This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch] Externalize data-ref analysis - part 1
- From: Daniel Berlin <dberlin at dberlin dot org>
- To: Ira Rosen <IRAR at il dot ibm dot com>
- Cc: gcc-patches at gcc dot gnu dot org, Sebastian Pop <sebastian dot pop at cri dot ensmp dot fr>
- Date: Thu, 07 Jul 2005 08:50:57 -0400
- Subject: Re: [patch] Externalize data-ref analysis - part 1
- References: <OF5D9795CB.C18C7089-ONC2257037.002590AC-C2257037.0029956D@il.ibm.com>
On Thu, 2005-07-07 at 10:34 +0300, Ira Rosen wrote:
>
>
>
> > If this is information about what the vectorizer requires in terms of
> > alignment for a given loop, that should not be passed to this generic
> > function. *That* part (and whatever misalignment info it needs) should
> > be computed separately by the vectorizer.
> >
>
> I understand that this is not the best solution. But otherwise I can't see
> any other way to calculate misalignment in the vectorizer rather than to do
> all the analysis again.
Why can you not simply record the alignment you have (not what you want
the alignment to be) in the dr structure (since this what
get_inner_reference, etc, give you), and then determine whether this is
misaligned in the vectorizer?
Then nothing needs to be passed to these functions.
ISTM you can always record the biggest alignment you know something has
during analysis, and then later see if this is okay for your purposes.
> Misalignment information is retrieved from the
> offset expression returned by get_inner_reference, along with the offset,
> step, etc.
>
> Thanks,
> Ira
>
>
>