Blob Blame History Raw
2019-11-22  Jonathan Wakely  <jwakely@redhat.com>

	Backport from mainline
	2019-10-29  Jonathan Wakely  <jwakely@redhat.com>

	PR libstdc++/92267
	* include/bits/stl_deque.h (_Deque_iterator(const _Deque_iterator&)):
	Do not define as defaulted.
	* testsuite/23_containers/deque/types/92267.cc: New test.

2019-11-21  Jakub Jelinek  <jakub@redhat.com>

	PR tree-optimization/91355
	* tree-ssa-sink.c (select_best_block): Use >= rather than >
	for early_bb scaled count with best_bb count comparison.

2019-11-21  Richard Biener  <rguenther@suse.de>

	Revert
	2019-09-17  Richard Biener  <rguenther@suse.de>

	PR tree-optimization/91790
	* tree-vect-stmts.c (vectorizable_load): For BB vectorization
	use the correct DR for setting up realignment.

2019-11-20  Peter Bergner <bergner@linux.ibm.com>

	Backport from mainline
	2019-11-07  Peter Bergner <bergner@linux.ibm.com>

	PR other/92090
	* config/rs6000/predicates.md (input_operand): Allow MODE_PARTIAL_INT
	modes for integer constants.

2019-11-20  Michael Matz  <matz@suse.de>

	Backport from mainline
	PR middle-end/90796
	* gimple-loop-jam.c (any_access_function_variant_p): New function.
	(adjust_unroll_factor): Use it to constrain safety, new parameter.
	(tree_loop_unroll_and_jam): Adjust call and profitable unroll factor.

2019-11-20  Joseph Myers  <joseph@codesourcery.com>

	* doc/invoke.texi (-Wc11-c2x-compat): Document.

--- libstdc++-v3/include/bits/stl_deque.h	(revision 278492)
+++ libstdc++-v3/include/bits/stl_deque.h	(revision 278614)
@@ -158,13 +158,16 @@
 #else
       // Conversion from iterator to const_iterator.
       template<typename _Iter,
-              typename = _Require<is_same<_Self, const_iterator>,
-                                  is_same<_Iter, iterator>>>
+	       typename = _Require<is_same<_Self, const_iterator>,
+				   is_same<_Iter, iterator>>>
        _Deque_iterator(const _Iter& __x) noexcept
        : _M_cur(__x._M_cur), _M_first(__x._M_first),
-         _M_last(__x._M_last), _M_node(__x._M_node) { }
+	 _M_last(__x._M_last), _M_node(__x._M_node) { }
 
-      _Deque_iterator(const _Deque_iterator&) = default;
+      _Deque_iterator(const _Deque_iterator& __x) noexcept
+       : _M_cur(__x._M_cur), _M_first(__x._M_first),
+	 _M_last(__x._M_last), _M_node(__x._M_node) { }
+
       _Deque_iterator& operator=(const _Deque_iterator&) = default;
 #endif
 
--- libstdc++-v3/testsuite/23_containers/deque/types/92267.cc	(nonexistent)
+++ libstdc++-v3/testsuite/23_containers/deque/types/92267.cc	(revision 278614)
@@ -0,0 +1,27 @@
+// Copyright (C) 2019 Free Software Foundation, Inc.
+//
+// This file is part of the GNU ISO C++ Library.  This library is free
+// software; you can redistribute it and/or modify it under the
+// terms of the GNU General Public License as published by the
+// Free Software Foundation; either version 3, or (at your option)
+// any later version.
+
+// This library is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+
+// You should have received a copy of the GNU General Public License along
+// with this library; see the file COPYING3.  If not see
+// <http://www.gnu.org/licenses/>.
+
+// { dg-do compile { target c++11 } }
+
+#include <deque>
+
+using std::deque;
+using std::is_trivially_copy_constructible;
+
+// PR libstdc++/92267
+static_assert(!is_trivially_copy_constructible<deque<int>::iterator>::value);
+static_assert(!is_trivially_copy_constructible<deque<int>::const_iterator>::value);
--- gcc/doc/invoke.texi	(revision 278492)
+++ gcc/doc/invoke.texi	(revision 278614)
@@ -292,6 +292,7 @@
 -Wbool-compare  -Wbool-operation @gol
 -Wno-builtin-declaration-mismatch @gol
 -Wno-builtin-macro-redefined  -Wc90-c99-compat  -Wc99-c11-compat @gol
+-Wc11-c2x-compat @gol
 -Wc++-compat  -Wc++11-compat  -Wc++14-compat  -Wc++17-compat  @gol
 -Wcast-align  -Wcast-align=strict  -Wcast-function-type  -Wcast-qual  @gol
 -Wchar-subscripts  -Wcatch-value  -Wcatch-value=@var{n} @gol
@@ -6698,6 +6699,14 @@
 and so on.  This option is independent of the standards mode.  Warnings are
 disabled in the expression that follows @code{__extension__}.
 
