[RFC][IPA-VRP] ADDR_EXPR and nonnull

[RFC][IPA-VRP] ADDR_EXPR and nonnull

[kugan]

[-- Attachment #1: Type: text/plain, Size: 767 bytes --]

Hi,

While computing jump function value range for pointer, I am wondering if 
we can assume that any tree with ADDR_EXPR will be nonnull.

That is, in cases like:

int arr[10];
foo (&arr[1]);

OR

struct st
{
   int a;
   int b;
};
struct st s2;
foo (&s2.a);

Attached patch tries to do this. I am not sure if this can be wrong. Any 
thoughts?

Attached patch bootstraps and regression testing didn't introduce any 
new regressions.

Thanks,
Kugan


gcc/ChangeLog:

2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>

	* ipa-prop.c (ipa_compute_jump_functions_for_edge): Set
	value range to nonull for ADDR_EXPR.

gcc/testsuite/ChangeLog:

2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>

	* gcc.dg/ipa/vrp5.c: New test.
	* gcc.dg/ipa/vrp6.c: New test.

[-- Attachment #2: p.txt --]
[-- Type: text/plain, Size: 1919 bytes --]

diff --git a/gcc/ipa-prop.c b/gcc/ipa-prop.c
index 353b638..b795d30 100644
--- a/gcc/ipa-prop.c
+++ b/gcc/ipa-prop.c
@@ -1670,9 +1670,10 @@ ipa_compute_jump_functions_for_edge (struct ipa_func_body_info *fbi,
 
       if (POINTER_TYPE_P (TREE_TYPE (arg)))
 	{
-	  if (TREE_CODE (arg) == SSA_NAME
-	      && param_type
-	      && get_ptr_nonnull (arg))
+	  if ((TREE_CODE (arg) == SSA_NAME
+	       && param_type
+	       && get_ptr_nonnull (arg))
+	      || (TREE_CODE (arg) == ADDR_EXPR))
 	    {
 	      jfunc->vr_known = true;
 	      jfunc->m_vr.type = VR_ANTI_RANGE;
diff --git a/gcc/testsuite/gcc.dg/ipa/vrp5.c b/gcc/testsuite/gcc.dg/ipa/vrp5.c
index e69de29..8e43a65 100644
--- a/gcc/testsuite/gcc.dg/ipa/vrp5.c
+++ b/gcc/testsuite/gcc.dg/ipa/vrp5.c
@@ -0,0 +1,28 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-ipa-cp-details -fdump-tree-vrp1" } */
+
+static __attribute__((noinline, noclone))
+int foo (int *p)
+{
+  if (!p)
+    return 0;
+  *p = 1;
+}
+
+struct st
+{
+  int a;
+  int b;
+};
+
+int bar (struct st *s)
+{
+int arr[10];
+  if (!s)
+    return 0;
+  foo (&s->a);
+  foo (&arr[1]);
+}
+
+/* { dg-final { scan-ipa-dump "Setting nonnull for 0" "cp" } } */
+/* { dg-final { scan-tree-dump-times "if" 1 "vrp1" } } */
diff --git a/gcc/testsuite/gcc.dg/ipa/vrp6.c b/gcc/testsuite/gcc.dg/ipa/vrp6.c
index e69de29..f837758 100644
--- a/gcc/testsuite/gcc.dg/ipa/vrp6.c
+++ b/gcc/testsuite/gcc.dg/ipa/vrp6.c
@@ -0,0 +1,28 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-ipa-cp-details -fdump-tree-vrp1" } */
+
+static __attribute__((noinline, noclone))
+int foo (int *p)
+{
+  if (!p)
+    return 0;
+  *p = 1;
+}
+
+struct st
+{
+  int a;
+  int b;
+};
+
+int bar (struct st *s)
+{
+struct st s2;
+  if (!s)
+    return 0;
+  foo (&s->a);
+  foo (&s2.a);
+}
+
+/* { dg-final { scan-ipa-dump "Setting nonnull for 0" "cp" } } */
+/* { dg-final { scan-tree-dump-times "if" 1 "vrp1" } } */

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[Richard Biener]

On Wed, 19 Oct 2016, kugan wrote:

> Hi,
> 
> While computing jump function value range for pointer, I am wondering if we
> can assume that any tree with ADDR_EXPR will be nonnull.
> 
> That is, in cases like:
> 
> int arr[10];
> foo (&arr[1]);
> 
> OR
> 
> struct st
> {
>   int a;
>   int b;
> };
> struct st s2;
> foo (&s2.a);
> 
> Attached patch tries to do this. I am not sure if this can be wrong. Any
> thoughts?

It can be wrong for weak symbols for example.

Richard.

> Attached patch bootstraps and regression testing didn't introduce any new
> regressions.
> 
> Thanks,
> Kugan
> 
> 
> gcc/ChangeLog:
> 
> 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
> 
> 	* ipa-prop.c (ipa_compute_jump_functions_for_edge): Set
> 	value range to nonull for ADDR_EXPR.
> 
> gcc/testsuite/ChangeLog:
> 
> 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
> 
> 	* gcc.dg/ipa/vrp5.c: New test.
> 	* gcc.dg/ipa/vrp6.c: New test.
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nuernberg)

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[kugan]

Hi Richard,


On 19/10/16 19:23, Richard Biener wrote:
> On Wed, 19 Oct 2016, kugan wrote:
>
>> Hi,
>>
>> While computing jump function value range for pointer, I am wondering if we
>> can assume that any tree with ADDR_EXPR will be nonnull.
>>
>> That is, in cases like:
>>
>> int arr[10];
>> foo (&arr[1]);
>>
>> OR
>>
>> struct st
>> {
>>   int a;
>>   int b;
>> };
>> struct st s2;
>> foo (&s2.a);
>>
>> Attached patch tries to do this. I am not sure if this can be wrong. Any
>> thoughts?
>
> It can be wrong for weak symbols for example.
Would excluding weak symbols (I believe I can check DECL_WEAK for this) 
good enough. Or looking for acceptable subset would work?

