This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] misalignment and loop peeling (vectorizer)
- From: Ira Rosen <IRAR at il dot ibm dot com>
- To: Devang Patel <dpatel at apple dot com>
- Cc: Dorit Nuzman <DORIT at il dot ibm dot com>, gcc-patches at gcc dot gnu dot org
- Date: Thu, 2 Feb 2006 11:46:03 +0200
- Subject: Re: [PATCH] misalignment and loop peeling (vectorizer)
Devang Patel <dpatel@apple.com> wrote on 01/02/2006 20:53:37:
>
> On Feb 1, 2006, at 12:30 AM, Dorit Nuzman wrote:
>
> > DR_MISALIGNMENT is probably what you want to use - that's where the
> > misalignment of the first access in the loop is stored, and what
> > peeling
> > tries to align (it's -1 if it's unknown at compile time).
>
> IIUC, DR_OFFSET_MISALIGNMENT also represents the same thing. One is
> an int value, where as DR_OFFSET_MISALIGNMENT is a tree node. Is
> there any other difference between these two? Either way it is fine
> for me.
>
DR_MISALIGNMENT is DR_OFFSET_MISALIGNMENT modulo vector size.
DR_MISALIGNMENT is a vectorizer specific field, cause it depends on the
vector size, and DR_OFFSET_MISALIGNMENT is a general tree-data-refs field.
We probably should improve the documentation for these fields.
Ira