diff mbox series

tree-optimization/98291 - allow SLP more vectorization of reductions

Message ID nycvar.YFH.7.76.2101041047120.20213@elmra.sevgm.obk
State New
Headers show
Series tree-optimization/98291 - allow SLP more vectorization of reductions | expand

Commit Message

Richard Biener Jan. 4, 2021, 9:47 a.m. UTC 
When the VF is one a SLP reduction is in-order and thus we can
vectorize even when the reduction op is not associative.

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

2021-01-04  Richard Biener  <rguenther@suse.de>

	PR tree-optimization/98291
	* tree-vect-loop.c (vectorizable_reduction): Bypass
	associativity check for SLP reductions with VF 1.

	* gcc.dg/vect/slp-reduc-11.c: New testcase.
	* gcc.dg/vect/vect-reduc-in-order-4.c: Adjust.
---
 gcc/testsuite/gcc.dg/vect/slp-reduc-11.c      | 20 +++++++++++++++++++
 .../gcc.dg/vect/vect-reduc-in-order-4.c       |  2 --
 gcc/tree-vect-loop.c                          | 10 ++++++++--
 3 files changed, 28 insertions(+), 4 deletions(-)
 create mode 100644 gcc/testsuite/gcc.dg/vect/slp-reduc-11.c
diff mbox series

Patch

diff --git a/gcc/testsuite/gcc.dg/vect/slp-reduc-11.c b/gcc/testsuite/gcc.dg/vect/slp-reduc-11.c
new file mode 100644
index 00000000000..a2f86fb8d66
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/slp-reduc-11.c
@@ -0,0 +1,20 @@ 
+/* { dg-do compile } */
+/* { dg-require-effective-target vect_double } */
+
+double dotprod(const double *a, const double *b, unsigned long long n)
+{
+  double d1 = 0.0;
+  double d2 = 0.0;
+
+  for (unsigned long long i = 0; i < n; i += 2) {
+    d1 += a[i] * b[i];
+    d2 += a[i + 1] * b[i + 1];
+  }
+
+  return (d1 + d2);
+}
+
+/* We should use a SLP reduction even without -ffast-math by using a
+   VF of one.  */
+/* { dg-final { scan-tree-dump "vectorizing stmts using SLP" "vect" } } */
+/* { dg-final { scan-tree-dump "vectorized 1 loops" "vect" } } */
diff --git a/gcc/testsuite/gcc.dg/vect/vect-reduc-in-order-4.c b/gcc/testsuite/gcc.dg/vect/vect-reduc-in-order-4.c
index 7706a2dc5b2..eff3994a335 100644
--- a/gcc/testsuite/gcc.dg/vect/vect-reduc-in-order-4.c
+++ b/gcc/testsuite/gcc.dg/vect/vect-reduc-in-order-4.c
@@ -41,6 +41,4 @@  main ()
   return 0;
 }
 
-/* { dg-final { scan-tree-dump {in-order unchained SLP reductions not supported} "vect" } } */
-/* { dg-final { scan-tree-dump-not {vectorizing stmts using SLP} "vect" } } */
 /* { dg-final { scan-tree-dump-times "VECT_PERM_EXPR" 0 "vect" } } */
diff --git a/gcc/tree-vect-loop.c b/gcc/tree-vect-loop.c
index d6f1ffcd386..4f5e3fe20cb 100644
--- a/gcc/tree-vect-loop.c
+++ b/gcc/tree-vect-loop.c
@@ -6868,8 +6868,14 @@  vectorizable_reduction (loop_vec_info loop_vinfo,
 	 cases, so we need to check that this is ok.  One exception is when
 	 vectorizing an outer-loop: the inner-loop is executed sequentially,
 	 and therefore vectorizing reductions in the inner-loop during
-	 outer-loop vectorization is safe.  */
-      if (needs_fold_left_reduction_p (scalar_type, orig_code))
+	 outer-loop vectorization is safe.  Likewise when we are vectorizing
+	 a series of reductions using SLP and the VF is one the reductions
+	 are performed in scalar order.  */
+      if (slp_node
+	  && !REDUC_GROUP_FIRST_ELEMENT (stmt_info)
+	  && known_eq (LOOP_VINFO_VECT_FACTOR (loop_vinfo), 1u))
+	;
+      else if (needs_fold_left_reduction_p (scalar_type, orig_code))
 	{
 	  /* When vectorizing a reduction chain w/o SLP the reduction PHI
 	     is not directy used in stmt.  */