Thanks,
Kugan

>
> Richard.
>
>> Attached patch bootstraps and regression testing didn't introduce any new
>> regressions.
>>
>> Thanks,
>> Kugan
>>
>>
>> gcc/ChangeLog:
>>
>> 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
>>
>> 	* ipa-prop.c (ipa_compute_jump_functions_for_edge): Set
>> 	value range to nonull for ADDR_EXPR.
>>
>> gcc/testsuite/ChangeLog:
>>
>> 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
>>
>> 	* gcc.dg/ipa/vrp5.c: New test.
>> 	* gcc.dg/ipa/vrp6.c: New test.
>>
>

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[Richard Biener]

On Wed, 19 Oct 2016, kugan wrote:

> Hi Richard,
> 
> 
> On 19/10/16 19:23, Richard Biener wrote:
> > On Wed, 19 Oct 2016, kugan wrote:
> > 
> > > Hi,
> > > 
> > > While computing jump function value range for pointer, I am wondering if
> > > we
> > > can assume that any tree with ADDR_EXPR will be nonnull.
> > > 
> > > That is, in cases like:
> > > 
> > > int arr[10];
> > > foo (&arr[1]);
> > > 
> > > OR
> > > 
> > > struct st
> > > {
> > >   int a;
> > >   int b;
> > > };
> > > struct st s2;
> > > foo (&s2.a);
> > > 
> > > Attached patch tries to do this. I am not sure if this can be wrong. Any
> > > thoughts?
> > 
> > It can be wrong for weak symbols for example.
> Would excluding weak symbols (I believe I can check DECL_WEAK for this) good
> enough. Or looking for acceptable subset would work?

I think we should add a symtab helper to tell if address_nonzero_p (if
that doesn't aleady exist).

Richard.

> Thanks,
> Kugan
> 
> > 
> > Richard.
> > 
> > > Attached patch bootstraps and regression testing didn't introduce any new
> > > regressions.
> > > 
> > > Thanks,
> > > Kugan
> > > 
> > > 
> > > gcc/ChangeLog:
> > > 
> > > 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
> > > 
> > > 	* ipa-prop.c (ipa_compute_jump_functions_for_edge): Set
> > > 	value range to nonull for ADDR_EXPR.
> > > 
> > > gcc/testsuite/ChangeLog:
> > > 
> > > 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
> > > 
> > > 	* gcc.dg/ipa/vrp5.c: New test.
> > > 	* gcc.dg/ipa/vrp6.c: New test.
> > > 
> > 
> 
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nuernberg)

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[Jan Hubicka]

> > Would excluding weak symbols (I believe I can check DECL_WEAK for this) good
> > enough. Or looking for acceptable subset would work?
> 
> I think we should add a symtab helper to tell if address_nonzero_p (if
> that doesn't aleady exist).

We have node->nonzero_address()

Honza
> 
> Richard.
> 
> > Thanks,
> > Kugan
> > 
> > > 
> > > Richard.
> > > 
> > > > Attached patch bootstraps and regression testing didn't introduce any new
> > > > regressions.
> > > > 
> > > > Thanks,
> > > > Kugan
> > > > 
> > > > 
> > > > gcc/ChangeLog:
> > > > 
> > > > 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
> > > > 
> > > > 	* ipa-prop.c (ipa_compute_jump_functions_for_edge): Set
> > > > 	value range to nonull for ADDR_EXPR.
> > > > 
> > > > gcc/testsuite/ChangeLog:
> > > > 
> > > > 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
> > > > 
> > > > 	* gcc.dg/ipa/vrp5.c: New test.
> > > > 	* gcc.dg/ipa/vrp6.c: New test.
> > > > 
> > > 
> > 
> > 
> 
> -- 
> Richard Biener <rguenther@suse.de>
> SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nuernberg)

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[kugan]

[-- Attachment #1: Type: text/plain, Size: 1836 bytes --]



On 20/10/16 01:26, Jan Hubicka wrote:
>>> Would excluding weak symbols (I believe I can check DECL_WEAK for this) good
>>> enough. Or looking for acceptable subset would work?
>>
>> I think we should add a symtab helper to tell if address_nonzero_p (if
>> that doesn't aleady exist).
>
> We have node->nonzero_address()

Thanks for the pointer. Here is an attempt. Does this look OK?

+	  if (TREE_CODE (arg) == ADDR_EXPR)
+	    {
+	      /* See if the AADR_EXPR is nonnull.  */
+	      varpool_node *node = NULL;
+	      tree base = TREE_OPERAND (arg, 0);
+	      base = get_base_address (base);
+
+	      if (decl_address_ip_invariant_p (base)
+		  || !is_global_var (base))
+		{
+		  /* If the symbol address is local or
+		     constant.  */
+		  addr_nonzero = true;
+		}
+	      else
+		{
+		  /* If symbol address is nonzero_address.  */
+		  node = varpool_node::get (base);
+		  if (node && node->nonzero_address ())
+		    addr_nonzero = true;
+		}
+	    }

Attached patch passes bootstrap and regression testing on x86_64-linu-gnu.

Thanks,
Kugan

>
> Honza
>>
>> Richard.
>>
>>> Thanks,
>>> Kugan
>>>
>>>>
>>>> Richard.
>>>>
>>>>> Attached patch bootstraps and regression testing didn't introduce any new
>>>>> regressions.
>>>>>
>>>>> Thanks,
>>>>> Kugan
>>>>>
>>>>>
>>>>> gcc/ChangeLog:
>>>>>
>>>>> 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
>>>>>
>>>>> 	* ipa-prop.c (ipa_compute_jump_functions_for_edge): Set
>>>>> 	value range to nonull for ADDR_EXPR.
>>>>>
>>>>> gcc/testsuite/ChangeLog:
>>>>>
>>>>> 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
>>>>>
>>>>> 	* gcc.dg/ipa/vrp5.c: New test.
>>>>> 	* gcc.dg/ipa/vrp6.c: New test.
>>>>>
>>>>
>>>
>>>
>>
>> --
>> Richard Biener <rguenther@suse.de>
>> SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nuernberg)

