[PATCH GCC]Make pointer overflow always undefined and remove the macro

[PATCH GCC]Make pointer overflow always undefined and remove the macro

[Bin Cheng]

[-- Attachment #1: Type: text/plain, Size: 876 bytes --]

Hi,
This is a followup patch to PR81388's fix.  According to Richi,
POINTER_TYPE_OVERFLOW_UNDEFINED was added in -fstrict-overflow
warning work.  Given:
  A) strict-overflow was removed;
  B) memory object can not wrap in address space;
  C) existing code doesn't take it in consideration, as in nowrap_type_p.
This patch makes it always true thus removes definition/usage of the macro.
Bootstrap and test on x86_64 and AArch64.  Is it OK?

Thanks,
bin
2017-07-20  Bin Cheng  <bin.cheng@arm.com>

	* tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
	* fold-const.c (fold_comparison, fold_binary_loc): Delete use of
	above macro.
	* match.pd: Ditto in address comparison pattern.

gcc/testsuite/ChangeLog
2017-07-20  Bin Cheng  <bin.cheng@arm.com>

	* gcc.dg/no-strict-overflow-7.c: Revise comment and test string.
	* gcc.dg/tree-ssa/pr81388-1.c: Ditto.

[-- Attachment #2: undefined-overflow-pointer-20170721.txt --]
[-- Type: text/plain, Size: 5118 bytes --]

diff --git a/gcc/fold-const.c b/gcc/fold-const.c
index 1bcbbb5..78bb326 100644
--- a/gcc/fold-const.c
+++ b/gcc/fold-const.c
@@ -8505,14 +8505,9 @@ fold_comparison (location_t loc, enum tree_code code, tree type,
 	{
 	  /* We can fold this expression to a constant if the non-constant
 	     offset parts are equal.  */
-	  if ((offset0 == offset1
-	       || (offset0 && offset1
-		   && operand_equal_p (offset0, offset1, 0)))
-	      && (equality_code
-		  || (indirect_base0
-		      && (DECL_P (base0) || CONSTANT_CLASS_P (base0)))
-		  || POINTER_TYPE_OVERFLOW_UNDEFINED))
-
+	  if (offset0 == offset1
+	      || (offset0 && offset1
+		  && operand_equal_p (offset0, offset1, 0)))
 	    {
 	      if (!equality_code
 		  && bitpos0 != bitpos1
@@ -8547,11 +8542,7 @@ fold_comparison (location_t loc, enum tree_code code, tree type,
 	     because pointer arithmetic is restricted to retain within an
 	     object and overflow on pointer differences is undefined as of
 	     6.5.6/8 and /9 with respect to the signed ptrdiff_t.  */
-	  else if (bitpos0 == bitpos1
-		   && (equality_code
-		       || (indirect_base0
-			   && (DECL_P (base0) || CONSTANT_CLASS_P (base0)))
-		       || POINTER_TYPE_OVERFLOW_UNDEFINED))
+	  else if (bitpos0 == bitpos1)
 	    {
 	      /* By converting to signed sizetype we cover middle-end pointer
 	         arithmetic which operates on unsigned pointer types of size
@@ -9651,7 +9642,7 @@ fold_binary_loc (location_t loc,
 
 	  /* With undefined overflow prefer doing association in a type
 	     which wraps on overflow, if that is one of the operand types.  */
-	  if ((POINTER_TYPE_P (type) && POINTER_TYPE_OVERFLOW_UNDEFINED)
+	  if (POINTER_TYPE_P (type)
 	      || (INTEGRAL_TYPE_P (type) && !TYPE_OVERFLOW_WRAPS (type)))
 	    {
 	      if (INTEGRAL_TYPE_P (TREE_TYPE (arg0))
@@ -9665,7 +9656,7 @@ fold_binary_loc (location_t loc,
 
 	  /* With undefined overflow we can only associate constants with one
 	     variable, and constants whose association doesn't overflow.  */
-	  if ((POINTER_TYPE_P (atype) && POINTER_TYPE_OVERFLOW_UNDEFINED)
+	  if (POINTER_TYPE_P (atype)
 	      || (INTEGRAL_TYPE_P (atype) && !TYPE_OVERFLOW_WRAPS (atype)))
 	    {
 	      if (var0 && var1)
diff --git a/gcc/match.pd b/gcc/match.pd
index 979085a..b89aed3 100644
--- a/gcc/match.pd
+++ b/gcc/match.pd
@@ -3129,14 +3129,7 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
 		    || TREE_CODE (base1) == STRING_CST))
          equal = (base0 == base1);
      }
-     (if (equal == 1
-	  && (cmp == EQ_EXPR || cmp == NE_EXPR
-	      /* If the offsets are equal we can ignore overflow.  */
-	      || off0 == off1
-	      || POINTER_TYPE_OVERFLOW_UNDEFINED
-	      /* Or if we compare using pointers to decls or strings.  */
-	      || (POINTER_TYPE_P (TREE_TYPE (@2))
-		  && (DECL_P (base0) || TREE_CODE (base0) == STRING_CST))))
+     (if (equal == 1)
       (switch
        (if (cmp == EQ_EXPR)
 	{ constant_boolean_node (off0 == off1, type); })
diff --git a/gcc/testsuite/gcc.dg/no-strict-overflow-7.c b/gcc/testsuite/gcc.dg/no-strict-overflow-7.c
index 19e1b55..0e73d48 100644
--- a/gcc/testsuite/gcc.dg/no-strict-overflow-7.c
+++ b/gcc/testsuite/gcc.dg/no-strict-overflow-7.c
@@ -3,8 +3,8 @@
 
