On Fri, 21 Jul 2023, Juzhe-Zhong wrote:
> Hi, Richard and Richi.
>
> I have double check the recent codes for len && mask support again.
>
> Some places code structure:
>
> if (len_mask_fn)
> ...
> else if (mask_fn)
> ...
>
> some places code structure:
>
> if (mask_len_fn)
> ...
> else if (mask)
>
> Base on previous review comment from Richi:
> https://gcc.gnu.org/pipermail/gcc-patches/2023-July/625067.html
>
> len mask stuff should be checked before mask.
>
> So I reorder all condition order to check LEN MASK stuff before MASK.
>
> This is the last clean up patch.
>
> Boostrap and Regression is on the way.
OK.
> gcc/ChangeLog:
>
> * tree-vect-stmts.cc (check_load_store_for_partial_vectors): Change
> condition order.
> (vectorizable_operation): Ditto.
>
> ---
> gcc/tree-vect-stmts.cc | 24 ++++++++++++------------
> 1 file changed, 12 insertions(+), 12 deletions(-)
>
> diff --git a/gcc/tree-vect-stmts.cc b/gcc/tree-vect-stmts.cc
> index d5b4f020332..2fe856db9ab 100644
> --- a/gcc/tree-vect-stmts.cc
> +++ b/gcc/tree-vect-stmts.cc
> @@ -1635,17 +1635,17 @@ check_load_store_for_partial_vectors (loop_vec_info
> loop_vinfo, tree vectype,
> internal_fn len_ifn = (is_load
> ? IFN_MASK_LEN_GATHER_LOAD
> : IFN_MASK_LEN_SCATTER_STORE);
> - if (internal_gather_scatter_fn_supported_p (ifn, vectype,
> + if (internal_gather_scatter_fn_supported_p (len_ifn, vectype,
> gs_info->memory_type,
> gs_info->offset_vectype,
> gs_info->scale))
> - vect_record_loop_mask (loop_vinfo, masks, nvectors, vectype,
> - scalar_mask);
> - else if (internal_gather_scatter_fn_supported_p (len_ifn, vectype,
> + vect_record_loop_len (loop_vinfo, lens, nvectors, vectype, 1);
> + else if (internal_gather_scatter_fn_supported_p (ifn, vectype,
> gs_info->memory_type,
> gs_info->offset_vectype,
> gs_info->scale))
> - vect_record_loop_len (loop_vinfo, lens, nvectors, vectype, 1);
> + vect_record_loop_mask (loop_vinfo, masks, nvectors, vectype,
> + scalar_mask);
> else
> {
> if (dump_enabled_p ())
> @@ -6596,16 +6596,16 @@ vectorizable_operation (vec_info *vinfo,
> && LOOP_VINFO_CAN_USE_PARTIAL_VECTORS_P (loop_vinfo)
> && mask_out_inactive)
> {
> - if (cond_fn != IFN_LAST
> - && direct_internal_fn_supported_p (cond_fn, vectype,
> + if (cond_len_fn != IFN_LAST
> + && direct_internal_fn_supported_p (cond_len_fn, vectype,
> OPTIMIZE_FOR_SPEED))
> - vect_record_loop_mask (loop_vinfo, masks, ncopies * vec_num,
> - vectype, NULL);
> - else if (cond_len_fn != IFN_LAST
> - && direct_internal_fn_supported_p (cond_len_fn, vectype,
> - OPTIMIZE_FOR_SPEED))
> vect_record_loop_len (loop_vinfo, lens, ncopies * vec_num, vectype,
> 1);
> + else if (cond_fn != IFN_LAST
> + && direct_internal_fn_supported_p (cond_fn, vectype,
> + OPTIMIZE_FOR_SPEED))
> + vect_record_loop_mask (loop_vinfo, masks, ncopies * vec_num,
> + vectype, NULL);
> else
> {
> if (dump_enabled_p ())
>
--
Richard Biener <rguent...@suse.de>
SUSE Software Solutions Germany GmbH, Frankenstrasse 146, 90461 Nuernberg,
Germany; GF: Ivo Totev, Andrew Myers, Andrew McDonald, Boudien Moerman;
HRB 36809 (AG Nuernberg)