[-- Attachment #2: p.txt --]
[-- Type: text/plain, Size: 2677 bytes --]

diff --git a/gcc/ipa-prop.c b/gcc/ipa-prop.c
index 353b638..affd8b9 100644
--- a/gcc/ipa-prop.c
+++ b/gcc/ipa-prop.c
@@ -1670,9 +1670,35 @@ ipa_compute_jump_functions_for_edge (struct ipa_func_body_info *fbi,
 
       if (POINTER_TYPE_P (TREE_TYPE (arg)))
 	{
-	  if (TREE_CODE (arg) == SSA_NAME
-	      && param_type
-	      && get_ptr_nonnull (arg))
+	  bool addr_nonzero = false;
+
+	  if (TREE_CODE (arg) == ADDR_EXPR)
+	    {
+	      /* See if the AADR_EXPR is nonnull.  */
+	      varpool_node *node = NULL;
+	      tree base = TREE_OPERAND (arg, 0);
+	      base = get_base_address (base);
+
+	      if (decl_address_ip_invariant_p (base)
+		  || !is_global_var (base))
+		{
+		  /* If the symbol address is local or
+		     constant.  */
+		  addr_nonzero = true;
+		}
+	      else
+		{
+		  /* If symbol address is nonzero_address.  */
+		  node = varpool_node::get (base);
+		  if (node && node->nonzero_address ())
+		    addr_nonzero = true;
+		}
+	    }
+
+	  if ((TREE_CODE (arg) == SSA_NAME
+	       && param_type
+	       && get_ptr_nonnull (arg))
+	      || addr_nonzero)
 	    {
 	      jfunc->vr_known = true;
 	      jfunc->m_vr.type = VR_ANTI_RANGE;
diff --git a/gcc/testsuite/gcc.dg/ipa/vrp5.c b/gcc/testsuite/gcc.dg/ipa/vrp5.c
index e69de29..571798d 100644
--- a/gcc/testsuite/gcc.dg/ipa/vrp5.c
+++ b/gcc/testsuite/gcc.dg/ipa/vrp5.c
@@ -0,0 +1,34 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-ipa-cp-details -fdump-tree-vrp1" } */
+
+static __attribute__((noinline, noclone))
+int foo (int *p)
+{
+  if (!p)
+    return 0;
+  *p = 1;
+}
+
+struct st
+{
+  int a;
+  int b;
+};
+
+int arr1[10];
+int a;
+int bar (struct st *s)
+{
+  int arr2[10];
+  int b;
+  if (!s)
+    return 0;
+  foo (&s->a);
+  foo (&a);
+  foo (&b);
+  foo (&arr1[1]);
+  foo (&arr2[1]);
+}
+
+/* { dg-final { scan-ipa-dump "Setting nonnull for 0" "cp" } } */
+/* { dg-final { scan-tree-dump-times "if" 1 "vrp1" } } */
diff --git a/gcc/testsuite/gcc.dg/ipa/vrp6.c b/gcc/testsuite/gcc.dg/ipa/vrp6.c
index e69de29..971db44 100644
--- a/gcc/testsuite/gcc.dg/ipa/vrp6.c
+++ b/gcc/testsuite/gcc.dg/ipa/vrp6.c
@@ -0,0 +1,34 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-ipa-cp-details -fdump-tree-vrp1" } */
+
+static __attribute__((noinline, noclone))
+int foo (int *p)
+{
+  if (!p)
+    return 0;
+  *p = 1;
+}
+
+struct st
+{
+  int a;
+  int b;
+};
+
+struct st s2;
+int a;
+int bar (struct st *s)
+{
+  struct st s3;
+  int b;
+  if (!s)
+    return 0;
+  foo (&s->a);
+  foo (&s2.a);
+  foo (&s3.a);
+  foo (&a);
+  foo (&b);
+}
+
+/* { dg-final { scan-ipa-dump "Setting nonnull for 0" "cp" } } */
+/* { dg-final { scan-tree-dump-times "if" 1 "vrp1" } } */

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[Richard Biener]

On Thu, 20 Oct 2016, kugan wrote:

> 
> 
> On 20/10/16 01:26, Jan Hubicka wrote:
> > > > Would excluding weak symbols (I believe I can check DECL_WEAK for this)
> > > > good
> > > > enough. Or looking for acceptable subset would work?
> > > 
> > > I think we should add a symtab helper to tell if address_nonzero_p (if
> > > that doesn't aleady exist).
> > 
> > We have node->nonzero_address()
> 
> Thanks for the pointer. Here is an attempt. Does this look OK?
> 
> +	  if (TREE_CODE (arg) == ADDR_EXPR)
> +	    {
> +	      /* See if the AADR_EXPR is nonnull.  */
> +	      varpool_node *node = NULL;
> +	      tree base = TREE_OPERAND (arg, 0);
> +	      base = get_base_address (base);
> +
> +	      if (decl_address_ip_invariant_p (base)
> +		  || !is_global_var (base))
> +		{
> +		  /* If the symbol address is local or
> +		     constant.  */

"constant" doesn't matter.  You want

   if (TREE_CODE (base) == CONST_DECL
|| TREE_CODE (base) == PARM_DECL
|| TREE_CODE (base) == RESULT_DECL)
  addr_nonzero = true;
   else if (VAR_P (base))
     addr_nonzero = ! is_global_var (base) || (varpool_node::get 
(base)->nonzero_address ());

Richard.

> +		  addr_nonzero = true;
> +		}
> +	      else
> +		{
> +		  /* If symbol address is nonzero_address.  */
> +		  node = varpool_node::get (base);
> +		  if (node && node->nonzero_address ())
> +		    addr_nonzero = true;
> +		}
> +	    }
> 
> Attached patch passes bootstrap and regression testing on x86_64-linu-gnu.
> 
> Thanks,
> Kugan
> 
> > 
> > Honza
> > > 
> > > Richard.
> > > 
> > > > Thanks,
> > > > Kugan
> > > > 
> > > > > 
> > > > > Richard.
> > > > > 
> > > > > > Attached patch bootstraps and regression testing didn't introduce
> > > > > > any new
> > > > > > regressions.
> > > > > > 
> > > > > > Thanks,
> > > > > > Kugan
> > > > > > 
> > > > > > 
> > > > > > gcc/ChangeLog:
> > > > > > 
> > > > > > 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
> > > > > > 
> > > > > > 	* ipa-prop.c (ipa_compute_jump_functions_for_edge): Set
> > > > > > 	value range to nonull for ADDR_EXPR.
> > > > > > 
> > > > > > gcc/testsuite/ChangeLog:
> > > > > > 
> > > > > > 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
> > > > > > 
> > > > > > 	* gcc.dg/ipa/vrp5.c: New test.
> > > > > > 	* gcc.dg/ipa/vrp6.c: New test.
> > > > > > 
> > > > > 
> > > > 
> > > > 
> > > 
> > > --
> > > Richard Biener <rguenther@suse.de>
> > > SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB
> > > 21284 (AG Nuernberg)
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nuernberg)

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[kugan]

[-- Attachment #1: Type: text/plain, Size: 2574 bytes --]

