tree-vect-data-refs.c (vect_analyze_data_ref_access): Allow consecutive accesses within outer-loop with force_vectorize for...

gcc/

	* tree-vect-data-refs.c (vect_analyze_data_ref_access): Allow
	consecutive accesses within outer-loop with force_vectorize
	for references with zero step in inner-loop.

gcc/testsuite/

	* gcc.dg/vect/vect-outer-simd-1.c: New test.

From-SVN: r223993

gcc/ChangeLog

@@ -1,3 +1,9 @@
+2015-06-01  Yuri Rumyantsev  <ysrumyan@gmail.com>
+
+	* tree-vect-data-refs.c (vect_analyze_data_ref_access): Allow
+	consecutive accesses within outer-loop with force_vectorize
+	for references with zero step in inner-loop.
+
 2015-06-01  Vidya Praveen  <vidyapraveen@arm.com>
 
 	* Makefile.in: Pick up gcov-dump dependencies from gcc/ directory

gcc/testsuite/ChangeLog

@@ -1,3 +1,7 @@
+2015-06-01  Yuri Rumyantsev  <ysrumyan@gmail.com>
+
+	* gcc.dg/vect/vect-outer-simd-1.c: New test.
+
 2015-06-01  Matthew Wahab  <matthew.wahab@arm.com>
 
 	PR target/65697

gcc/testsuite/gcc.dg/vect/vect-outer-simd-1.c

@@ -0,0 +1,75 @@
+/* { dg-require-effective-target vect_simd_clones } */
+/* { dg-additional-options "-fopenmp-simd -ffast-math" } */
+#include <stdlib.h>
+#include "tree-vect.h"
+#define N 64
+
+float *px, *py;
+float *tx, *ty;
+float *x1, *z1, *t1, *t2;
+
+static void inline bar(const float cx, float cy,
+                         float *vx, float *vy)
+{
+  int j;
+    for (j = 0; j < N; ++j)
+    {
+        const float dx  = cx - px[j];
+        const float dy  = cy - py[j];
+        *vx               -= dx * tx[j];
+        *vy               -= dy * ty[j];
+    }
+}
+
+__attribute__((noinline, noclone)) void foo1 ()
+{
+  int i;
+#pragma omp simd
+  for (i=0; i<N; i++)
+    bar(px[i], py[i], x1+i, z1+i);
+}
+
+__attribute__((noinline, noclone)) void foo2 ()
+{
+  volatile int i;
+  for (i=0; i<N; i++)
+    bar(px[i], py[i], x1+i, z1+i);
+}
+
+
+int main()
+{
+  float *X = (float*)malloc(N * 8 * sizeof (float));
+  int i;
+  check_vect ();
+  px = &X[0];
+  py = &X[N * 1];
+  tx = &X[N * 2];
+  ty = &X[N * 3];
+  x1 = &X[N * 4];
+  z1 = &X[N * 5];
+  t1 = &X[N * 6];
+  t2 = &X[N * 7];
+
+  for (i=0; i<N; i++)
+    {
+      px[i] = (float) (i+2);
+      tx[i] = (float) (i+1);
+      py[i] = (float) (i+4);
+      ty[i] = (float) (i+3);
+      x1[i] = z1[i] = 1.0f;
+    }
+  foo1 ();  /* vector variant.  */
+  for (i=0; i<N;i++)
+    {
+      t1[i] = x1[i]; x1[i] = 1.0f;
+      t2[i] = z1[i]; z1[i] = 1.0f;
+    }
+  foo2 ();  /* scalar variant.  */
+  for (i=0; i<N; i++)
+    if (x1[i] != t1[i] || z1[i] != t2[i])
+      abort ();	
+  return 0;
+} 
+/* { dg-final { scan-tree-dump "OUTER LOOP VECTORIZED" "vect" } } */
+/* { dg-final { cleanup-tree-dump "vect" } } */

gcc/tree-vect-data-refs.c

@@ -2287,18 +2287,22 @@ vect_analyze_data_ref_access (struct data_reference *dr)
       return false;
     }
 
-  /* Allow invariant loads in not nested loops.  */
+  /* Allow loads with zero step in inner-loop vectorization.  */
   if (loop_vinfo && integer_zerop (step))
     {
       GROUP_FIRST_ELEMENT (vinfo_for_stmt (stmt)) = NULL;
-      if (nested_in_vect_loop_p (loop, stmt))
+      if (!nested_in_vect_loop_p (loop, stmt))
+	return DR_IS_READ (dr);
+      /* Allow references with zero step for outer loops marked
+	 with pragma omp simd only - it guarantees absence of
+	 loop-carried dependencies between inner loop iterations.  */
+      if (!loop->force_vectorize)
 	{
 	  if (dump_enabled_p ())
 	    dump_printf_loc (MSG_NOTE, vect_location,
 			     "zero step in inner loop of nest\n");
 	  return false;
 	}
-      return DR_IS_READ (dr);
     }
 
   if (loop && nested_in_vect_loop_p (loop, stmt))