re PR tree-optimization/28364 (poor optimization choices when iterating over a…

re PR tree-optimization/28364 (poor optimization choices when iterating over a std::string (probably not c++-specific))

2018-04-30  Richard Biener  <rguenther@suse.de>

	PR tree-optimization/28364
	PR tree-optimization/85275
	* tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
	copying first exit test.

	* gcc.dg/tree-ssa/copy-headers-5.c: New testcase.
	* gcc.dg/tree-ssa/predcom-8.c: Likewise.
	* gcc.dg/tree-ssa/cunroll-13.c: Rewrite to gimple testcase.
	* gcc.dg/tree-ssa/ivopt_mult_1.c: XFAIL.
	* gcc.dg/tree-ssa/ivopt_mult_1g.c: Add gimple variant that
	still passes.
	* gcc.dg/tree-ssa/ivopt_mult_2.c: XFAIL.
	* gcc.dg/tree-ssa/ivopt_mult_2g.c: Add gimple variant that
	still passes.
	* gcc.dg/tree-ssa/ssa-dom-thread-7.c: Adjust.
	* gcc.dg/tree-ssa/20030710-1.c: Likewise.
	* gcc.dg/tree-ssa/20030711-1.c: Likewise.

From-SVN: r259754

gcc/ChangeLog

+2018-04-30  Richard Biener  <rguenther@suse.de>
+
+	PR tree-optimization/28364
+	PR tree-optimization/85275
+	* tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
+	copying first exit test.
+
 2018-04-28  Mark Wielaard  <mark@klomp.org>
 
 	* dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for

gcc/testsuite/ChangeLog

+2018-04-30  Richard Biener  <rguenther@suse.de>
+
+	PR tree-optimization/28364
+	PR tree-optimization/85275
+	* gcc.dg/tree-ssa/copy-headers-5.c: New testcase.
+	* gcc.dg/tree-ssa/predcom-8.c: Likewise.
+	* gcc.dg/tree-ssa/cunroll-13.c: Rewrite to gimple testcase.
+	* gcc.dg/tree-ssa/ivopt_mult_1.c: XFAIL.
+	* gcc.dg/tree-ssa/ivopt_mult_1g.c: Add gimple variant that
+	still passes.
+	* gcc.dg/tree-ssa/ivopt_mult_2.c: XFAIL.
+	* gcc.dg/tree-ssa/ivopt_mult_2g.c: Add gimple variant that
+	still passes.
+	* gcc.dg/tree-ssa/ssa-dom-thread-7.c: Adjust.
+	* gcc.dg/tree-ssa/20030710-1.c: Likewise.
+	* gcc.dg/tree-ssa/20030711-1.c: Likewise.
+
 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
 
 	PR target/84431

gcc/testsuite/gcc.dg/tree-ssa/20030710-1.c

 /* { dg-do compile } */
-/* { dg-options "-O1 -fdump-tree-dom2" } */
+/* { dg-options "-O1 -fdump-tree-phicprop1" } */
 
 extern void abort (void);
 extern void blah (void);
@@ -42,14 +42,14 @@ record_component_aliases (type)
 /* The call to blah should have been eliminated.  If the call is not
    eliminated, then dominator optimizations failed and it'll be
    impossible to delete other unnecessary code.  */
-/* { dg-final { scan-tree-dump-not "blah \\(\\)" "dom2" } } */
+/* { dg-final { scan-tree-dump-not "blah \\(\\)" "phicprop1" } } */
   
 /* There should be two IF conditionals.  */
-/* { dg-final { scan-tree-dump-times "if " 2 "dom2"} } */
+/* { dg-final { scan-tree-dump-times "if " 2 "phicprop1"} } */
                                                                                 
 /* There should be a single load of type.binfo.  */
-/* { dg-final { scan-tree-dump-times "type\\.binfo" 1 "dom2"} } */
+/* { dg-final { scan-tree-dump-times "type\\.binfo" 1 "phicprop1"} } */
 
