXPR_WFL node (instead of what was inside of it).
So I have to add a STRIP_WFL and rebootstrap from the beginning and
re-regtest.
Repeat 50 times a day (I'm pretty sure i've written or had a hand in
the largest number of tree-ssa optimization passes right now, with or
without including the ones not yet submitted, so i'm probably a pretty
good case study here).
I've gotten to the point where i now preemptively STRIP_WFL everything,
everywhere, now unless i know it can't be a WFL (IE I just passed it
from somewhere where i STRIP_WFL'd it).
--Dan