This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Only test STMT_VINFO_STRIDED_P for the first statement in a group
- From: Richard Biener <richard dot guenther at gmail dot com>
- To: GCC Patches <gcc-patches at gcc dot gnu dot org>, Richard Sandiford <richard dot sandiford at arm dot com>
- Date: Wed, 22 Aug 2018 14:05:16 +0200
- Subject: Re: Only test STMT_VINFO_STRIDED_P for the first statement in a group
- References: <871saqwy2e.fsf@arm.com>
On Wed, Aug 22, 2018 at 11:27 AM Richard Sandiford
<richard.sandiford@arm.com> wrote:
>
> get_load_store_type & co were testing STMT_VINFO_STRIDED_P on individual
> statements in a group instead of the first. This has no effect on
> its own, but is needed by a later patch.
>
> Tested on aarch64-linux-gnu (with and without SVE), aarch64_be-elf
> and x86_64-linux-gnu. OK to install?
OK.
Richard.
> Richard
>
>
> 2018-08-22 Richard Sandiford <richard.sandiford@arm.com>
>
> gcc/
> * tree-vect-stmts.c (get_group_load_store_type)
> (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
> first statement in a group.
>
> Index: gcc/tree-vect-stmts.c
> ===================================================================
> --- gcc/tree-vect-stmts.c 2018-08-09 15:37:12.802984133 +0100
> +++ gcc/tree-vect-stmts.c 2018-08-22 10:23:50.562747188 +0100
> @@ -2191,14 +2191,14 @@ get_group_load_store_type (stmt_vec_info
>
> /* There can only be a gap at the end of the group if the stride is
> known at compile time. */
> - gcc_assert (!STMT_VINFO_STRIDED_P (stmt_info) || gap == 0);
> + gcc_assert (!STMT_VINFO_STRIDED_P (first_stmt_info) || gap == 0);
>
> /* Stores can't yet have gaps. */
> gcc_assert (slp || vls_type == VLS_LOAD || gap == 0);
>
> if (slp)
> {
> - if (STMT_VINFO_STRIDED_P (stmt_info))
> + if (STMT_VINFO_STRIDED_P (first_stmt_info))
> {
> /* Try to use consecutive accesses of DR_GROUP_SIZE elements,
> separated by the stride, until we have a complete vector.
> @@ -2255,7 +2255,7 @@ get_group_load_store_type (stmt_vec_info
> / vect_get_scalar_dr_size (first_dr_info)))
> would_overrun_p = false;
>
> - if (!STMT_VINFO_STRIDED_P (stmt_info)
> + if (!STMT_VINFO_STRIDED_P (first_stmt_info)
> && (can_overrun_p || !would_overrun_p)
> && compare_step_with_zero (stmt_info) > 0)
> {
> @@ -2466,8 +2466,11 @@ get_load_store_type (stmt_vec_info stmt_
> /* FIXME: At the moment the cost model seems to underestimate the
> cost of using elementwise accesses. This check preserves the
> traditional behavior until that can be fixed. */
> + stmt_vec_info first_stmt_info = DR_GROUP_FIRST_ELEMENT (stmt_info);
> + if (!first_stmt_info)
> + first_stmt_info = stmt_info;
> if (*memory_access_type == VMAT_ELEMENTWISE
> - && !STMT_VINFO_STRIDED_P (stmt_info)
> + && !STMT_VINFO_STRIDED_P (first_stmt_info)
> && !(stmt_info == DR_GROUP_FIRST_ELEMENT (stmt_info)
> && !DR_GROUP_NEXT_ELEMENT (stmt_info)
> && !pow2p_hwi (DR_GROUP_SIZE (stmt_info))))