+@item -Wc11-c2x-compat @r{(C and Objective-C only)}
+@opindex Wc11-c2x-compat
+@opindex Wno-c11-c2x-compat
+Warn about features not present in ISO C11, but present in ISO C2X.
+For instance, warn about omitting the string in @code{_Static_assert}.
+This option is independent of the standards mode.  Warnings are
+disabled in the expression that follows @code{__extension__}.
+
 @item -Wc++-compat @r{(C and Objective-C only)}
 @opindex Wc++-compat
 @opindex Wno-c++-compat
--- gcc/testsuite/gcc.target/powerpc/pr92090-2.c	(nonexistent)
+++ gcc/testsuite/gcc.target/powerpc/pr92090-2.c	(revision 278614)
@@ -0,0 +1,45 @@
+/* { dg-do compile } */
+/* { dg-options "-mdejagnu-cpu=power8 -Os -w" } */
+/* { dg-additional-options "-mbig" { target powerpc64le-*-* } } */
+
+/* Verify that we don't ICE.  */
+
+int a;
+static _Atomic long double b, c, d, m;
+double n;
+extern int foo (void);
+extern void bar (int, int, int, int);
+
+void
+bug (void)
+{
+  b = 1.79769313486231580793728971405301199e308L;
+  for (int i = 0; i < 10000; i++)
+    if (__builtin_isinf (n))
+      b;
+  c = 1;
+  int e, f, g, h;
+  while (a)
+    ;
+  for (int i; i; i++)
+    {
+      double j = c /= foo ();
+      if (__builtin_isinf (j))
+	{
+	  if (foo == 1 << 31)
+	    e++;
+	  f++;
+	  c = 0;
+	}
+      else
+	{
+	  if (foo == 1 << 30)
+	    g++;
+	  h++;
+	  c = 1;
+	}
+    }
+  bar (e, f, g, h);
+  d = 1.79769313486231580793728971405301199e308L;
+  m = 1;
+}
--- gcc/testsuite/gcc.target/powerpc/pr92090.c	(nonexistent)
+++ gcc/testsuite/gcc.target/powerpc/pr92090.c	(revision 278614)
@@ -0,0 +1,43 @@
+/* { dg-do compile } */
+/* { dg-options "-mdejagnu-cpu=power8 -Os" } */
+/* { dg-additional-options "-mbig" { target powerpc64le-*-* } } */
+
+/* Verify that we don't ICE.  */
+
+_Atomic int a;
+_Atomic long double b, c;
+int j;
+void foo (void);
+void bar (int, int, int, int);
+
+void
+bug (void)
+{
+  b = 1;
+  int d, e, f, g;
+  while (a)
+    ;
+  for (int h = 0; h < 10000; h++)
+    {
+      double i = b /= 3;
+      foo ();
+      if (i)
+	{
+	  if (i == 1)
+	    d++;
+	  e++;
+	  b = 0;
+	}
+      else
+	{
+	  if (i == 2)
+	    f++;
+	  g++;
+	  b = 1;
+	}
+    }
+  bar (d, e, f, g);
+  c = 1;
+  for (int h; h; h++)
+    j = 0;
+}
--- gcc/testsuite/gcc.dg/unroll-and-jam.c	(revision 278492)
+++ gcc/testsuite/gcc.dg/unroll-and-jam.c	(revision 278614)
@@ -1,5 +1,5 @@
 /* { dg-do run } */
-/* { dg-options "-O3 -floop-unroll-and-jam --param unroll-jam-min-percent=0 -fdump-tree-unrolljam-details" } */
+/* { dg-options "-O3 -floop-unroll-and-jam -fno-tree-loop-im --param unroll-jam-min-percent=0 -fdump-tree-unrolljam-details" } */
 /* { dg-require-effective-target int32plus } */
 
