This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: optimization/6162: gcc 3.0.4: certain i386 asm reloader ice
- From: Michael Matz <matz at suse dot de>
- To: Nathanael Nerode <neroden at twcny dot rr dot com>,Kevin Ryde <user42 at zip dot com dot au>
- Cc: <gcc-gnats at gcc dot gnu dot org>, <gcc-prs at gcc dot gnu dot org>,<gcc-bugs at gcc dot gnu dot org>, <gcc at gcc dot gnu dot org>
- Date: Wed, 30 Oct 2002 23:09:20 +0100 (CET)
- Subject: Re: optimization/6162: gcc 3.0.4: certain i386 asm reloader ice
Hi,
On Wed, 30 Oct 2002, Nathanael Nerode wrote:
> I know this is a regression, but:
> * it's a subtle and difficult bug in reload, which may take a long time
> and a lot of work to find
Well, to find it is quite easy. reload is inherently incapable of
handling more than one commutative operand, but there are two of them in
the asm. If you are unlucky both of them would need to be swapped to make
the insn valid, and this is what breaks reload, as it only can swap the
last pair. I don't know why this constraint isn't documented, but anyway,
that's the reason.
To fix this indeed would be quite involved, as we would need to handle
then all permutations of swappings (of which there are 4 with two pairs).
Well, _I_ would fix this bug by documenting that behaviour. Note, that
also, if we get rid of reload, the functionality will likely not change in
that respect.
> * reload is intended to go away for 3.4
On Thu, 31 Oct 2002, Kevin Ryde wrote:
> The second (simpler) example I posted does it without that.
>
> > I don't think it's reasonable to hold up 3.2.1 or 3.3 for this bug.
>
> As far as I can tell you have to be pretty lucky, or unlucky, to
> actually strike it. I've only had it two (or maybe three) times, and
> not on finished code.
Simply don't use more than one commutative pair, and you are safe (except
if you really hit a bug of course ;-) )
Ciao,
Michael.