This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch] Fix PR26859
- From: Roger Sayle <roger at eyesopen dot com>
- To: Sebastian Pop <sebpop at gmail dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Mon, 3 Apr 2006 13:44:03 -0600 (MDT)
- Subject: Re: [patch] Fix PR26859
On Mon, 3 Apr 2006, Sebastian Pop wrote:
> Sorry, here is the part that was missing from the previous patch.
> I bootstrapped and tested this patch on amd64-linux. Committed to trunk.
Hi Sebastian,
Just a quick comment on the middle-end APIs you're using in this patch.
You seem to be using a mixture of build_int_cst and build_int_cst_type.
The difference is that the second one performs extra work to correctly
set the high order bits of a constant based upon the type. So for the
constant one, it's better to consistently use build_int_cst (type, 1)
instead of build_int_cst_type (type, 1).
Finally, you also introduce a new use of the front-end function
"convert" in this change (the last hunk), where tree-data-ref.c
should instead use "fold_convert".
Thanks,
Roger
--