 /* Source: Ian Lance Taylor.  Dual of strict-overflow-6.c.  */
 
-/* We can only simplify the conditional when using strict overflow
-   semantics.  */
+/* We can simplify the conditional because pointer overflow always has
+   undefined semantics.  */
 
 int
 foo (char* p)
@@ -12,4 +12,4 @@ foo (char* p)
   return p + 1000 < p;
 }
 
-/* { dg-final { scan-tree-dump "\[+\]\[ \]*1000" "optimized" } } */
+/* { dg-final { scan-tree-dump "return 0" "optimized" } } */
diff --git a/gcc/testsuite/gcc.dg/tree-ssa/pr81388-1.c b/gcc/testsuite/gcc.dg/tree-ssa/pr81388-1.c
index ecfe129..0beb510 100644
--- a/gcc/testsuite/gcc.dg/tree-ssa/pr81388-1.c
+++ b/gcc/testsuite/gcc.dg/tree-ssa/pr81388-1.c
@@ -1,5 +1,5 @@
 /* { dg-do compile } */
-/* { dg-options "-O2 -fno-strict-overflow -fdump-tree-ivcanon-details" } */
+/* { dg-options "-O2 -fno-strict-overflow -fdump-tree-tailc-details" } */
 
 void bar();
 void foo(char *dst)
@@ -11,4 +11,6 @@ void foo(char *dst)
   } while (dst < end);
 }
 
-/* { dg-final { scan-tree-dump-times " zero if " 1 "ivcanon" } } */
+/* The loop only iterates once because pointer overflow always has undefined
+   semantics.  As a result, call to bar becomes tail call.  */
+/* { dg-final { scan-tree-dump-times "Found tail call " 1 "tailc" } } */
diff --git a/gcc/tree.h b/gcc/tree.h
index 91cf253..4e4edf2 100644
--- a/gcc/tree.h
+++ b/gcc/tree.h
@@ -859,9 +859,6 @@ extern void omp_clause_range_check_failed (const_tree, const char *, int,
    && !TYPE_OVERFLOW_WRAPS (TYPE)			\
    && (flag_sanitize & SANITIZE_SI_OVERFLOW))
 
-/* True if pointer types have undefined overflow.  */
-#define POINTER_TYPE_OVERFLOW_UNDEFINED (!flag_wrapv)
-
 /* Nonzero in a VAR_DECL or STRING_CST means assembler code has been written.
    Nonzero in a FUNCTION_DECL means that the function has been compiled.
    This is interesting in an inline function, since it might not need

Re: [PATCH GCC]Make pointer overflow always undefined and remove the macro

[Richard Biener]

On Mon, Jul 24, 2017 at 10:43 AM, Bin Cheng <Bin.Cheng@arm.com> wrote:
> Hi,
> This is a followup patch to PR81388's fix.  According to Richi,
> POINTER_TYPE_OVERFLOW_UNDEFINED was added in -fstrict-overflow
> warning work.  Given:
>   A) strict-overflow was removed;
>   B) memory object can not wrap in address space;
>   C) existing code doesn't take it in consideration, as in nowrap_type_p.
> This patch makes it always true thus removes definition/usage of the macro.
> Bootstrap and test on x86_64 and AArch64.  Is it OK?

Ok.

Please give others 24h to comment.

Thanks,
Richard.

> Thanks,
> bin
> 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
>
>         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
>         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
>         above macro.
>         * match.pd: Ditto in address comparison pattern.
>
> gcc/testsuite/ChangeLog
> 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
>
>         * gcc.dg/no-strict-overflow-7.c: Revise comment and test string.
>         * gcc.dg/tree-ssa/pr81388-1.c: Ditto.

Re: [PATCH GCC]Make pointer overflow always undefined and remove the macro

[Bin.Cheng]

On Tue, Jul 25, 2017 at 8:26 AM, Richard Biener
<richard.guenther@gmail.com> wrote:
> On Mon, Jul 24, 2017 at 10:43 AM, Bin Cheng <Bin.Cheng@arm.com> wrote:
>> Hi,
>> This is a followup patch to PR81388's fix.  According to Richi,
>> POINTER_TYPE_OVERFLOW_UNDEFINED was added in -fstrict-overflow
>> warning work.  Given:
>>   A) strict-overflow was removed;
>>   B) memory object can not wrap in address space;
>>   C) existing code doesn't take it in consideration, as in nowrap_type_p.
>> This patch makes it always true thus removes definition/usage of the macro.
>> Bootstrap and test on x86_64 and AArch64.  Is it OK?
>
> Ok.
>
> Please give others 24h to comment.
Hi,
I committed the patch as:

r250765 | amker | 2017-08-01 10:28:18 +0100 (Tue, 01 Aug 2017) | 9 lines

        * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
        * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
        above macro.
        * match.pd: Ditto in address comparison pattern.

        gcc/testsuite
        * gcc.dg/no-strict-overflow-7.c: Revise comment and test string.
        * gcc.dg/tree-ssa/pr81388-1.c: Ditto.

We can always revert it if there is any different opinions.

Thanks,
bin