On Thu, Jul 8, 2021 at 2:42 PM Richard Sandiford via Gcc-patches
<gcc-patches@gcc.gnu.org> wrote:
>
> This patch constructs an array_slice of the scalar statements that
> produce live-out reduction results in the original unvectorised loop.
> There are three cases:
>
> - SLP reduction chains: the final SLP stmt is live-out
> - full SLP reductions: all SLP stmts are live-out
> - non-SLP reductions: the single scalar stmt is live-out
>
> This is a slight simplification on its own, mostly because it maans
> “group_size” has a consistent meaning throughout the function.
> The main justification though is that it helps with later patches.

OK

> gcc/
>         * tree-vect-loop.c (vect_create_epilog_for_reduction): Truncate
>         scalar_results to group_size elements after reducing down from
>         N*group_size elements.  Construct an array_slice of the live-out
>         stmts and assert that there is one stmt per scalar result.
> ---
>  gcc/tree-vect-loop.c | 61 +++++++++++++++-----------------------------
>  1 file changed, 21 insertions(+), 40 deletions(-)
>
> diff --git a/gcc/tree-vect-loop.c b/gcc/tree-vect-loop.c
> index 7c3e3352b43..8390ac80ca0 100644
> --- a/gcc/tree-vect-loop.c
> +++ b/gcc/tree-vect-loop.c
> @@ -5010,7 +5010,12 @@ vect_create_epilog_for_reduction (loop_vec_info 
> loop_vinfo,
>    auto_vec<tree> scalar_results;
>    unsigned int group_size = 1, k;
>    auto_vec<gimple *> phis;
> -  bool slp_reduc = false;
> +  /* SLP reduction without reduction chain, e.g.,
> +     # a1 = phi <a2, a0>
> +     # b1 = phi <b2, b0>
> +     a2 = operation (a1)
> +     b2 = operation (b1)  */
> +  bool slp_reduc = (slp_node && !REDUC_GROUP_FIRST_ELEMENT (stmt_info));
>    bool direct_slp_reduc;
>    tree new_phi_result;
>    tree induction_index = NULL_TREE;
> @@ -5050,6 +5055,16 @@ vect_create_epilog_for_reduction (loop_vec_info 
> loop_vinfo,
>         adjustment_def = STMT_VINFO_REDUC_EPILOGUE_ADJUSTMENT (reduc_info);
>      }
>
> +  stmt_vec_info single_live_out_stmt[] = { stmt_info };
> +  array_slice<const stmt_vec_info> live_out_stmts = single_live_out_stmt;
> +  if (slp_reduc)
> +    /* All statements produce live-out values.  */
> +    live_out_stmts = SLP_TREE_SCALAR_STMTS (slp_node);
> +  else if (slp_node)
> +    /* The last statement in the reduction chain produces the live-out
> +       value.  */
> +    single_live_out_stmt[0] = SLP_TREE_SCALAR_STMTS (slp_node)[group_size - 
> 1];
> +
>    unsigned vec_num;
>    int ncopies;
>    if (slp_node)
> @@ -5248,13 +5263,6 @@ vect_create_epilog_for_reduction (loop_vec_info 
> loop_vinfo,
>    new_scalar_dest = vect_create_destination_var (scalar_dest, NULL);
>    bitsize = TYPE_SIZE (scalar_type);
>
> -  /* SLP reduction without reduction chain, e.g.,
> -     # a1 = phi <a2, a0>
> -     # b1 = phi <b2, b0>
> -     a2 = operation (a1)
> -     b2 = operation (b1)  */
> -  slp_reduc = (slp_node && !REDUC_GROUP_FIRST_ELEMENT (stmt_info));
> -
>    /* True if we should implement SLP_REDUC using native reduction operations
>       instead of scalar operations.  */
>    direct_slp_reduc = (reduc_fn != IFN_LAST
> @@ -5877,6 +5885,7 @@ vect_create_epilog_for_reduction (loop_vec_info 
> loop_vinfo,
>                                           first_res, res);
>                    scalar_results[j % group_size] = new_res;
>                  }
> +             scalar_results.truncate (group_size);
>               for (k = 0; k < group_size; k++)
>                 scalar_results[k] = gimple_convert (&stmts, scalar_type,
>                                                     scalar_results[k]);
> @@ -5969,39 +5978,11 @@ vect_create_epilog_for_reduction (loop_vec_info 
> loop_vinfo,
>            use <s_out4>
>            use <s_out4> */
>
> -
> -  /* In SLP reduction chain we reduce vector results into one vector if
> -     necessary, hence we set here REDUC_GROUP_SIZE to 1.  SCALAR_DEST is the
> -     LHS of the last stmt in the reduction chain, since we are looking for
> -     the loop exit phi node.  */
> -  if (REDUC_GROUP_FIRST_ELEMENT (stmt_info))
> -    {
> -      stmt_vec_info dest_stmt_info
> -       = vect_orig_stmt (SLP_TREE_SCALAR_STMTS (slp_node)[group_size - 1]);
> -      scalar_dest = gimple_assign_lhs (dest_stmt_info->stmt);
> -      group_size = 1;
> -    }
> -
> -  /* In SLP we may have several statements in NEW_PHIS and REDUCTION_PHIS (in
> -     case that REDUC_GROUP_SIZE is greater than vectorization factor).
> -     Therefore, we need to match SCALAR_RESULTS with corresponding 
> statements.
> -     The first (REDUC_GROUP_SIZE / number of new vector stmts) scalar results
> -     correspond to the first vector stmt, etc.
> -     (RATIO is equal to (REDUC_GROUP_SIZE / number of new vector stmts)).  */
> -  if (group_size > new_phis.length ())
> -    gcc_assert (!(group_size % new_phis.length ()));
> -
> -  for (k = 0; k < group_size; k++)
> +  gcc_assert (live_out_stmts.size () == scalar_results.length ());
> +  for (k = 0; k < live_out_stmts.size (); k++)
>      {
> -      if (slp_reduc)
> -        {
> -         stmt_vec_info scalar_stmt_info = SLP_TREE_SCALAR_STMTS 
> (slp_node)[k];
> -
> -         orig_stmt_info = STMT_VINFO_RELATED_STMT (scalar_stmt_info);
> -         /* SLP statements can't participate in patterns.  */
> -         gcc_assert (!orig_stmt_info);
> -         scalar_dest = gimple_assign_lhs (scalar_stmt_info->stmt);
> -        }
> +      stmt_vec_info scalar_stmt_info = vect_orig_stmt (live_out_stmts[k]);
> +      scalar_dest = gimple_assign_lhs (scalar_stmt_info->stmt);
>
>        phis.create (3);
>        /* Find the loop-closed-use at the loop exit of the original scalar

Reply via email to