 #include <stdio.h>
@@ -34,7 +34,7 @@
 #define TEST(name, body, test) \
 static void __attribute__((noinline,noclone)) name (unsigned long n, unsigned long m) \
 { \
-  unsigned long i, j; \
+  unsigned i, j; \
   for (i = 1; i < m; i++) { \
       for (j = 1; j < n; j++) { \
 	  body; \
@@ -58,9 +58,14 @@
 TEST(foo4, aa[i][j] = aa[i-1][j+1] * aa[i-1][j+1] / 2, checkaa()) //notok, -1,1
 TEST(foo5, aa[i][j] = aa[i+1][j+1] * aa[i+1][j+1] / 2, checkaa()) //ok, 1,1
 TEST(foo6, aa[i][j] = aa[i+1][j] * aa[i+1][j] / 2, checkaa()) //ok, -1,0
+TEST(foo61, aa[i][0] = aa[i+1][0] * aa[i+1][0] / 2, checkaa()) //notok, -1,0
+TEST(foo62, aa[i][j/2] = aa[i+1][j/2] * aa[i+1][j/2] / 2, checkaa()) //notok, not affine
+TEST(foo63, aa[i][j%2] = aa[i+1][j%2] * aa[i+1][j%2] / 2, checkaa()) //notok, not affine
 TEST(foo7, aa[i+1][j] = aa[i][j] * aa[i][j] / 2, checkaa()) //ok, 1,0
 TEST(foo9, b[j] = 3*b[j+1] + 1, checkb()) //notok, 0,-1
 TEST(foo10, b[j] = 3*b[j] + 1, checkb()) //ok, 0,0
+extern int f;
+TEST(foo11, f = b[i-1] = 1 + 3* b[i+1], checkb()) //ok, 2,0 but must reduce unroll factor to 2, (it would be incorrect with unroll-by-3, which the profitability would suggest)
 
 /* foo8 should work as well, but currently doesn't because the distance
    vectors we compute are too pessimistic.  We compute
@@ -68,6 +73,7 @@
    and the last one causes us to lose.  */
 TEST(foo8, b[j+1] = 3*b[j] + 1, checkb()) //ok, 0,1
 
+int f;
 unsigned int a[1024];
 unsigned int b[1024];
 unsigned int aa[16][1024];
@@ -88,10 +94,12 @@
     printf(" %s\n", #name); \
     init();for(i=0;i<4;i++)name##noopt(32,8); checka = checksum; \
     init();for(i=0;i<4;i++)name(32,8); \
+    if (checka != checksum) fail = 1; \
     printf("%sok %s\n", checka != checksum ? "NOT " : "", #name);
 
 int main()
 {
+  int fail = 0;
   int i;
   unsigned checka;
   RUN(foo1);
@@ -100,12 +108,18 @@
   RUN(foo4);
   RUN(foo5);
   RUN(foo6);
+  RUN(foo61);
+  RUN(foo62);
+  RUN(foo63);
   RUN(foo7);
   RUN(foo8);
   RUN(foo9);
   RUN(foo10);
-  return 0;
+  RUN(foo11);
+  if (fail)
+    __builtin_abort();
+  return fail;
 }
 
