This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: [ tree-ssa] PRE problem
- From: law at redhat dot com
- To: Diego Novillo <dnovillo at redhat dot com>
- Cc: Daniel Berlin <dberlin at dberlin dot org>, "gcc at gcc dot gnu dot org" <gcc at gcc dot gnu dot org>
- Date: Fri, 20 Feb 2004 13:31:55 -0700
- Subject: Re: [ tree-ssa] PRE problem
- Reply-to: law at redhat dot com
In message <1077307706.3169.194.camel@localhost.localdomain>, Diego Novillo wri
tes:
>
>--=-mRW9vtO4kFf6Ff1teIUo
>Content-Type: text/plain
>Content-Transfer-Encoding: 7bit
>
>On Fri, 2004-02-20 at 14:59, law@redhat.com wrote:
>> In message <1077304036.3169.162.camel@localhost.localdomain>, Diego Novillo
> wri
>> tes:
>> >
>> >See if this (untested) patch fixes your problem. We were not calling
>> >propagate_value in generate_expr_as_of_bb. This fixes PR14153, but I
>> >haven't checked other places in PRE where we may be doing the same.
>> So if you really think this is going to work, build c, c++ (no need to
>> fully bootstrap), then try to build libstdc++.
>>
>I just did :) Attached.
Bloody odd since I just repeated my test and couldn't get it to fail!! That
included with and without the two versions of my dominator improvements
and with and without the tree loop header copying.
It must be Friday... Time to get away and recharge.
>Have you got any other local changes?
Not that I'm aware of :-)
>Your idea of having PRE check
>whether the propagation is valid sounded good to me. So, perhaps I'm
>just not tickling the bug with a pristine branch.
I think it's the right thing to do, particularly since the destination
could be a DECL_HARD_REGISTER and I didn't see a way to necessarily
detect that elsewhere.
Jeff