public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc/devel/autopar_devel] tree-optimization/92260 - improve fix
@ 2020-08-22 21:17 Giuliano Belinassi
  0 siblings, 0 replies; only message in thread
From: Giuliano Belinassi @ 2020-08-22 21:17 UTC (permalink / raw)
  To: gcc-cvs

https://gcc.gnu.org/g:d8e6d84b3c915eb485aadd1de48045fdbd5b112b

commit d8e6d84b3c915eb485aadd1de48045fdbd5b112b
Author: Richard Biener <rguenther@suse.de>
Date:   Fri May 15 11:14:53 2020 +0200

    tree-optimization/92260 - improve fix
    
    This improves the fix for PR92260 changing the number of vector
    computation to the canonical one, not needing to look at the
    using stmt.
    
    2020-05-15  Richard Biener  <rguenther@suse.de>
    
            PR tree-optimization/92260
            * tree-vect-slp.c (vect_get_constant_vectors): Compute
            the number of vector stmts in a canonical way.

Diff:
---
 gcc/ChangeLog       |  6 ++++++
 gcc/tree-vect-slp.c | 19 ++++++-------------
 2 files changed, 12 insertions(+), 13 deletions(-)

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index cc761f2c6b6..5eaa1b8af27 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,9 @@
+2020-05-15  Richard Biener  <rguenther@suse.de>
+
+	PR tree-optimization/92260
+	* tree-vect-slp.c (vect_get_constant_vectors): Compute
+	the number of vector stmts in a canonical way.
+
 2020-05-15  Martin Liska  <mliska@suse.cz>
 
 	* hsa-gen.c (get_symbol_for_decl): Fix misleading indentation
diff --git a/gcc/tree-vect-slp.c b/gcc/tree-vect-slp.c
index 6f623955ce5..15eea74d8d1 100644
--- a/gcc/tree-vect-slp.c
+++ b/gcc/tree-vect-slp.c
@@ -3561,19 +3561,12 @@ vect_get_constant_vectors (vec_info *vinfo,
   else
     vector_type = get_vectype_for_scalar_type (vinfo, TREE_TYPE (op), op_node);
 
-  /* ???  For lane-reducing ops we should also have the required number
-     of vector stmts initialized rather than second-guessing here.  */
-  unsigned int number_of_vectors;
-  if (is_gimple_assign (stmt_vinfo->stmt)
-      && (gimple_assign_rhs_code (stmt_vinfo->stmt) == SAD_EXPR
-	  || gimple_assign_rhs_code (stmt_vinfo->stmt) == DOT_PROD_EXPR
-	  || gimple_assign_rhs_code (stmt_vinfo->stmt) == WIDEN_SUM_EXPR))
-    number_of_vectors = SLP_TREE_NUMBER_OF_VEC_STMTS (slp_node);
-  else
-    number_of_vectors
-      = vect_get_num_vectors (SLP_TREE_NUMBER_OF_VEC_STMTS (slp_node)
-			      * TYPE_VECTOR_SUBPARTS (stmt_vectype),
-			      vector_type);
+  poly_uint64 vf = 1;
+  if (loop_vec_info loop_vinfo = dyn_cast <loop_vec_info> (vinfo))
+    vf = loop_vinfo->vectorization_factor;
+  unsigned int number_of_vectors
+    = vect_get_num_vectors (vf * group_size, vector_type);
+
   vec_oprnds->create (number_of_vectors);
   auto_vec<tree> voprnds (number_of_vectors);


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

only message in thread, other threads:[~2020-08-22 21:17 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-08-22 21:17 [gcc/devel/autopar_devel] tree-optimization/92260 - improve fix Giuliano Belinassi

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).