On Tue, Jan 3, 2012 at 12:26 PM, Eric Botcazou <ebotca...@adacore.com> wrote:
>> Shouldn't we go back and unconditionally return false if lhs and (or?) rhs
>> are BLKmode? I suppose they are, in this case. Or should we truncate
>> the alignment from get_object_alignment to TYPE_ALIGN if it is bigger
>> than that to avoid this situation?
>
> Sorry, I didn't read your message before sending mine. Yes, the alignment cap
> appears to be the best solution, but of course TYPE_ALIGN (access->type) won't
> work (that was the original code), so TYPE_ALIGN (TREE_TYPE (access->base)):
>
> Index: tree-sra.c
> ===================================================================
> --- tree-sra.c (revision 182780)
> +++ tree-sra.c (working copy)
> @@ -1124,7 +1124,9 @@ build_accesses_from_assign (gimple stmt)
> {
> lacc->grp_assignment_write = 1;
> if (STRICT_ALIGNMENT
> - && tree_non_aligned_mem_p (rhs, get_object_alignment (lhs)))
> + && tree_non_aligned_mem_p (rhs,
> + MIN (TYPE_ALIGN (TREE_TYPE (lacc->base)),
> + get_object_alignment (lhs))))
> lacc->grp_unscalarizable_region = 1;
> }
>
> @@ -1135,7 +1137,9 @@ build_accesses_from_assign (gimple stmt)
> && !is_gimple_reg_type (racc->type))
> bitmap_set_bit (should_scalarize_away_bitmap, DECL_UID (racc->base));
> if (STRICT_ALIGNMENT
> - && tree_non_aligned_mem_p (lhs, get_object_alignment (rhs)))
> + && tree_non_aligned_mem_p (lhs,
> + MIN (TYPE_ALIGN (TREE_TYPE (racc->base)),
> + get_object_alignment (rhs))))
> racc->grp_unscalarizable_region = 1;
I suppose we should move this logic to tree_non_aligned_mem_p, maybe
renaming it to tree_non_matching_align_mems_p or so, thus passing
down both lhs and rhs.
But I agree - let's sort out correctness issues first, then try to address the
missed optimizations we caused by that.
Richard.
> }
>
> --
> Eric Botcazou