Hi Richard,


On 20/10/16 18:41, Richard Biener wrote:
> On Thu, 20 Oct 2016, kugan wrote:
>
>>
>>
>> On 20/10/16 01:26, Jan Hubicka wrote:
>>>>> Would excluding weak symbols (I believe I can check DECL_WEAK for this)
>>>>> good
>>>>> enough. Or looking for acceptable subset would work?
>>>>
>>>> I think we should add a symtab helper to tell if address_nonzero_p (if
>>>> that doesn't aleady exist).
>>>
>>> We have node->nonzero_address()
>>
>> Thanks for the pointer. Here is an attempt. Does this look OK?
>>
>> +	  if (TREE_CODE (arg) == ADDR_EXPR)
>> +	    {
>> +	      /* See if the AADR_EXPR is nonnull.  */
>> +	      varpool_node *node = NULL;
>> +	      tree base = TREE_OPERAND (arg, 0);
>> +	      base = get_base_address (base);
>> +
>> +	      if (decl_address_ip_invariant_p (base)
>> +		  || !is_global_var (base))
>> +		{
>> +		  /* If the symbol address is local or
>> +		     constant.  */
>
> "constant" doesn't matter.  You want
>
>    if (TREE_CODE (base) == CONST_DECL
> || TREE_CODE (base) == PARM_DECL
> || TREE_CODE (base) == RESULT_DECL)
>   addr_nonzero = true;
>    else if (VAR_P (base))
>      addr_nonzero = ! is_global_var (base) || (varpool_node::get
> (base)->nonzero_address ());

Attached patch does this. Bootstrapped and regression tested on 
x86_64-linux-gnu. Is this OK?

Thanks,
Kugan

>
> Richard.
>
>> +		  addr_nonzero = true;
>> +		}
>> +	      else
>> +		{
>> +		  /* If symbol address is nonzero_address.  */
>> +		  node = varpool_node::get (base);
>> +		  if (node && node->nonzero_address ())
>> +		    addr_nonzero = true;
>> +		}
>> +	    }
>>
>> Attached patch passes bootstrap and regression testing on x86_64-linu-gnu.
>>
>> Thanks,
>> Kugan
>>
>>>
>>> Honza
>>>>
>>>> Richard.
>>>>
>>>>> Thanks,
>>>>> Kugan
>>>>>
>>>>>>
>>>>>> Richard.
>>>>>>
>>>>>>> Attached patch bootstraps and regression testing didn't introduce
>>>>>>> any new
>>>>>>> regressions.
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Kugan
>>>>>>>
>>>>>>>
>>>>>>> gcc/ChangeLog:
>>>>>>>
>>>>>>> 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
>>>>>>>
>>>>>>> 	* ipa-prop.c (ipa_compute_jump_functions_for_edge): Set
>>>>>>> 	value range to nonull for ADDR_EXPR.
>>>>>>>
>>>>>>> gcc/testsuite/ChangeLog:
>>>>>>>
>>>>>>> 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
>>>>>>>
>>>>>>> 	* gcc.dg/ipa/vrp5.c: New test.
>>>>>>> 	* gcc.dg/ipa/vrp6.c: New test.
>>>>>>>
>>>>>>
>>>>>
>>>>>
>>>>
>>>> --
>>>> Richard Biener <rguenther@suse.de>
>>>> SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB
>>>> 21284 (AG Nuernberg)
>>
>

[-- Attachment #2: p.txt --]
[-- Type: text/plain, Size: 2535 bytes --]

diff --git a/gcc/ipa-prop.c b/gcc/ipa-prop.c
index 353b638..3fe60d6 100644
--- a/gcc/ipa-prop.c
+++ b/gcc/ipa-prop.c
@@ -1670,9 +1670,26 @@ ipa_compute_jump_functions_for_edge (struct ipa_func_body_info *fbi,
 
       if (POINTER_TYPE_P (TREE_TYPE (arg)))
 	{
-	  if (TREE_CODE (arg) == SSA_NAME
-	      && param_type
-	      && get_ptr_nonnull (arg))
+	  bool addr_nonzero = false;
+
+	  if (TREE_CODE (arg) == ADDR_EXPR)
+	    {
+	      /* See if the ADDR_EXPR is nonnull.  */
+	      tree base = get_base_address (TREE_OPERAND (arg, 0));
+
+	      if (TREE_CODE (base) == CONST_DECL
+		  || TREE_CODE (base) == PARM_DECL
+		  || TREE_CODE (base) == RESULT_DECL)
+		addr_nonzero = true;
+	      else if (VAR_P (base))
+		addr_nonzero = ! is_global_var (base)
+		  || (varpool_node::get (base)->nonzero_address ());
+	    }
+
+	  if ((TREE_CODE (arg) == SSA_NAME
+	       && param_type
+	       && get_ptr_nonnull (arg))
+	      || addr_nonzero)
 	    {
 	      jfunc->vr_known = true;
 	      jfunc->m_vr.type = VR_ANTI_RANGE;
diff --git a/gcc/testsuite/gcc.dg/ipa/vrp5.c b/gcc/testsuite/gcc.dg/ipa/vrp5.c
index e69de29..571798d 100644
--- a/gcc/testsuite/gcc.dg/ipa/vrp5.c
+++ b/gcc/testsuite/gcc.dg/ipa/vrp5.c
@@ -0,0 +1,34 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-ipa-cp-details -fdump-tree-vrp1" } */
+
+static __attribute__((noinline, noclone))
+int foo (int *p)
+{
+  if (!p)
+    return 0;
+  *p = 1;
+}
+
+struct st
+{
+  int a;
+  int b;
+};
+
+int arr1[10];
+int a;
+int bar (struct st *s)
+{
+  int arr2[10];
+  int b;
+  if (!s)
+    return 0;
+  foo (&s->a);
+  foo (&a);
+  foo (&b);
+  foo (&arr1[1]);
+  foo (&arr2[1]);
+}
+
+/* { dg-final { scan-ipa-dump "Setting nonnull for 0" "cp" } } */
+/* { dg-final { scan-tree-dump-times "if" 1 "vrp1" } } */
diff --git a/gcc/testsuite/gcc.dg/ipa/vrp6.c b/gcc/testsuite/gcc.dg/ipa/vrp6.c
index e69de29..971db44 100644
--- a/gcc/testsuite/gcc.dg/ipa/vrp6.c
+++ b/gcc/testsuite/gcc.dg/ipa/vrp6.c
@@ -0,0 +1,34 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-ipa-cp-details -fdump-tree-vrp1" } */
+
+static __attribute__((noinline, noclone))
+int foo (int *p)
+{
+  if (!p)
+    return 0;
+  *p = 1;
+}
+
+struct st
+{
+  int a;
+  int b;
+};
+
+struct st s2;
+int a;
+int bar (struct st *s)
+{
+  struct st s3;
+  int b;
+  if (!s)
+    return 0;
+  foo (&s->a);
+  foo (&s2.a);
+  foo (&s3.a);
+  foo (&a);
+  foo (&b);
+}
+
+/* { dg-final { scan-ipa-dump "Setting nonnull for 0" "cp" } } */
+/* { dg-final { scan-tree-dump-times "if" 1 "vrp1" } } */

