This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH, take 2] Fix PR tree-optimization/49960 ,Fix self data dependence
- From: Razya Ladelsky <RAZYA at il dot ibm dot com>
- To: Jakub Jelinek <jakub at redhat dot com>
- Cc: GCC Patches <gcc-patches at gcc dot gnu dot org>, Richard Guenther <richard dot guenther at gmail dot com>
- Date: Mon, 21 Nov 2011 15:50:10 +0200
- Subject: Re: [PATCH, take 2] Fix PR tree-optimization/49960 ,Fix self data dependence
- References: <OF662F6234.DE8C60EA-ONC2257926.0048D53A-C225792C.00231C15@il.ibm.com> <CAFiYyc15E0jqzPC+SD40cWTj18Tes0ZbEm56BXHHe-eZnqSRvw@mail.gmail.com> <OF746BCB18.CF82809F-ONC225792E.0051CE3F-C225792E.00564975@il.ibm.com> <CAFiYyc2ykFPCW8A8vW=f5UbNa7zFRQObwL13D9ioXjCd_em9pQ@mail.gmail.com> <CAFiYyc0D8n6BU-CLqB132FBuxOqJdmiRBDTiZ6Spf8P0J5rbJg@mail.gmail.com> <OFE59B30C0.27072E7A-ONC2257949.0035FEEA-C2257949.0039D985@il.ibm.com> <20111121125707.GG27242@tyan-ft48-01.lab.bos.redhat.com>
gcc-patches-owner@gcc.gnu.org wrote on 21/11/2011 02:57:07 PM:
> From: Jakub Jelinek <jakub@redhat.com>
> To: Razya Ladelsky/Haifa/IBM@IBMIL
> Cc: Richard Guenther <richard.guenther@gmail.com>, GCC Patches <gcc-
> patches@gcc.gnu.org>
> Date: 21/11/2011 02:57 PM
> Subject: Re: [PATCH, take 2] Fix PR tree-optimization/49960 ,Fix
> self data dependence
> Sent by: gcc-patches-owner@gcc.gnu.org
>
> On Tue, Nov 15, 2011 at 12:31:39PM +0200, Razya Ladelsky wrote:
> > This patch fixes the failures described in
> > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=49960
> > It also fixes bzips when run with autopar enabled.
>
> Sorry, I've been away and couldn't react to this earlier, but the patch
> looks wrong to me. The reason why I've been calling
compute_self_dependence
> from the gather handling code is to do there what we did for normal data
> references in compute_all_dependences, except for the affine stuff.
Hi Jakub,
what do you mean by 'except for the affine stuff'?
Before having this patch, compute_self_depepndence just marked the
distance zero
and returned, while now it calls compute_affine_dependence.
So I think you do get the right outcome now.
> So, when compute_all_dependences no longer calls it, neither should
> vect_analyze_data_refs.
So do you want to replace it with the call to compute_affine_dependence?
Thanks,
Razya
>
> Jakub
>