On Mon, Apr 16, 2018 at 6:16 PM, Andreas Krebbel
<kreb...@linux.vnet.ibm.com> wrote:
> I did run into an ICE with a single element vector triggered by
> dividing the number of elements by 2 with exact_div here:
>
> tree-vect-data-refs.c:5132
>
> else
> {
> /* If length is not equal to 3 then only power of 2 is supported.
> */
> gcc_assert (pow2p_hwi (count));
> poly_uint64 nelt = GET_MODE_NUNITS (mode);
>
> /* The encoding has 2 interleaved stepped patterns. */
> vec_perm_builder sel (nelt, 2, 3);
> sel.quick_grow (6);
> for (i = 0; i < 3; i++)
> {
> sel[i * 2] = i;
> sel[i * 2 + 1] = i + nelt;
> }
> vec_perm_indices indices (sel, 2, nelt);
> if (can_vec_perm_const_p (mode, indices))
> {
> for (i = 0; i < 6; i++)
> sel[i] += exact_div (nelt, 2); <-----
> indices.new_vector (sel, 2, nelt);
> if (can_vec_perm_const_p (mode, indices))
> return true;
> }
> }
>
> The patch adds a check to prevent this.
Testcase?
What's the group size?
Did it work before the poly-int stuff? Single-element vectors are
somewhat "special" - where do they appear for you?
Richard.
> Ok?
>
> -Andreas-
>
> gcc/ChangeLog:
>
> 2018-04-16 Andreas Krebbel <kreb...@linux.ibm.com>
>
> * tree-vect-data-refs.c (vect_grouped_store_supported): Exit for
> single element vectors.
> ---
> gcc/tree-vect-data-refs.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/gcc/tree-vect-data-refs.c b/gcc/tree-vect-data-refs.c
> index 161a886..01e28ca 100644
> --- a/gcc/tree-vect-data-refs.c
> +++ b/gcc/tree-vect-data-refs.c
> @@ -5135,6 +5135,9 @@ vect_grouped_store_supported (tree vectype, unsigned
> HOST_WIDE_INT count)
> gcc_assert (pow2p_hwi (count));
> poly_uint64 nelt = GET_MODE_NUNITS (mode);
>
> + if (maybe_eq (nelt, 1U))
> + return false;
> +
> /* The encoding has 2 interleaved stepped patterns. */
> vec_perm_builder sel (nelt, 2, 3);
> sel.quick_grow (6);
> --
> 2.9.1
>
