This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [Graphite] Fix type problems in loop ivs.
- From: Richard Guenther <richard dot guenther at gmail dot com>
- To: Sebastian Pop <sebpop at gmail dot com>
- Cc: Tobias Grosser <grosser at fim dot uni-passau dot de>, gcc-patches at gcc dot gnu dot org, gcc-graphite at googlegroups dot com
- Date: Wed, 2 Jun 2010 11:03:01 +0200
- Subject: Re: [Graphite] Fix type problems in loop ivs.
- References: <1267798447-4699-1-git-send-email-grosser@fim.uni-passau.de> <cb9d34b21003101233q28787e42sb84f3725ae33c640@mail.gmail.com> <84fc9c001003110248l1aae7864k9f9375f03c65ce66@mail.gmail.com> <cb9d34b21003110554w5acbd3baoe6275f6afa52ebb5@mail.gmail.com> <84fc9c001003110616w5da6c538t606d6e108f6b8ab4@mail.gmail.com> <AANLkTimTpH67CihYVQJdXAcsDCGl5DtZ0H8qyZQWcNLf@mail.gmail.com>
On Wed, Jun 2, 2010 at 1:09 AM, Sebastian Pop <sebpop@gmail.com> wrote:
> On Thu, Mar 11, 2010 at 09:16, Richard Guenther
> <richard.guenther@gmail.com> wrote:
>> On Thu, Mar 11, 2010 at 2:54 PM, Sebastian Pop <sebpop@gmail.com> wrote:
>>> On Thu, Mar 11, 2010 at 04:48, Richard Guenther
>>> <richard.guenther@gmail.com> wrote:
>>>> Java indeed fails to initialize most of the common trees.
>>>> I suggest to use ssizetype if long_long_type_node is NULL.
>>>
>>> Here is the fix. ?Does this look good?
>>
>> Hm. ?I don't like the #define. ?And Java doesn't initialize
>> long_integer_type_node either, so I think the patch won't fix
>> the bootstrap issue. ?Basically the C kind types are not
>> really supposed to be used in the middle-end.
>>
>> What you probably want is to use
>> smallest_mode_for_size (type_precision + 1, MODE_INT)
>> and fail if that returns BLKmode (ugh, it doesn't ...).
>> From the modes precision then build a type using
>> build_nonstandard_integer_type. And possibly you
>> want to restrict yourself to maximum BITS_PER_WORD
>> size to avoid long long or TImode arithmetic.
>>
>
> Sorry for having neglected this issue so long.
> Here is the patch that fixes this induction variable type computation.
> Ok for trunk after the automatic tests in the graphite branch?
Ok.
Thanks,
Richard.
> Thanks,
> Sebastian
>