This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [RFC] [autovect patch] Implement vectorization hints
- From: Richard Henderson <rth at redhat dot com>
- To: Zack Weinberg <zack at codesourcery dot com>
- Cc: Steven Bosscher <stevenb at suse dot de>, gcc-patches at gcc dot gnu dot org, Devang Patel <dpatel at apple dot com>, joseph at codesourcery dot com
- Date: Mon, 7 Mar 2005 06:36:26 -0800
- Subject: Re: [RFC] [autovect patch] Implement vectorization hints
- References: <B397FD6A-85F5-42B2-8F66-E2BF57641F31@apple.com> <200502271226.15372.stevenb@suse.de> <87is4dst97.fsf@codesourcery.com>
On Sun, Feb 27, 2005 at 11:22:12AM -0800, Zack Weinberg wrote:
> For things like #pragma STDC FP_CONTRACT OFF it's a little
> simpler, because the semantics are "has effect from point of
> appearance to end of immediately enclosing block", so the optimizers
> can just look for a PRAGMA_EXPR inside each lexical block.
Except that lexical blocks are explicit decomposed early.
We need to associate (or re-associate) this data with basic
blocks asap.
> I'm not sure how to tie a pragma to the next language-level loop.
We'll definitely need language-level help for this. Consider
#pragma doit
for (i = start(); i < end; i++)
where start() expands inline and contains loops. Either that
or we have to delay inlining from where we do presently. Not
that that would be a bad thing necessarily...
r~