-/* Five loops should be unroll-jammed (actually six, but see above).  */
-/* { dg-final { scan-tree-dump-times "applying unroll and jam" 5 "unrolljam" } } */
+/* Six loops should be unroll-jammed (actually seven, but see above).  */
+/* { dg-final { scan-tree-dump-times "applying unroll and jam" 6 "unrolljam" } } */
--- gcc/testsuite/g++.dg/torture/pr91355.C	(nonexistent)
+++ gcc/testsuite/g++.dg/torture/pr91355.C	(revision 278614)
@@ -0,0 +1,28 @@
+// PR tree-optimization/91355
+// { dg-do run }
+// { dg-options "-std=c++14" }
+
+unsigned int d = 0;
+
+struct S {
+  S () { d++; }
+  S (const S &) { d++; }
+  ~S () { d--; }
+};
+
+void
+foo (int i) throw (int) // { dg-warning "dynamic exception specifications are deprecated" }
+{
+  if (i == 0)
+    throw 3;
+  S d;
+  throw 3;
+}
+
+int
+main ()
+{
+  try { foo (1); } catch (...) {}
+  if (d)
+    __builtin_abort ();
+}
--- gcc/tree-ssa-sink.c	(revision 278492)
+++ gcc/tree-ssa-sink.c	(revision 278614)
@@ -229,7 +229,7 @@
       /* If result of comparsion is unknown, preffer EARLY_BB.
 	 Thus use !(...>=..) rather than (...<...)  */
       && !(best_bb->count.apply_scale (100, 1)
-	   > (early_bb->count.apply_scale (threshold, 1))))
+	   >= early_bb->count.apply_scale (threshold, 1)))
     return best_bb;
 
   /* No better block found, so return EARLY_BB, which happens to be the
--- gcc/tree-vect-stmts.c	(revision 278492)
+++ gcc/tree-vect-stmts.c	(revision 278614)
@@ -8276,9 +8276,7 @@
        || alignment_support_scheme == dr_explicit_realign)
       && !compute_in_loop)
     {
-      msq = vect_setup_realignment (first_stmt_info_for_drptr
-				    ? first_stmt_info_for_drptr
-				    : first_stmt_info, gsi, &realignment_token,
+      msq = vect_setup_realignment (first_stmt_info, gsi, &realignment_token,
 				    alignment_support_scheme, NULL_TREE,
 				    &at_loop);
       if (alignment_support_scheme == dr_explicit_realign_optimized)
--- gcc/gimple-loop-jam.c	(revision 278492)
+++ gcc/gimple-loop-jam.c	(revision 278614)
@@ -360,9 +360,26 @@
   rewrite_into_loop_closed_ssa_1 (NULL, 0, SSA_OP_USE, loop);
 }
 
+/* Return true if any of the access functions for dataref A
+   isn't invariant with respect to loop LOOP_NEST.  */
+static bool
+any_access_function_variant_p (const struct data_reference *a,
+			       const class loop *loop_nest)
+{
+  unsigned int i;
+  vec<tree> fns = DR_ACCESS_FNS (a);
+  tree t;
+
+  FOR_EACH_VEC_ELT (fns, i, t)
+    if (!evolution_function_is_invariant_p (t, loop_nest->num))
+      return true;
+
+  return false;
+}
+
 /* Returns true if the distance in DDR can be determined and adjusts
    the unroll factor in *UNROLL to make unrolling valid for that distance.
-   Otherwise return false.
+   Otherwise return false.  DDR is with respect to the outer loop of INNER.
 
    If this data dep can lead to a removed memory reference, increment
    *REMOVED and adjust *PROFIT_UNROLL to be the necessary unroll factor
@@ -369,7 +386,7 @@
    for this to happen.  */
 