 /* There should be two loads of vec.length.  */
-/* { dg-final { scan-tree-dump-times "vec.length" 2 "dom2"} } */
+/* { dg-final { scan-tree-dump-times "vec.length" 2 "phicprop1"} } */
 

gcc/testsuite/gcc.dg/tree-ssa/20030711-1.c

@@ -44,12 +44,12 @@ record_component_aliases (type)
 /* The call to blah can not be eliminated.  */
 /* { dg-final { scan-tree-dump-times "blah \\(\\)" 1 "dom2" } } */
    
-/* There should be four IF conditionals.  */
-/* { dg-final { scan-tree-dump-times "if " 4 "dom2"} } */
+/* There should be three IF conditionals.  */
+/* { dg-final { scan-tree-dump-times "if " 3 "dom2"} } */
                                                                                 
 /* There should be two loads of type.binfo.  */
 /* { dg-final { scan-tree-dump-times "type\\.binfo" 2 "dom2"} } */
  
-/* There should be four loads of vec.length.  */
-/* { dg-final { scan-tree-dump-times "vec.length" 4 "dom2"} } */
+/* There should be three loads of vec.length.  */
+/* { dg-final { scan-tree-dump-times "vec.length" 3 "dom2"} } */
 

gcc/testsuite/gcc.dg/tree-ssa/copy-headers-5.c

+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-ch2-details" } */
+
+int is_sorted(int *a, int n)
+{
+  for (int i = 0; i < n - 1; i++)
+    if (a[i] > a[i + 1])
+      return 0;
+  return 1;
+}
+
+/* Verify we apply loop header copying but only copy the IV test and
+   not the alternate exit test.  */
+
+/* { dg-final { scan-tree-dump "is now do-while loop" "ch2" } } */
+/* { dg-final { scan-tree-dump-times "  if " 3 "ch2" } } */

gcc/testsuite/gcc.dg/tree-ssa/cunroll-13.c

 /* { dg-do compile } */
-/* { dg-options "-O3 -fdisable-tree-evrp -fdisable-tree-cunrolli -fdisable-tree-vrp1 -fdump-tree-cunroll-blocks-details" } */
+/* { dg-options "-O3 -fgimple -fdump-tree-cunroll-blocks-details" } */
 
 #if __SIZEOF_INT__ < 4
 __extension__ typedef __INT32_TYPE__ i32;
@@ -7,15 +7,53 @@ __extension__ typedef __INT32_TYPE__ i32;
 typedef int i32;
 #endif
 
-struct a {int a[8];int b;};
-void
-t(struct a *a)
+struct a {i32 a[8];i32 b;};
+
+void __GIMPLE (startwith("fix_loops"))
+t (struct a * a)
 {
-  for (i32 i=0;i<123456 && a->a[i];i++)
-    a->a[i]++;
+  i32 i;
+  i32 _1;
+  i32 _2;
+  i32 _9;
+  i32 _11;
+
+bb_2:
+  _11 = a_6(D)->a[0];
+  if (_11 != _Literal (i32) 0)
+    goto bb_6;
+  else
+    goto bb_3;
+
+bb_3:
+  return;
+
+bb_4:
+  _1 = _2 + 1;
+  a_6(D)->a[i_19] = _1;
+  i_8 = i_19 + _Literal (i32) 1;
+  if (i_8 <= _Literal (i32) 123455)
+    goto bb_5;
+  else
+    goto bb_3;
+
+bb_5:
+  i_19 = __PHI (bb_6: _Literal (i32) 1, bb_4: i_8);
+  _2 = a_6(D)->a[i_19];
+  if (_2 != _Literal (i32) 0)
+    goto bb_4;
+  else
+    goto bb_3;
+
+bb_6:
+  _9 = _11 + _Literal (i32) 1;
+  a_6(D)->a[0] = _9;
+  goto bb_5;
 }
-/* This pass relies on the fact that we do not eliminate the redundant test for i early.
-   It is necessary to disable all passes that do so.  At the moment it is evrp, vrp1 and cunrolli.  */
+
+/* This testcase relies on the fact that we do not eliminate the redundant test
+   for i early.  It is necessary to disable all passes that do so, for the
+   moment starting with the loop pipeline is good enough.  */
 /* { dg-final { scan-tree-dump-times "Loop 1 iterates 123454 times" 1 "cunroll" } } */
 /* { dg-final { scan-tree-dump-times "Last iteration exit edge was proved true" 1 "cunroll" } } */
 /* { dg-final { scan-tree-dump-times "Exit condition of peeled iterations was eliminated" 1 "cunroll" } } */

