public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
* [PATCH] Optimize vector codegen for invariant loads, fix SLP support
@ 2023-06-23  8:11 Richard Biener
  0 siblings, 0 replies; only message in thread
From: Richard Biener @ 2023-06-23  8:11 UTC (permalink / raw)
  To: gcc-patches

The following avoids creating duplicate stmts for invariant loads
which was necessary when the vector stmts were in a linked list.
It also fixes SLP support which didn't correctly create the
appropriate number of copies.

Bootstrapped and tested on x86_64-unknown-linux-gnu, pushed.

	* tree-vect-stmts.cc (vectorizable_load): Avoid useless
	copies of VMAT_INVARIANT vectorized stmts, fix SLP support.
---
 gcc/tree-vect-stmts.cc | 39 +++++++++++++++++++--------------------
 1 file changed, 19 insertions(+), 20 deletions(-)

diff --git a/gcc/tree-vect-stmts.cc b/gcc/tree-vect-stmts.cc
index 1b160cecfce..b1b08238dc3 100644
--- a/gcc/tree-vect-stmts.cc
+++ b/gcc/tree-vect-stmts.cc
@@ -9612,27 +9612,26 @@ vectorizable_load (vec_info *vinfo,
 	  gimple_set_vuse (new_stmt, vuse);
 	  gsi_insert_on_edge_immediate (pe, new_stmt);
 	}
-      /* These copies are all equivalent, but currently the representation
-	 requires a separate STMT_VINFO_VEC_STMT for each one.  */
-      gimple_stmt_iterator gsi2 = *gsi;
-      gsi_next (&gsi2);
-      for (j = 0; j < ncopies; j++)
+      /* These copies are all equivalent.  */
+      if (hoist_p)
+	new_temp = vect_init_vector (vinfo, stmt_info, scalar_dest,
+				     vectype, NULL);
+      else
 	{
-	  if (hoist_p)
-	    new_temp = vect_init_vector (vinfo, stmt_info, scalar_dest,
-					 vectype, NULL);
-	  else
-	    new_temp = vect_init_vector (vinfo, stmt_info, scalar_dest,
-					 vectype, &gsi2);
-	  gimple *new_stmt = SSA_NAME_DEF_STMT (new_temp);
-	  if (slp)
-	    SLP_TREE_VEC_STMTS (slp_node).quick_push (new_stmt);
-	  else
-	    {
-	      if (j == 0)
-		*vec_stmt = new_stmt;
-	      STMT_VINFO_VEC_STMTS (stmt_info).safe_push (new_stmt);
-	    }
+	  gimple_stmt_iterator gsi2 = *gsi;
+	  gsi_next (&gsi2);
+	  new_temp = vect_init_vector (vinfo, stmt_info, scalar_dest,
+				       vectype, &gsi2);
+	}
+      gimple *new_stmt = SSA_NAME_DEF_STMT (new_temp);
+      if (slp)
+	for (j = 0; j < (int) SLP_TREE_NUMBER_OF_VEC_STMTS (slp_node); ++j)
+	  SLP_TREE_VEC_STMTS (slp_node).quick_push (new_stmt);
+      else
+	{
+	  for (j = 0; j < ncopies; ++j)
+	    STMT_VINFO_VEC_STMTS (stmt_info).safe_push (new_stmt);
+	  *vec_stmt = new_stmt;
 	}
       return true;
     }
-- 
2.35.3

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2023-06-23  8:11 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-06-23  8:11 [PATCH] Optimize vector codegen for invariant loads, fix SLP support Richard Biener

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).