public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
From: Richard Biener <rguenth@gcc.gnu.org>
To: gcc-cvs@gcc.gnu.org
Subject: [gcc r12-3042] Refactor BB splitting of DRs for SLP group analysis
Date: Fri, 20 Aug 2021 10:37:35 +0000 (GMT)	[thread overview]
Message-ID: <20210820103735.3B95F3894410@sourceware.org> (raw)

https://gcc.gnu.org/g:37744f8260857005c8409c9e2e633a05c768a7dd

commit r12-3042-g37744f8260857005c8409c9e2e633a05c768a7dd
Author: Richard Biener <rguenther@suse.de>
Date:   Fri Aug 20 11:32:00 2021 +0200

    Refactor BB splitting of DRs for SLP group analysis
    
    This uses the group_id computed to ensure DRs in different BBs do
    not get merged into a DR group.  To achieve this we seed the
    group from the BB index when group_ids are not computed and we
    make sure to bump the group_id when advancing to the next BB for
    BB SLP analysis.
    
    This paves the way for relaxing the grouping for BB vectorization
    by adjusting its group_id computation.
    
    2021-08-20  Richard Biener  <rguenther@suse.de>
    
            * tree-vect-data-refs.c (dr_group_sort_cmp): Do not compare
            BBs.
            (vect_analyze_data_ref_accesses): Likewise.  Assign the BB
            index as group_id when dataref_groups were not computed.
            * tree-vect-slp.c (vect_slp_bbs): Bump current_group when
            we advace to the next BB.

Diff:
---
 gcc/tree-vect-data-refs.c | 21 ++++++++-------------
 gcc/tree-vect-slp.c       |  2 ++
 2 files changed, 10 insertions(+), 13 deletions(-)

diff --git a/gcc/tree-vect-data-refs.c b/gcc/tree-vect-data-refs.c
index b7dde07fc5e..37f46d1aaa3 100644
--- a/gcc/tree-vect-data-refs.c
+++ b/gcc/tree-vect-data-refs.c
@@ -2831,12 +2831,6 @@ dr_group_sort_cmp (const void *dra_, const void *drb_)
   if (dra == drb)
     return 0;
 
-  /* DRs in different basic-blocks never belong to the same group.  */
-  int bb_index1 = gimple_bb (DR_STMT (dra))->index;
-  int bb_index2 = gimple_bb (DR_STMT (drb))->index;
-  if (bb_index1 != bb_index2)
-    return bb_index1 < bb_index2 ? -1 : 1;
-
   /* Different group IDs lead never belong to the same group.  */
   if (dra_pair.second != drb_pair.second)
     return dra_pair.second < drb_pair.second ? -1 : 1;
@@ -2963,7 +2957,13 @@ vect_analyze_data_ref_accesses (vec_info *vinfo,
   datarefs_copy.create (datarefs.length ());
   for (unsigned i = 0; i < datarefs.length (); i++)
     {
-      int group_id = dataref_groups ? (*dataref_groups)[i] : 0;
+      int group_id;
+      /* If the caller computed DR grouping use that, otherwise group by
+	 basic blocks.  */
+      if (dataref_groups)
+	group_id = (*dataref_groups)[i];
+      else
+	group_id = gimple_bb (DR_STMT (datarefs[i]))->index;
       datarefs_copy.quick_push (data_ref_pair (datarefs[i], group_id));
     }
   datarefs_copy.qsort (dr_group_sort_cmp);
@@ -2999,13 +2999,8 @@ vect_analyze_data_ref_accesses (vec_info *vinfo,
 	     matters we can push those to a worklist and re-iterate
 	     over them.  The we can just skip ahead to the next DR here.  */
 
-	  /* DRs in a different BBs should not be put into the same
+	  /* DRs in a different DR group should not be put into the same
 	     interleaving group.  */
-	  int bb_index1 = gimple_bb (DR_STMT (dra))->index;
-	  int bb_index2 = gimple_bb (DR_STMT (drb))->index;
-	  if (bb_index1 != bb_index2)
-	    break;
-
 	  if (dra_group_id != drb_group_id)
 	    break;
 
diff --git a/gcc/tree-vect-slp.c b/gcc/tree-vect-slp.c
index 4dcc70c5dd7..d2f6a16f220 100644
--- a/gcc/tree-vect-slp.c
+++ b/gcc/tree-vect-slp.c
@@ -5985,6 +5985,8 @@ vect_slp_bbs (const vec<basic_block> &bbs)
 					      &dataref_groups, current_group))
 	    ++current_group;
 	}
+      /* New BBs always start a new DR group.  */
+      ++current_group;
     }
 
   return vect_slp_region (bbs, datarefs, &dataref_groups, insns);


                 reply	other threads:[~2021-08-20 10:37 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210820103735.3B95F3894410@sourceware.org \
    --to=rguenth@gcc.gnu.org \
    --cc=gcc-cvs@gcc.gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).