gcc/testsuite/gcc.dg/tree-ssa/ivopt_mult_1.c

@@ -20,4 +20,4 @@ long foo(long* p, long* p2, int N1, int N2)
   return s;
 }
 
-/* { dg-final { scan-tree-dump-times "Replacing" 1 "ivopts"} } */
+/* { dg-final { scan-tree-dump-times "Replacing" 1 "ivopts" { xfail *-*-* } } } */

gcc/testsuite/gcc.dg/tree-ssa/ivopt_mult_1g.c

+/* { dg-do compile { target {{ i?86-*-* x86_64-*-* } && lp64 } } } */
+/* { dg-options "-O2 -fgimple -m64 -fdump-tree-ivopts-details" } */
+
+/* The test 'if (p2 > p_limit2)' can be replaced, so iv p2 can be
+ * eliminated.  */
+long int __GIMPLE (startwith("fix_loops"))
+foo (long int * p, long int * p2, int N1, int N2)
+{
+  long int s;
+  long int * p_limit2;
+  long int * p_limit;
+  long unsigned int _1;
+  long unsigned int _2;
+  long unsigned int _3;
+  long unsigned int _4;
+  long int _5;
+
+  bb_2:
+  _1 = (long unsigned int) N1_10(D);
+  _2 = _1 * 8ul;
+  p_limit_12 = p_11(D) + _2;
+  _3 = (long unsigned int) N2_13(D);
+  _4 = _3 * 8ul;
+  p_limit2_15 = p2_14(D) + _4;
+  if (p_11(D) <= p_limit_12)
+    goto bb_3;
+  else
+    goto bb_13;
+
+  bb_13:
+
+  bb_9:
+  goto bb_6;
+
+  bb_3:
+  p_20 = p_11(D) + 8ul;
+  p2_23 = p2_14(D) + 8ul;
+  if (p_limit2_15 < p2_23)
+    goto bb_14;
+  else
+    goto bb_7;
+
+  bb_14:
+  goto bb_9;
+
+  bb_7:
+  goto bb_5;
+
+  bb_4:
+  p_16 = p_26 + 8ul;
+  p2_17 = p2_27 + 8ul;
+  if (p_limit2_15 < p2_17)
+    goto bb_11;
+  else
+    goto bb_8;
+
+  bb_11:
+  goto bb_6;
+
+  bb_8:
+  ;
+
+  bb_5:
+  s_24 = __PHI (bb_7: 0l, bb_8: s_19);
+  p_26 = __PHI (bb_7: p_20, bb_8: p_16);
+  p2_27 = __PHI (bb_7: p2_23, bb_8: p2_17);
+  _5 = __MEM <long int> (p_26);
+  s_19 = _5 + s_24;
+  if (p_limit_12 >= p_26)
+    goto bb_4;
+  else
+    goto bb_12;
+
+  bb_12:
+  ;
+
+  bb_6:
+  s_25 = __PHI (bb_12: s_19, bb_11: s_19, bb_9: 0l);
+  return s_25;
+}
+
+/* { dg-final { scan-tree-dump-times "Replacing" 1 "ivopts" } } */

gcc/testsuite/gcc.dg/tree-ssa/ivopt_mult_2.c

@@ -21,4 +21,4 @@ long foo(long* p, long* p2, int N1, int N2)
   return s;
 }
 
-/* { dg-final { scan-tree-dump-times "Replacing" 2 "ivopts"} } */
+/* { dg-final { scan-tree-dump-times "Replacing" 2 "ivopts" { xfail *-*-* } } } */