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[Richard Biener]

On Thu, 20 Oct 2016, kugan wrote:

> Hi Richard,
> 
> 
> On 20/10/16 18:41, Richard Biener wrote:
> > On Thu, 20 Oct 2016, kugan wrote:
> > 
> > > 
> > > 
> > > On 20/10/16 01:26, Jan Hubicka wrote:
> > > > > > Would excluding weak symbols (I believe I can check DECL_WEAK for
> > > > > > this)
> > > > > > good
> > > > > > enough. Or looking for acceptable subset would work?
> > > > > 
> > > > > I think we should add a symtab helper to tell if address_nonzero_p (if
> > > > > that doesn't aleady exist).
> > > > 
> > > > We have node->nonzero_address()
> > > 
> > > Thanks for the pointer. Here is an attempt. Does this look OK?
> > > 
> > > +	  if (TREE_CODE (arg) == ADDR_EXPR)
> > > +	    {
> > > +	      /* See if the AADR_EXPR is nonnull.  */
> > > +	      varpool_node *node = NULL;
> > > +	      tree base = TREE_OPERAND (arg, 0);
> > > +	      base = get_base_address (base);
> > > +
> > > +	      if (decl_address_ip_invariant_p (base)
> > > +		  || !is_global_var (base))
> > > +		{
> > > +		  /* If the symbol address is local or
> > > +		     constant.  */
> > 
> > "constant" doesn't matter.  You want
> > 
> >    if (TREE_CODE (base) == CONST_DECL
> > || TREE_CODE (base) == PARM_DECL
> > || TREE_CODE (base) == RESULT_DECL)
> >   addr_nonzero = true;
> >    else if (VAR_P (base))
> >      addr_nonzero = ! is_global_var (base) || (varpool_node::get
> > (base)->nonzero_address ());
> 
> Attached patch does this. Bootstrapped and regression tested on
> x86_64-linux-gnu. Is this OK?

Ok with

+           }
+
+         if ((TREE_CODE (arg) == SSA_NAME
+              && param_type
+              && get_ptr_nonnull (arg))
+             || addr_nonzero)
            {


re-factored as

+           }
          else if (TREE_CODE (arg) == SSA_NAME
+              && param_type
+              && get_ptr_nonnull (arg))
            addr_nonzero = true;

          if (addr_nonzero)

Richard.

> Thanks,
> Kugan
> 
> > 
> > Richard.
> > 
> > > +		  addr_nonzero = true;
> > > +		}
> > > +	      else
> > > +		{
> > > +		  /* If symbol address is nonzero_address.  */
> > > +		  node = varpool_node::get (base);
> > > +		  if (node && node->nonzero_address ())
> > > +		    addr_nonzero = true;
> > > +		}
> > > +	    }
> > > 
> > > Attached patch passes bootstrap and regression testing on x86_64-linu-gnu.
> > > 
> > > Thanks,
> > > Kugan
> > > 
> > > > 
> > > > Honza
> > > > > 
> > > > > Richard.
> > > > > 
> > > > > > Thanks,
> > > > > > Kugan
> > > > > > 
> > > > > > > 
> > > > > > > Richard.
> > > > > > > 
> > > > > > > > Attached patch bootstraps and regression testing didn't
> > > > > > > > introduce
> > > > > > > > any new
> > > > > > > > regressions.
> > > > > > > > 
> > > > > > > > Thanks,
> > > > > > > > Kugan
> > > > > > > > 
> > > > > > > > 
> > > > > > > > gcc/ChangeLog:
> > > > > > > > 
> > > > > > > > 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
> > > > > > > > 
> > > > > > > > 	* ipa-prop.c (ipa_compute_jump_functions_for_edge): Set
> > > > > > > > 	value range to nonull for ADDR_EXPR.
> > > > > > > > 
> > > > > > > > gcc/testsuite/ChangeLog:
> > > > > > > > 
> > > > > > > > 2016-10-19  Kugan Vivekanandarajah  <kuganv@linaro.org>
> > > > > > > > 
> > > > > > > > 	* gcc.dg/ipa/vrp5.c: New test.
> > > > > > > > 	* gcc.dg/ipa/vrp6.c: New test.
> > > > > > > > 
> > > > > > > 
> > > > > > 
> > > > > > 
> > > > > 
> > > > > --
> > > > > Richard Biener <rguenther@suse.de>
> > > > > SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton,
> > > > > HRB
> > > > > 21284 (AG Nuernberg)
> > > 
> > 
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nuernberg)

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[Jan Hubicka]

> Hi Richard,
> 
> 
> On 20/10/16 18:41, Richard Biener wrote:
> >On Thu, 20 Oct 2016, kugan wrote:
> >
> >>
> >>
> >>On 20/10/16 01:26, Jan Hubicka wrote:
> >>>>>Would excluding weak symbols (I believe I can check DECL_WEAK for this)
> >>>>>good
> >>>>>enough. Or looking for acceptable subset would work?
> >>>>
> >>>>I think we should add a symtab helper to tell if address_nonzero_p (if
> >>>>that doesn't aleady exist).
> >>>
> >>>We have node->nonzero_address()
> >>
> >>Thanks for the pointer. Here is an attempt. Does this look OK?
> >>
> >>+	  if (TREE_CODE (arg) == ADDR_EXPR)
> >>+	    {
> >>+	      /* See if the AADR_EXPR is nonnull.  */
> >>+	      varpool_node *node = NULL;
> >>+	      tree base = TREE_OPERAND (arg, 0);
> >>+	      base = get_base_address (base);
> >>+
> >>+	      if (decl_address_ip_invariant_p (base)
> >>+		  || !is_global_var (base))
> >>+		{
> >>+		  /* If the symbol address is local or
> >>+		     constant.  */
> >
> >"constant" doesn't matter.  You want
> >
> >   if (TREE_CODE (base) == CONST_DECL
> >|| TREE_CODE (base) == PARM_DECL
> >|| TREE_CODE (base) == RESULT_DECL)
> >  addr_nonzero = true;
> >   else if (VAR_P (base))
Better to check decl_in_symtab_p (decl) 
> >     addr_nonzero = ! is_global_var (base) || (varpool_node::get
> >(base)->nonzero_address ());
and symtab_node::get.

