|
|
On Wed, Jun 3, 2009 at 3:19 PM, Razya Ladelsky <RAZYA@xxxxxxxxxx> wrote:
> Zdenek Dvorak <rakdver@xxxxxxxxxxxxxxx> wrote on 03/06/2009 15:58:22:
>
>> Hi,
>>
>> > The patch committed to GCC4.4 following PR38156 was slightly different
>
>> > than what was submitted here:
>> > http://gcc.gnu.org/ml/gcc-patches/2008-11/msg00882.html, although it
>
>> > fixed the failure as well.
>> > (but creating unnecessary structures)
>> >
>> > I would like to commit the more appropriate patch now to GCC4.5, so
> that
>> > arg_struct and new_arg_struct will not be
>> > built when there's no reason for it to happen.
>>
>> the patch is OK.
>
> Thanks Zdenek,
> and regarding the backporting?
> In GCC4.3 ,the testcase is not failing by default, only when
> -ftree-parallelize-loops
> is enabled.
It's not a regression, but if the patch below is all what needs to
be backported I guess it's fine (supposed it passes bootstrap
and regtest on the branch).
Richard.
> Thanks,
> Razya
>
>
>>
>> > Index: tree-parloops.c
>> > ===================================================================
>> > --- tree-parloops.c (revision 147835)
>> > +++ tree-parloops.c (working copy)
>> > @@ -1193,7 +1193,7 @@
>> >
>> > VEC_free (basic_block, heap, body);
>> >
>> > - if (htab_elements (name_copies) == 0 && reduction_list == 0)
>> > + if (htab_elements (name_copies) == 0 && htab_elements
>> (reduction_list)==0)
>> > {
>>
>> Add space before and after ==
>>
>> Zdenek
>
>
|
|