This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Fix PR45199: do not aggregate memory accesses to the same array for -ftree-loop-distribute-patterns
On Wed, Dec 1, 2010 at 09:26, Sebastian Pop <sebpop@gmail.com> wrote:
> On Wed, Dec 1, 2010 at 04:21, Richard Guenther <rguenther@suse.de> wrote:
>> Comments inline (non-quoted patch - attachments suck).
>
> Sorry. ?This is the last time I will use gmail to send out a patch.
>
>> + ?return dr_analyze_innermost (dr) &&
>> + ? ?(integer_zerop (size_binop (MINUS_EXPR,
>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? fold_convert (sizetype, DR_STEP (dr)),
>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? TYPE_SIZE_UNIT (TREE_TYPE (op0))))
>> + ? ? || integer_zerop (size_binop (PLUS_EXPR,
>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?TYPE_SIZE_UNIT (TREE_TYPE (op0)),
>> + ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?fold_convert (sizetype, DR_STEP (dr)))));
>>
>> You leak dr. ?&& goes to the next line.
>
> Right.
>
>> You want
>> tree_int_cst_equal (fold_unary (ABS_EXPR, TREE_TYPE (DR_STEP (dr)),
>> ? ? ? ? ? ? ? ? ? ? ? ? ? ? ? ?DR_STEP (dr)), TYPE_SIZE_UNIT (...))
>>
>
> I will replace this also in generate_memset_zero.
>
>> if this is the same test as in generate_memset_zero please put this
>> helper in tree-loop-distribution and use it there as well.
>
> The problem is that in generate_memset_zero we do something different
> in the branches:
Oh never mind this comment, I see what you meant before.
Sebastian