gcc/testsuite/gcc.dg/tree-ssa/ivopt_mult_2g.c

+/* { dg-do compile { target {{ i?86-*-* x86_64-*-* } && lp64 } } } */
+/* { dg-options "-O2 -fgimple -m64 -fdump-tree-ivopts-details" } */
+
+/* Exit tests 'i < N1' and 'p2 > p_limit2' can be replaced, so
+ * two ivs i and p2 can be eliminate.  */
+long int __GIMPLE (startwith("fix_loops"))
+foo (long int * p, long int * p2, int N1, int N2)
+{
+  long int s;
+  long int * p_limit2;
+  int i;
+  long unsigned int _1;
+  long unsigned int _2;
+  long int _3;
+
+  bb_2:
+  _1 = (long unsigned int) N2_9(D);
+  _2 = _1 * 8ul;
+  p_limit2_11 = p2_10(D) + _2;
+  if (N1_13(D) > 0)
+    goto bb_3;
+  else
+    goto bb_13;
+
+  bb_13:
+
+  bb_9:
+  goto bb_6;
+
+  bb_3:
+  p_22 = p_12(D) + 8ul;
+  p2_23 = p2_10(D) + 8ul;
+  if (p_limit2_11 < p2_23)
+    goto bb_14;
+  else
+    goto bb_7;
+
+  bb_14:
+  goto bb_9;
+
+  bb_7:
+  goto bb_5;
+
+  bb_4:
+  p_14 = p_27 + 8ul;
+  p2_15 = p2_28 + 8ul;
+  i_16 = i_29 + 1;
+  if (p_limit2_11 < p2_15)
+    goto bb_11;
+  else
+    goto bb_8;
+
+  bb_11:
+  goto bb_6;
+
+  bb_8:
+  ;
+
+  bb_5:
+  s_25 = __PHI (bb_7: 0l, bb_8: s_18);
+  p_27 = __PHI (bb_7: p_22, bb_8: p_14);
+  p2_28 = __PHI (bb_7: p2_23, bb_8: p2_15);
+  i_29 = __PHI (bb_7: 1, bb_8: i_16);
+  _3 = __MEM <long int> (p_27);
+  s_18 = _3 + s_25;
+  if (N1_13(D) > i_29)
+    goto bb_4;
+  else
+    goto bb_12;
+
+  bb_12:
+  ;
+
+  bb_6:
+  s_26 = __PHI (bb_12: s_18, bb_11: s_18, bb_9: 0l);
+  return s_26;
+}
+
+/* { dg-final { scan-tree-dump-times "Replacing" 2 "ivopts"} } */

gcc/testsuite/gcc.dg/tree-ssa/predcom-8.c

+/* { dg-do compile } */
+/* { dg-options "-O3 -fdump-tree-pcom-details" } */
+
+int is_sorted(int *a, int n)
+{
+  for (int i = 0; i < n - 1; i++)
+    if (a[i] > a[i + 1])
+      return 0;
+  return 1;
+}
+
+/* { dg-final { scan-tree-dump "Executing predictive commoning without unrolling" "pcom" } } */

gcc/testsuite/gcc.dg/tree-ssa/ssa-dom-thread-7.c

@@ -3,7 +3,7 @@
 /* { dg-final { scan-tree-dump "Jumps threaded: 16"  "thread1" } } */
 /* { dg-final { scan-tree-dump "Jumps threaded: 9" "thread2" } } */
 /* { dg-final { scan-tree-dump "Jumps threaded: 3" "thread3" } } */
-/* { dg-final { scan-tree-dump-not "Jumps threaded"  "dom2" } } */
+/* { dg-final { scan-tree-dump "Jumps threaded: 1"  "dom2" } } */
 /* { dg-final { scan-tree-dump-not "Jumps threaded"  "dom3" } } */
 /* { dg-final { scan-tree-dump-not "Jumps threaded"  "vrp2" } } */
 

gcc/tree-ssa-loop-ch.c

@@ -340,6 +340,11 @@ ch_base::copy_headers (function *fun)
 	  bbs[n_bbs++] = header;
 	  gcc_assert (bbs_size > n_bbs);
 	  header = exit->dest;
+	  /* Make sure to stop copying after we copied the first exit test.
+	     Without further heuristics we do not want to rotate the loop
+	     any further.  */
+	  if (loop_exits_from_bb_p (loop, exit->src))
+	    break;
 	}
 
       if (!exit)