This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PATCH to remove unnecessary NOTE_INSN_DELETED
- To: Jeffrey A Law <law at cygnus dot com>
- Subject: Re: PATCH to remove unnecessary NOTE_INSN_DELETED
- From: Bernd Schmidt <bernds at cygnus dot co dot uk>
- Date: Thu, 4 Nov 1999 18:34:52 +0000 (GMT)
- cc: Mark Mitchell <mark at codesourcery dot com>, kenner at vlsi1 dot ultra dot nyu dot edu, gcc-patches at gcc dot gnu dot org
On Thu, 4 Nov 1999, Jeffrey A Law wrote:
> In message <Pine.LNX.4.10.9911041808370.6182-100000@biriani.cygnus.co.uk>you
> write:
> > Just FYI, I started to look into this. I've hit one stumbling block so far
> > (expressions of the form uid_luid[REGNO_FIRST_UID (xxx)], but I guess that
> > can be worked around somehow.
> Instead of recording the UID record the actual insn perhaps.
That breaks down if the insn gets deleted (loop has some code to deal with
that problem by making uid_luid "approximately right" for insns that have been
deleted).
> Or if we relax the "luids are for within a pass only" decree slightly we
> could go ahead and record the first/last luid in regscan.
That might work better.
Apart from those problems, the mechanisms used e.g. in loop which involve
comparisons between insn uids/luids to determine which insn is executed first
seem exceedingly brittle; I suspect it might not be hard to fool parts of the
loop optimizer with clever use of gotos.
Bernd