On Mon, Aug 14, 2023 at 10:52 AM Kewen.Lin wrote:
>
> Hi,
>
> In function vectorizable_load, there is one hunk which is
> dedicated for the handlings on VMAT_INVARIANT and return
> early, it means we shouldn't encounter any cases with
> memory_access_type VMAT_INVARIANT in the following code
> after that. This patch is to clean up several useless
> checks on VMAT_INVARIANT. There should be no functional
> changes.
>
> Bootstrapped and regtested on x86_64-redhat-linux,
> aarch64-linux-gnu and powerpc64{,le}-linux-gnu.
OK.
> gcc/ChangeLog:
>
> * tree-vect-stmts.cc (vectorizable_load): Remove some useless checks
> on VMAT_INVARIANT.
> ---
> gcc/tree-vect-stmts.cc | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/gcc/tree-vect-stmts.cc b/gcc/tree-vect-stmts.cc
> index 89607a98f99..d4e781531fd 100644
> --- a/gcc/tree-vect-stmts.cc
> +++ b/gcc/tree-vect-stmts.cc
> @@ -10499,7 +10499,7 @@ vectorizable_load (vec_info *vinfo,
> tree bias = NULL_TREE;
> if (!costing_p)
> {
> - if (loop_masks && memory_access_type != VMAT_INVARIANT)
> + if (loop_masks)
> final_mask
> = vect_get_loop_mask (loop_vinfo, gsi, loop_masks,
> vec_num * ncopies, vectype,
> @@ -10729,7 +10729,7 @@ vectorizable_load (vec_info *vinfo,
> bias = build_int_cst (intQI_type_node, biasval);
> }
>
> - if (final_len && memory_access_type != VMAT_INVARIANT)
> + if (final_len)
> {
> tree ptr
> = build_int_cst (ref_type, align * BITS_PER_UNIT);
> --
> 2.31.1