re PR tree-optimization/68529 (scev failed for while(i--))

	PR tree-optimization/68529
	* tree-ssa-loop-niter.c (number_of_iterations_ne): Add new param.
	Compute no-overflow information for control iv.
	(number_of_iterations_lt, number_of_iterations_le): Add new param.
	(number_of_iterations_cond): Pass new argument to above functions.

	PR tree-optimization/68529
	* gcc.dg/tree-ssa/pr68529-1.c: New test.
	* gcc.dg/tree-ssa/pr68529-2.c: New test.
	* gcc.dg/tree-ssa/pr68529-3.c: New test.

From-SVN: r231097

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index a1c1962..5f675f1 100644 (file)
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,11 @@
+2015-12-01  Bin Cheng  <bin.cheng@arm.com>
+
+       PR tree-optimization/68529
+       * tree-ssa-loop-niter.c (number_of_iterations_ne): Add new param.
+       Compute no-overflow information for control iv.
+       (number_of_iterations_lt, number_of_iterations_le): Add new param.
+       (number_of_iterations_cond): Pass new argument to above functions.
+
 2015-11-30  Jan Hubicka  <hubicka@ucw.cz>
 
        * ipa-inline-transform.c (inline_call): Drop -fstrict-aliasing when

diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 1a86d87..6e17cc4 100644 (file)
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,10 @@
+2015-12-01  Bin Cheng  <bin.cheng@arm.com>
+
+       PR tree-optimization/68529
+       * gcc.dg/tree-ssa/pr68529-1.c: New test.
+       * gcc.dg/tree-ssa/pr68529-2.c: New test.
+       * gcc.dg/tree-ssa/pr68529-3.c: New test.
+
 2015-11-30  Jan Hubicka  <hubicka@ucw.cz>
 
        * gcc.dg/lto/alias-1_0.c: New testcase.

diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr68529-1.c b/gcc/testsuite/gcc.dg/tree-ssa/pr68529-1.c
new file mode 100644 (file)
index 0000000..9dcc22f
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr68529-1.c
@@ -0,0 +1,20 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -ftree-loop-distribution -ftree-loop-distribute-patterns -fdump-tree-ldist-details" } */
+
+void bar(char *s);
+int foo()
+{
+  char c[10000] = {};
+  unsigned short nchar = 9999;
+
+  while(nchar-- != 0)
+    {
+      c[nchar] = 'A';
+    }
+
+  bar (c);
+  return 0;
+}
+
+/* { dg-final { scan-tree-dump "distributed: split to 0 loops and 1 library calls" "ldist" } } */
+/* { dg-final { scan-tree-dump "generated memset" "ldist" } } */

diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr68529-2.c b/gcc/testsuite/gcc.dg/tree-ssa/pr68529-2.c
new file mode 100644 (file)
index 0000000..6d57618
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr68529-2.c
@@ -0,0 +1,23 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -ftree-loop-distribution -ftree-loop-distribute-patterns -fdump-tree-ldist-details" } */
+
+void bar(char *s);
+int foo(unsigned short l)
+{
+  char c[10000] = {};
+  unsigned short nchar = 9999;
+
+  if (nchar <= l)
+    return -1;
+
+  while(nchar-- != l)
+    {
+      c[nchar] = 'A';
+    }
+
+  bar (c);
+  return 0;
+}
+
+/* { dg-final { scan-tree-dump "distributed: split to 0 loops and 1 library calls" "ldist" } } */
+/* { dg-final { scan-tree-dump "generated memset" "ldist" } } */

diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr68529-3.c b/gcc/testsuite/gcc.dg/tree-ssa/pr68529-3.c
new file mode 100644 (file)
index 0000000..583b4a6
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr68529-3.c
@@ -0,0 +1,47 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -ftree-loop-distribution -ftree-loop-distribute-patterns -fdump-tree-ldist-details" } */
+
+void bar(char *s);
+int foo1(unsigned short l)
+{
+  char c[10000] = {};
+  unsigned short nchar = 9999;
+
+  while(nchar-- != l)
+    {
+      c[nchar] = 'A';
+    }
+
+  bar (c);
+  return 0;
+}
+
+int foo2()
+{
+  char c[100000] = {};
+  unsigned short nchar;
+
+  for (nchar = 0; nchar != 1000; --nchar)
+    {
+      c[nchar] = 'A';
+    }
+
+  bar (c);
+  return 0;
+}
+
+int foo3()
+{
+  char c[100000] = {};
+  unsigned short nchar;
+
+  for (nchar = 0; nchar != 1000; nchar += 3)
+    {
+      c[nchar] = 'A';
+    }
+
+  bar (c);
+  return 0;
+}
+
+/* { dg-final { scan-tree-dump-times "failed: evolution of offset is not affine" 3 "ldist" } } */

diff --git a/gcc/tree-ssa-loop-niter.c b/gcc/tree-ssa-loop-niter.c
index 6d480c0..67080a3 100644 (file)
--- a/gcc/tree-ssa-loop-niter.c
+++ b/gcc/tree-ssa-loop-niter.c
@@ -957,13 +957,14 @@ number_of_iterations_ne_max (mpz_t bnd, bool no_overflow, tree c, tree s,
    bounds on the difference FINAL - IV->base.  */
 