 static bool
-adjust_unroll_factor (struct data_dependence_relation *ddr,
+adjust_unroll_factor (class loop *inner, struct data_dependence_relation *ddr,
 		      unsigned *unroll, unsigned *profit_unroll,
 		      unsigned *removed)
 {
@@ -392,9 +409,59 @@
 	    gcc_unreachable ();
 	  else if ((unsigned)dist >= *unroll)
 	    ;
-	  else if (lambda_vector_lexico_pos (dist_v + 1, DDR_NB_LOOPS (ddr) - 1)
-		   || (lambda_vector_zerop (dist_v + 1, DDR_NB_LOOPS (ddr) - 1)
-		       && dist > 0))
+	  else if (lambda_vector_zerop (dist_v + 1, DDR_NB_LOOPS (ddr) - 1))
+	    {
+	      /* We have (a,0) with a < N, so this will be transformed into
+	         (0,0) after unrolling by N.  This might potentially be a
+		 problem, if it's not a read-read dependency.  */
+	      if (DR_IS_READ (DDR_A (ddr)) && DR_IS_READ (DDR_B (ddr)))
+		;
+	      else
+		{
+		  /* So, at least one is a write, and we might reduce the
+		     distance vector to (0,0).  This is still no problem
+		     if both data-refs are affine with respect to the inner
+		     loops.  But if one of them is invariant with respect
+		     to an inner loop our reordering implicit in loop fusion
+		     corrupts the program, as our data dependences don't
+		     capture this.  E.g. for:
+		       for (0 <= i < n)
+		         for (0 <= j < m)
+		           a[i][0] = a[i+1][0] + 2;    // (1)
+		           b[i][j] = b[i+1][j] + 2;    // (2)
+		     the distance vector for both statements is (-1,0),
+		     but exchanging the order for (2) is okay, while
+		     for (1) it is not.  To see this, write out the original
+		     accesses (assume m is 2):
+		       a i j original
+		       0 0 0 r a[1][0] b[1][0]
+		       1 0 0 w a[0][0] b[0][0]
+		       2 0 1 r a[1][0] b[1][1]
+		       3 0 1 w a[0][0] b[0][1]
+		       4 1 0 r a[2][0] b[2][0]
+		       5 1 0 w a[1][0] b[1][0]
+		     after unroll-by-2 and fusion the accesses are done in
+		     this order (from column a): 0,1, 4,5, 2,3, i.e. this:
+		       a i j transformed
+		       0 0 0 r a[1][0] b[1][0]
+		       1 0 0 w a[0][0] b[0][0]
+		       4 1 0 r a[2][0] b[2][0]
+		       5 1 0 w a[1][0] b[1][0]
+		       2 0 1 r a[1][0] b[1][1]  
+		       3 0 1 w a[0][0] b[0][1]
+		     Note how access 2 accesses the same element as access 5
+		     for array 'a' but not for array 'b'.  */
+		  if (any_access_function_variant_p (DDR_A (ddr), inner)
+		      && any_access_function_variant_p (DDR_B (ddr), inner))
+		    ;
+		  else
+		    /* And if any dataref of this pair is invariant with
+		       respect to the inner loop, we have no chance than
+		       to reduce the unroll factor.  */
+		    *unroll = dist;
+		}
+	    }
+	  else if (lambda_vector_lexico_pos (dist_v + 1, DDR_NB_LOOPS (ddr) - 1))
 	    ;
 	  else
 	    *unroll = dist;
@@ -486,7 +553,7 @@
 	  /* Now check the distance vector, for determining a sensible
 	     outer unroll factor, and for validity of merging the inner
 	     loop copies.  */
-	  if (!adjust_unroll_factor (ddr, &unroll_factor, &profit_unroll,
+	  if (!adjust_unroll_factor (loop, ddr, &unroll_factor, &profit_unroll,
 				     &removed))
 	    {
 	      /* Couldn't get the distance vector.  For two reads that's
@@ -506,7 +573,7 @@
 	 to ignore all profitability concerns and apply the transformation
 	 always.  */
       if (!PARAM_VALUE (PARAM_UNROLL_JAM_MIN_PERCENT))
-	profit_unroll = 2;
+	profit_unroll = MAX(2, profit_unroll);
       else if (removed * 100 / datarefs.length ()
 	  < (unsigned)PARAM_VALUE (PARAM_UNROLL_JAM_MIN_PERCENT))
 	profit_unroll = 1;
--- gcc/config/rs6000/predicates.md	(revision 278492)
+++ gcc/config/rs6000/predicates.md	(revision 278614)
@@ -1053,8 +1053,7 @@
     return 1;
 
   /* Allow any integer constant.  */
-  if (GET_MODE_CLASS (mode) == MODE_INT
-      && CONST_SCALAR_INT_P (op))
+  if (SCALAR_INT_MODE_P (mode) && CONST_SCALAR_INT_P (op))
     return 1;
 
   /* Allow easy vector constants.  */