I wonder if we can't unify the logic with tree_expr_nonzero_warnv_p
and corresponding vrp code?

Otherwise the patch looks fine to me.
Honza

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[kugan]

[-- Attachment #1: Type: text/plain, Size: 1750 bytes --]

Hi,

On 20/10/16 23:15, Jan Hubicka wrote:
>> Hi Richard,
>>
>>
>> On 20/10/16 18:41, Richard Biener wrote:
>>> On Thu, 20 Oct 2016, kugan wrote:
>>>
>>>>
>>>>
>>>> On 20/10/16 01:26, Jan Hubicka wrote:
>>>>>>> Would excluding weak symbols (I believe I can check DECL_WEAK for this)
>>>>>>> good
>>>>>>> enough. Or looking for acceptable subset would work?
>>>>>>
>>>>>> I think we should add a symtab helper to tell if address_nonzero_p (if
>>>>>> that doesn't aleady exist).
>>>>>
>>>>> We have node->nonzero_address()
>>>>
>>>> Thanks for the pointer. Here is an attempt. Does this look OK?
>>>>
>>>> +	  if (TREE_CODE (arg) == ADDR_EXPR)
>>>> +	    {
>>>> +	      /* See if the AADR_EXPR is nonnull.  */
>>>> +	      varpool_node *node = NULL;
>>>> +	      tree base = TREE_OPERAND (arg, 0);
>>>> +	      base = get_base_address (base);
>>>> +
>>>> +	      if (decl_address_ip_invariant_p (base)
>>>> +		  || !is_global_var (base))
>>>> +		{
>>>> +		  /* If the symbol address is local or
>>>> +		     constant.  */
>>>
>>> "constant" doesn't matter.  You want
>>>
>>>   if (TREE_CODE (base) == CONST_DECL
>>> || TREE_CODE (base) == PARM_DECL
>>> || TREE_CODE (base) == RESULT_DECL)
>>>  addr_nonzero = true;
>>>   else if (VAR_P (base))
> Better to check decl_in_symtab_p (decl)
>>>     addr_nonzero = ! is_global_var (base) || (varpool_node::get
>>> (base)->nonzero_address ());
> and symtab_node::get.
>
> I wonder if we can't unify the logic with tree_expr_nonzero_warnv_p
> and corresponding vrp code?
Are you saying that we should export tree_expr_nonzero_warnv_p and use 
here with the logic from above added to tree_expr_nonzero_warnv_p as 
shown in the attached patch?

Thanks,
Kugan

>
> Otherwise the patch looks fine to me.
> Honza
>

[-- Attachment #2: p2.txt --]
[-- Type: text/plain, Size: 3638 bytes --]

diff --git a/gcc/fold-const.c b/gcc/fold-const.c
index 89ed89d..6fa42c9 100644
--- a/gcc/fold-const.c
+++ b/gcc/fold-const.c
@@ -8954,7 +8954,7 @@ mask_with_tz (tree type, const wide_int &x, const wide_int &y)
    is undefined, set *STRICT_OVERFLOW_P to true; otherwise, don't
    change *STRICT_OVERFLOW_P.  */
 
-static bool
+bool
 tree_expr_nonzero_warnv_p (tree t, bool *strict_overflow_p)
 {
   tree type = TREE_TYPE (t);
@@ -13402,9 +13402,15 @@ tree_single_nonzero_warnv_p (tree t, bool *strict_overflow_p)
 	  return true;
 
 	/* Constants are never weak.  */
-	if (CONSTANT_CLASS_P (base))
+	if (CONSTANT_CLASS_P (base)
+	    || TREE_CODE (base) == PARM_DECL
+	    || TREE_CODE (base) == RESULT_DECL)
 	  return true;
 
+	if (VAR_P (base)
+	    && (!is_global_var (base)
+		|| varpool_node::get (base)->nonzero_address ()))
+	  return true;
 	return false;
       }
 
diff --git a/gcc/fold-const.h b/gcc/fold-const.h
index bc22c88..a19a27a 100644
--- a/gcc/fold-const.h
+++ b/gcc/fold-const.h
@@ -198,6 +198,7 @@ extern tree fold_build_pointer_plus_loc (location_t loc, tree ptr, tree off);
 
 /* Build and fold a POINTER_PLUS_EXPR at LOC offsetting PTR by OFF.  */
 extern tree fold_build_pointer_plus_hwi_loc (location_t loc, tree ptr, HOST_WIDE_INT off);
+extern bool tree_expr_nonzero_warnv_p (tree t, bool *strict_overflow_p);
 
 #define fold_build_pointer_plus_hwi(p,o) \
 	fold_build_pointer_plus_hwi_loc (UNKNOWN_LOCATION, p, o)
diff --git a/gcc/ipa-prop.c b/gcc/ipa-prop.c
index 353b638..55a2ea1 100644
--- a/gcc/ipa-prop.c
+++ b/gcc/ipa-prop.c
@@ -1670,9 +1670,20 @@ ipa_compute_jump_functions_for_edge (struct ipa_func_body_info *fbi,
 