 static bool
-number_of_iterations_ne (tree type, affine_iv *iv, tree final,
-                        struct tree_niter_desc *niter, bool exit_must_be_taken,
-                        bounds *bnds)
+number_of_iterations_ne (struct loop *loop, tree type, affine_iv *iv,
+                        tree final, struct tree_niter_desc *niter,
+                        bool exit_must_be_taken, bounds *bnds)
 {
   tree niter_type = unsigned_type_for (type);
   tree s, c, d, bits, assumption, tmp, bound;
   mpz_t max;
+  tree e;
 
   niter->control = *iv;
   niter->bound = final;
@@ -998,6 +999,23 @@ number_of_iterations_ne (tree type, affine_iv *iv, tree final,
                                 TYPE_SIGN (niter_type));
   mpz_clear (max);
 
+  /* Compute no-overflow information for the control iv.  Note we are
+     handling NE_EXPR, if iv base equals to final value, the loop exits
+     immediately, and the iv does not overflow.  */
+  if (tree_int_cst_sign_bit (iv->step))
+    e = fold_build2 (GE_EXPR, boolean_type_node, iv->base, final);
+  else
+    e = fold_build2 (LE_EXPR, boolean_type_node, iv->base, final);
+  e = simplify_using_initial_conditions (loop, e);
+  if (integer_onep (e)
+      && (integer_onep (s)
+         || (TREE_CODE (c) == INTEGER_CST
+             && TREE_CODE (s) == INTEGER_CST
+             && wi::mod_trunc (c, s, TYPE_SIGN (type)) == 0)))
+    {
+      niter->control.no_overflow = true;
+    }
+
   /* First the trivial cases -- when the step is 1.  */
   if (integer_onep (s))
     {
@@ -1361,8 +1379,8 @@ assert_loop_rolls_lt (tree type, affine_iv *iv0, affine_iv *iv1,
    that the exit must be taken eventually.  */
 
 static bool
-number_of_iterations_lt (tree type, affine_iv *iv0, affine_iv *iv1,
-                        struct tree_niter_desc *niter,
+number_of_iterations_lt (struct loop *loop, tree type, affine_iv *iv0,
+                        affine_iv *iv1, struct tree_niter_desc *niter,
                         bool exit_must_be_taken, bounds *bnds)
 {
   tree niter_type = unsigned_type_for (type);
@@ -1434,7 +1452,8 @@ number_of_iterations_lt (tree type, affine_iv *iv0, affine_iv *iv1,
         zps does not overflow.  */
       zps.no_overflow = true;
 
-      return number_of_iterations_ne (type, &zps, delta, niter, true, bnds);
+      return number_of_iterations_ne (loop, type, &zps,
+                                     delta, niter, true, bnds);
     }
 
   /* Make sure that the control iv does not overflow.  */
@@ -1473,9 +1492,9 @@ number_of_iterations_lt (tree type, affine_iv *iv0, affine_iv *iv1,
    is the case).  BNDS bounds the difference IV1->base - IV0->base.  */
 
 static bool
-number_of_iterations_le (tree type, affine_iv *iv0, affine_iv *iv1,
-                        struct tree_niter_desc *niter, bool exit_must_be_taken,
-                        bounds *bnds)
+number_of_iterations_le (struct loop *loop, tree type, affine_iv *iv0,
+                        affine_iv *iv1, struct tree_niter_desc *niter,
+                        bool exit_must_be_taken, bounds *bnds)
 {
   tree assumption;
   tree type1 = type;
@@ -1523,7 +1542,7 @@ number_of_iterations_le (tree type, affine_iv *iv0, affine_iv *iv1,
 
   bounds_add (bnds, 1, type1);
 
-  return number_of_iterations_lt (type, iv0, iv1, niter, exit_must_be_taken,
+  return number_of_iterations_lt (loop, type, iv0, iv1, niter, exit_must_be_taken,
                                  bnds);
 }
 
@@ -1698,18 +1717,18 @@ number_of_iterations_cond (struct loop *loop,
     {
     case NE_EXPR:
       gcc_assert (integer_zerop (iv1->step));
-      ret = number_of_iterations_ne (type, iv0, iv1->base, niter,
+      ret = number_of_iterations_ne (loop, type, iv0, iv1->base, niter,
                                     exit_must_be_taken, &bnds);
       break;
 
     case LT_EXPR:
-      ret = number_of_iterations_lt (type, iv0, iv1, niter, exit_must_be_taken,
-                                    &bnds);
+      ret = number_of_iterations_lt (loop, type, iv0, iv1, niter,
+                                    exit_must_be_taken, &bnds);
       break;
 
     case LE_EXPR:
-      ret = number_of_iterations_le (type, iv0, iv1, niter, exit_must_be_taken,
-                                    &bnds);
+      ret = number_of_iterations_le (loop, type, iv0, iv1, niter,
+                                    exit_must_be_taken, &bnds);
       break;
 
     default: