This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] [4.3 projects] Verctorizer - versioning for alias
- From: Victor Kaplansky <VICTORK at il dot ibm dot com>
- To: Zdenek Dvorak <rakdver at kam dot mff dot cuni dot cz>
- Cc: gcc-patches at gnu dot org
- Date: Tue, 14 Aug 2007 15:00:43 +0300
- Subject: Re: [PATCH] [4.3 projects] Verctorizer - versioning for alias
Zdenek Dvorak <rakdver@kam.mff.cuni.cz> wrote on 14/08/2007 02:37:27 PM:
>
> could you also add some test that checks the situation where the IN and
> OUT pointers overlap in such a way that the vectorized copy of the loop
> would produce wrong results (if it were used)?
OK, I'll add some.
>
> There should be an empty line between the comment and the function.
>
OK.
> > + logical and operation.
>
> ^ with
>
> > + LOOP_VINFO - fields of the loop information are used.
>
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ I am not
> sure what this sentence is supposed to mean.
I'll fix this.
>
> > + part_cond_expr =
> > + fold_build2 (TRUTH_OR_EXPR, boolean_type_node,
> > + fold_build2 (LT_EXPR, boolean_type_node,
> > + fold_build2 (PLUS_EXPR, ptr_type_node,
>
> ^^^^^^^^^
> should be POINTER_PLUS_EXPR. It is a bit suspicious that this passes
> testing, as it should ICE in build2. Can you check why it does not?
Interesting indeed. I'll try to step thru build2 to see why it doesn't ICE.
Anyway, I'll replace PLUS_EXPR by POINTER_PLUS_EXPR.
> The params should be documented in invoke.texi.
OK, I'll ad the docimentation.
Otherwise, the patch is
> OK.
>
> Zdenek
Thanks, I'll resend fixed patch shortly.
-- Victor