       if (POINTER_TYPE_P (TREE_TYPE (arg)))
 	{
+	  bool addr_nonzero = false;
+	  bool strict_overflow = false;
+
 	  if (TREE_CODE (arg) == SSA_NAME
 	      && param_type
 	      && get_ptr_nonnull (arg))
+	    addr_nonzero = true;
+	  else if (tree_expr_nonzero_warnv_p (arg, &strict_overflow))
+	    {
+	      if (!strict_overflow)
+		addr_nonzero = true;
+	    }
+
+	  if (addr_nonzero)
 	    {
 	      jfunc->vr_known = true;
 	      jfunc->m_vr.type = VR_ANTI_RANGE;
diff --git a/gcc/testsuite/gcc.dg/ipa/vrp5.c b/gcc/testsuite/gcc.dg/ipa/vrp5.c
index e69de29..571798d 100644
--- a/gcc/testsuite/gcc.dg/ipa/vrp5.c
+++ b/gcc/testsuite/gcc.dg/ipa/vrp5.c
@@ -0,0 +1,34 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-ipa-cp-details -fdump-tree-vrp1" } */
+
+static __attribute__((noinline, noclone))
+int foo (int *p)
+{
+  if (!p)
+    return 0;
+  *p = 1;
+}
+
+struct st
+{
+  int a;
+  int b;
+};
+
+int arr1[10];
+int a;
+int bar (struct st *s)
+{
+  int arr2[10];
+  int b;
+  if (!s)
+    return 0;
+  foo (&s->a);
+  foo (&a);
+  foo (&b);
+  foo (&arr1[1]);
+  foo (&arr2[1]);
+}
+
+/* { dg-final { scan-ipa-dump "Setting nonnull for 0" "cp" } } */
+/* { dg-final { scan-tree-dump-times "if" 1 "vrp1" } } */
diff --git a/gcc/testsuite/gcc.dg/ipa/vrp6.c b/gcc/testsuite/gcc.dg/ipa/vrp6.c
index e69de29..971db44 100644
--- a/gcc/testsuite/gcc.dg/ipa/vrp6.c
+++ b/gcc/testsuite/gcc.dg/ipa/vrp6.c
@@ -0,0 +1,34 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-ipa-cp-details -fdump-tree-vrp1" } */
+
+static __attribute__((noinline, noclone))
+int foo (int *p)
+{
+  if (!p)
+    return 0;
+  *p = 1;
+}
+
+struct st
+{
+  int a;
+  int b;
+};
+
+struct st s2;
+int a;
+int bar (struct st *s)
+{
+  struct st s3;
+  int b;
+  if (!s)
+    return 0;
+  foo (&s->a);
+  foo (&s2.a);
+  foo (&s3.a);
+  foo (&a);
+  foo (&b);
+}
+
+/* { dg-final { scan-ipa-dump "Setting nonnull for 0" "cp" } } */
+/* { dg-final { scan-tree-dump-times "if" 1 "vrp1" } } */

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[Richard Biener]

On Fri, 21 Oct 2016, kugan wrote:

> Hi,
> 
> On 20/10/16 23:15, Jan Hubicka wrote:
> > > Hi Richard,
> > > 
> > > 
> > > On 20/10/16 18:41, Richard Biener wrote:
> > > > On Thu, 20 Oct 2016, kugan wrote:
> > > > 
> > > > > 
> > > > > 
> > > > > On 20/10/16 01:26, Jan Hubicka wrote:
> > > > > > > > Would excluding weak symbols (I believe I can check DECL_WEAK
> > > > > > > > for this)
> > > > > > > > good
> > > > > > > > enough. Or looking for acceptable subset would work?
> > > > > > > 
> > > > > > > I think we should add a symtab helper to tell if address_nonzero_p
> > > > > > > (if
> > > > > > > that doesn't aleady exist).
> > > > > > 
> > > > > > We have node->nonzero_address()
> > > > > 
> > > > > Thanks for the pointer. Here is an attempt. Does this look OK?
> > > > > 
> > > > > +	  if (TREE_CODE (arg) == ADDR_EXPR)
> > > > > +	    {
> > > > > +	      /* See if the AADR_EXPR is nonnull.  */
> > > > > +	      varpool_node *node = NULL;
> > > > > +	      tree base = TREE_OPERAND (arg, 0);
> > > > > +	      base = get_base_address (base);
> > > > > +
> > > > > +	      if (decl_address_ip_invariant_p (base)
> > > > > +		  || !is_global_var (base))
> > > > > +		{
> > > > > +		  /* If the symbol address is local or
> > > > > +		     constant.  */
> > > > 
> > > > "constant" doesn't matter.  You want
> > > > 
> > > >   if (TREE_CODE (base) == CONST_DECL
> > > > || TREE_CODE (base) == PARM_DECL
> > > > || TREE_CODE (base) == RESULT_DECL)
> > > >  addr_nonzero = true;
> > > >   else if (VAR_P (base))
> > Better to check decl_in_symtab_p (decl)
> > > >     addr_nonzero = ! is_global_var (base) || (varpool_node::get
> > > > (base)->nonzero_address ());
> > and symtab_node::get.
> > 
> > I wonder if we can't unify the logic with tree_expr_nonzero_warnv_p
> > and corresponding vrp code?
> Are you saying that we should export tree_expr_nonzero_warnv_p and use here
> with the logic from above added to tree_expr_nonzero_warnv_p as shown in the
> attached patch?

tree_single_nonzero_warnv_p already handles this correctly via
maybe_nonzero_address.  And tree_single_nonzero_warnv_p is already
exported.

Richard.

> Thanks,
> Kugan
> 
> > 
> > Otherwise the patch looks fine to me.
> > Honza
> > 
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nuernberg)

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[kugan]

[-- Attachment #1: Type: text/plain, Size: 2376 bytes --]



On 21/10/16 18:16, Richard Biener wrote:
> On Fri, 21 Oct 2016, kugan wrote:
>
>> Hi,
>>
>> On 20/10/16 23:15, Jan Hubicka wrote:
>>>> Hi Richard,
>>>>
>>>>
>>>> On 20/10/16 18:41, Richard Biener wrote:
>>>>> On Thu, 20 Oct 2016, kugan wrote:
>>>>>
>>>>>>
>>>>>>
>>>>>> On 20/10/16 01:26, Jan Hubicka wrote:
>>>>>>>>> Would excluding weak symbols (I believe I can check DECL_WEAK
>>>>>>>>> for this)
>>>>>>>>> good
>>>>>>>>> enough. Or looking for acceptable subset would work?
>>>>>>>>
>>>>>>>> I think we should add a symtab helper to tell if address_nonzero_p
>>>>>>>> (if
>>>>>>>> that doesn't aleady exist).
>>>>>>>
>>>>>>> We have node->nonzero_address()
>>>>>>
>>>>>> Thanks for the pointer. Here is an attempt. Does this look OK?
>>>>>>
>>>>>> +	  if (TREE_CODE (arg) == ADDR_EXPR)
>>>>>> +	    {
>>>>>> +	      /* See if the AADR_EXPR is nonnull.  */
>>>>>> +	      varpool_node *node = NULL;
>>>>>> +	      tree base = TREE_OPERAND (arg, 0);
>>>>>> +	      base = get_base_address (base);
>>>>>> +
>>>>>> +	      if (decl_address_ip_invariant_p (base)
>>>>>> +		  || !is_global_var (base))
>>>>>> +		{
>>>>>> +		  /* If the symbol address is local or
>>>>>> +		     constant.  */
>>>>>
>>>>> "constant" doesn't matter.  You want
>>>>>
>>>>>   if (TREE_CODE (base) == CONST_DECL
>>>>> || TREE_CODE (base) == PARM_DECL
>>>>> || TREE_CODE (base) == RESULT_DECL)
>>>>>  addr_nonzero = true;
>>>>>   else if (VAR_P (base))
>>> Better to check decl_in_symtab_p (decl)
>>>>>     addr_nonzero = ! is_global_var (base) || (varpool_node::get
>>>>> (base)->nonzero_address ());
>>> and symtab_node::get.
>>>
>>> I wonder if we can't unify the logic with tree_expr_nonzero_warnv_p
>>> and corresponding vrp code?
>> Are you saying that we should export tree_expr_nonzero_warnv_p and use here
>> with the logic from above added to tree_expr_nonzero_warnv_p as shown in the
>> attached patch?
>
> tree_single_nonzero_warnv_p already handles this correctly via
> maybe_nonzero_address.  And tree_single_nonzero_warnv_p is already
> exported.
Indeed. maybe_nonzero_address does look at node->nonzero_address () but 
we still seems to miss TREE_CODE (base) == PARM_DECL and TREE_CODE 
(base) == RESULT_DECL.

