backport: re PR tree-optimization/81740 (wrong code at -O3 in both 32-bit and 64...

2019-09-04  Richard Biener  <rguenther@suse.de>

	Backport from mainline
	2019-03-26  Bin Cheng  <bin.cheng@linux.alibaba.com>

	PR tree-optimization/81740
	* tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
	In case of outer loop vectorization, check for backward dependence
	at the inner loop if outer loop dependence is reversed.

	* gcc.dg/vect/pr81740-1.c: New testcase.
	* gcc.dg/vect/pr81740-2.c: Likewise.

From-SVN: r275372

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index f666940c6076ef631d7dba4671ff791e5be4d74d..475f194b71fb8efcdb7055088bae27f31b013b85 100644 (file)
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,13 @@
+2019-09-04  Richard Biener  <rguenther@suse.de>
+
+       Backport from mainline
+       2019-03-26  Bin Cheng  <bin.cheng@linux.alibaba.com>
+
+       PR tree-optimization/81740
+       * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
+       In case of outer loop vectorization, check for backward dependence
+       at the inner loop if outer loop dependence is reversed.
+
 2019-09-04  Richard Biener  <rguenther@suse.de>
 
        Backport from mainline

diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 1e848f2b3b578e86fbc961d9a3031923dfb2b91a..79b74c0b3f4c8e1670bbbec47515c2b0fb89c1fc 100644 (file)
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,12 @@
+2019-09-04  Richard Biener  <rguenther@suse.de>
+
+       Backport from mainline
+       2019-03-26  Bin Cheng  <bin.cheng@linux.alibaba.com>
+
+       PR tree-optimization/81740
+       * gcc.dg/vect/pr81740-1.c: New testcase.
+       * gcc.dg/vect/pr81740-2.c: Likewise.
+
 2019-09-04  Richard Biener  <rguenther@suse.de>
 
        Backport from mainline

diff --git a/gcc/testsuite/gcc.dg/vect/pr81740-1.c b/gcc/testsuite/gcc.dg/vect/pr81740-1.c
new file mode 100644 (file)
index 0000000..d2226fc
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/pr81740-1.c
@@ -0,0 +1,22 @@
+/* { dg-do run } */
+/* { dg-require-effective-target vect_int } */
+
+#include "tree-vect.h"
+
+int a[8][10] = { [2][5] = 4 }, c;
+
+int
+main ()
+{
+  short b;
+  int i, d;
+  check_vect ();
+  for (b = 4; b >= 0; b--)
+    for (c = 0; c <= 6; c++)
+      a[c + 1][b + 2] = a[c][b + 1];
+  for (i = 0; i < 8; i++)
+    for (d = 0; d < 10; d++)
+      if (a[i][d] != (i == 3 && d == 6) * 4)
+       __builtin_abort ();
+  return 0;
+}

diff --git a/gcc/testsuite/gcc.dg/vect/pr81740-2.c b/gcc/testsuite/gcc.dg/vect/pr81740-2.c
new file mode 100644 (file)
index 0000000..76637ad
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/pr81740-2.c
@@ -0,0 +1,24 @@
+/* { dg-do run } */
+/* { dg-require-effective-target vect_int } */
+
+#include "tree-vect.h"
+
+int a[8][10] = { [2][5] = 4 }, c;
+
+int
+main ()
+{
+  short b;
+  int i, d;
+  check_vect ();
+  for (b = 4; b >= 0; b--)
+    for (c = 6; c >= 0; c--)
+      a[c + 1][b + 2] = a[c][b + 1];
+  for (i = 0; i < 8; i++)
+    for (d = 0; d < 10; d++)
+      if (a[i][d] != (i == 3 && d == 6) * 4)
+       __builtin_abort ();
+  return 0;
+}
+
+/* { dg-final { scan-tree-dump "OUTER LOOP VECTORIZED" "vect" } } */

diff --git a/gcc/tree-vect-data-refs.c b/gcc/tree-vect-data-refs.c
index 022b7e3592cab0ee489fcdc20b3b521f4015ca2c..7915eeabac5a07642566dd2bf894a8ace63f054f 100644 (file)
--- a/gcc/tree-vect-data-refs.c
+++ b/gcc/tree-vect-data-refs.c
@@ -470,8 +470,19 @@ vect_analyze_data_ref_dependence (struct data_dependence_relation *ddr,
             reversed (to make distance vector positive), and the actual
             distance is negative.  */
          if (dump_enabled_p ())
-           dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
+           dump_printf_loc (MSG_NOTE, vect_location,
                             "dependence distance negative.\n");
+         /* When doing outer loop vectorization, we need to check if there is
+            a backward dependence at the inner loop level if the dependence
+            at the outer loop is reversed.  See PR81740.  */
+         if (nested_in_vect_loop_p (loop, DR_STMT (dra))
+             || nested_in_vect_loop_p (loop, DR_STMT (drb)))
+           {
+             unsigned inner_depth = index_in_loop_nest (loop->inner->num,
+                                                        DDR_LOOP_NEST (ddr));
+             if (dist_v[inner_depth] < 0)
+               return true;
+           }
          /* Record a negative dependence distance to later limit the
             amount of stmt copying / unrolling we can perform.
             Only need to handle read-after-write dependence.  */