This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: auto vectorization in gcc

From: Richard Henderson <rth at redhat dot com>
To: David Edelsohn <dje at watson dot ibm dot com>
Cc: Daniel Berlin <dberlin at dberlin dot org>, Dorit Naishlos <DORIT at il dot ibm dot com>, gcc at gcc dot gnu dot org, Diego Novillo <dnovillo at redhat dot com>
Date: Thu, 17 Jul 2003 11:19:56 -0700
Subject: Re: auto vectorization in gcc
References: <D29C4A9E-B881-11D7-93E3-000A95A34564@dberlin.org> <200307171815.OAA28902@makai.watson.ibm.com>

On Thu, Jul 17, 2003 at 02:15:37PM -0400, David Edelsohn wrote:
> 	I believe part of the concern is alignment issues.  If the code
> was not originally written with vector types, we cannot guarantee that
> objects will have the appropriate alignment for SIMD instructions.

Ok, sure, but that just means either (1) we have access to the
original DECL, and we *force* proper alignment or (2) we generate
extra code to ensure alignment or (3) we bail.

The only thing that happens by waiting to rtl level is that we
lose the ability to do (1).

r~

References:
- Re: auto vectorization in gcc
  - From: Daniel Berlin
- Re: auto vectorization in gcc
  - From: David Edelsohn

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]