2018-01-09 Richard Biener <rguenther@suse.de>

	PR tree-optimization/83668
	* graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
	move prologue...
	(canonicalize_loop_form): ... here, renamed from ...
	(canonicalize_loop_closed_ssa_form): ... this and amended to
	swap successor edges for loop exit blocks to make us use
	the RPO order we need for initial schedule generation.

	* gcc.dg/graphite/pr83668.c: New testcase.


git-svn-id: https://gcc.gnu.org/svn/gcc/trunk@256381 138bc75d-0d04-0410-961f-82ee72b054a4

4 files changed, 73 insertions, 8 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index 24fd1b4187f..f908b7cc2cc 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,13 @@
+2018-01-09  Richard Biener  <rguenther@suse.de>
+
+	PR tree-optimization/83668
+	* graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
+	move prologue...
+	(canonicalize_loop_form): ... here, renamed from ...
+	(canonicalize_loop_closed_ssa_form): ... this and amended to
+	swap successor edges for loop exit blocks to make us use
+	the RPO order we need for initial schedule generation.
+
 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
 
 	PR tree-optimization/64811
diff --git a/gcc/graphite.c b/gcc/graphite.c
index e5f318d44fd..3b7f46bebd6 100644
--- a/gcc/graphite.c
+++ b/gcc/graphite.c
@@ -227,15 +227,11 @@ free_scops (vec<scop_p> scops)
 /* Transforms LOOP to the canonical loop closed SSA form.  */
 
 static void
-canonicalize_loop_closed_ssa (loop_p loop)
+canonicalize_loop_closed_ssa (loop_p loop, edge e)
 {
-  edge e = single_exit (loop);
   basic_block bb;
   gphi_iterator psi;
 
-  if (!e || (e->flags & EDGE_COMPLEX))
-    return;
-
   bb = e->dest;
 
   /* Make the loop-close PHI node BB contain only PHIs and have a
@@ -314,14 +310,34 @@ canonicalize_loop_closed_ssa (loop_p loop)
    other statements.
 
    - there exist only one phi node per definition in the loop.
+
+   In addition to that we also make sure that loop exit edges are
+   first in the successor edge vector.  This is to make RPO order
+   as computed by pre_and_rev_post_order_compute be consistent with
+   what initial schedule generation expects.
 */
 
 static void
-canonicalize_loop_closed_ssa_form (void)
+canonicalize_loop_form (void)
 {
   loop_p loop;
   FOR_EACH_LOOP (loop, LI_FROM_INNERMOST)
-    canonicalize_loop_closed_ssa (loop);
+    {
+      edge e = single_exit (loop);
+      if (!e || (e->flags & EDGE_COMPLEX))
+	continue;
+
+      canonicalize_loop_closed_ssa (loop, e);
+
+      /* If the exit is not first in the edge vector make it so.  */
+      if (e != EDGE_SUCC (e->src, 0))
+	{
+	  unsigned ei;
+	  for (ei = 0; EDGE_SUCC (e->src, ei) != e; ++ei)
+	    ;
+	  std::swap (EDGE_SUCC (e->src, ei), EDGE_SUCC (e->src, 0));
+	}
+    }
 
   /* We can end up releasing duplicate exit PHIs and also introduce
      additional copies so the cached information isn't correct anymore.  */
@@ -360,7 +376,7 @@ graphite_transform_loops (void)
   the_isl_ctx = ctx;
 
   sort_sibling_loops (cfun);
-  canonicalize_loop_closed_ssa_form ();
+  canonicalize_loop_form ();
 
   /* Print the loop structure.  */
   if (dump_file && (dump_flags & TDF_DETAILS))
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 679b22a68d4..bbf9c760647 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,8 @@
+2018-01-09  Richard Biener  <rguenther@suse.de>
+
+	PR tree-optimization/83668
+	* gcc.dg/graphite/pr83668.c: New testcase.
+
 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
 
 	PR tree-optimization/64811
diff --git a/gcc/testsuite/gcc.dg/graphite/pr83668.c b/gcc/testsuite/gcc.dg/graphite/pr83668.c
new file mode 100644
index 00000000000..22b01b8c4ae
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/graphite/pr83668.c
@@ -0,0 +1,34 @@
+/* { dg-do run } */
+/* { dg-options "-O -fno-tree-dominator-opts -fgraphite-identity" } */
+
+int a, b, c;
+int d[14];
+
+int
+main (void)
+{
+  short e;
+  char f;
+
+  for (; b >= 0; b--)
+    {
+      e = 0;
+      for (; e < 2; e++)
+	{
+	  a = 0;
+	  for (; a < 7; a++)
+	    d[a] = 1;
+	}
+      if (c)
+	{
+	  f = 0;
+	  for (; f >= 0; f--)
+	    ;
+	}
+    }
+
+  if (a != 7)
+    __builtin_abort ();
+
+  return 0;
+}