Does the attached patch looks OK if no regressions?

Thanks,
Kugan


>
> Richard.
>
>> Thanks,
>> Kugan
>>
>>>
>>> Otherwise the patch looks fine to me.
>>> Honza
>>>
>>
>

[-- Attachment #2: p3.txt --]
[-- Type: text/plain, Size: 2636 bytes --]

diff --git a/gcc/fold-const.c b/gcc/fold-const.c
index 89ed89d..3c1803f 100644
--- a/gcc/fold-const.c
+++ b/gcc/fold-const.c
@@ -13402,7 +13402,9 @@ tree_single_nonzero_warnv_p (tree t, bool *strict_overflow_p)
 	  return true;
 
 	/* Constants are never weak.  */
-	if (CONSTANT_CLASS_P (base))
+	if (CONSTANT_CLASS_P (base)
+	    || TREE_CODE (base) == PARM_DECL
+	    || TREE_CODE (base) == RESULT_DECL)
 	  return true;
 
 	return false;
diff --git a/gcc/ipa-prop.c b/gcc/ipa-prop.c
index 353b638..939fd1d 100644
--- a/gcc/ipa-prop.c
+++ b/gcc/ipa-prop.c
@@ -1670,9 +1670,20 @@ ipa_compute_jump_functions_for_edge (struct ipa_func_body_info *fbi,
 
       if (POINTER_TYPE_P (TREE_TYPE (arg)))
 	{
+	  bool addr_nonzero = false;
+	  bool strict_overflow = false;
+
 	  if (TREE_CODE (arg) == SSA_NAME
 	      && param_type
 	      && get_ptr_nonnull (arg))
+	    addr_nonzero = true;
+	  else if (tree_single_nonzero_warnv_p (arg, &strict_overflow))
+	    {
+	      if (!strict_overflow)
+		addr_nonzero = true;
+	    }
+
+	  if (addr_nonzero)
 	    {
 	      jfunc->vr_known = true;
 	      jfunc->m_vr.type = VR_ANTI_RANGE;
diff --git a/gcc/testsuite/gcc.dg/ipa/vrp5.c b/gcc/testsuite/gcc.dg/ipa/vrp5.c
index e69de29..571798d 100644
--- a/gcc/testsuite/gcc.dg/ipa/vrp5.c
+++ b/gcc/testsuite/gcc.dg/ipa/vrp5.c
@@ -0,0 +1,34 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-ipa-cp-details -fdump-tree-vrp1" } */
+
+static __attribute__((noinline, noclone))
+int foo (int *p)
+{
+  if (!p)
+    return 0;
+  *p = 1;
+}
+
+struct st
+{
+  int a;
+  int b;
+};
+
+int arr1[10];
+int a;
+int bar (struct st *s)
+{
+  int arr2[10];
+  int b;
+  if (!s)
+    return 0;
+  foo (&s->a);
+  foo (&a);
+  foo (&b);
+  foo (&arr1[1]);
+  foo (&arr2[1]);
+}
+
+/* { dg-final { scan-ipa-dump "Setting nonnull for 0" "cp" } } */
+/* { dg-final { scan-tree-dump-times "if" 1 "vrp1" } } */
diff --git a/gcc/testsuite/gcc.dg/ipa/vrp6.c b/gcc/testsuite/gcc.dg/ipa/vrp6.c
index e69de29..971db44 100644
--- a/gcc/testsuite/gcc.dg/ipa/vrp6.c
+++ b/gcc/testsuite/gcc.dg/ipa/vrp6.c
@@ -0,0 +1,34 @@
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-ipa-cp-details -fdump-tree-vrp1" } */
+
+static __attribute__((noinline, noclone))
+int foo (int *p)
+{
+  if (!p)
+    return 0;
+  *p = 1;
+}
+
+struct st
+{
+  int a;
+  int b;
+};
+
+struct st s2;
+int a;
+int bar (struct st *s)
+{
+  struct st s3;
+  int b;
+  if (!s)
+    return 0;
+  foo (&s->a);
+  foo (&s2.a);
+  foo (&s3.a);
+  foo (&a);
+  foo (&b);
+}
+
+/* { dg-final { scan-ipa-dump "Setting nonnull for 0" "cp" } } */
+/* { dg-final { scan-tree-dump-times "if" 1 "vrp1" } } */

Re: [RFC][IPA-VRP] ADDR_EXPR and nonnull

[Richard Biener]

On Fri, 21 Oct 2016, kugan wrote:

> 
> 
> On 21/10/16 18:16, Richard Biener wrote:
> > On Fri, 21 Oct 2016, kugan wrote:
> > 
> > > Hi,
> > > 
> > > On 20/10/16 23:15, Jan Hubicka wrote:
> > > > > Hi Richard,
> > > > > 
> > > > > 
> > > > > On 20/10/16 18:41, Richard Biener wrote:
> > > > > > On Thu, 20 Oct 2016, kugan wrote:
> > > > > > 
> > > > > > > 
> > > > > > > 
> > > > > > > On 20/10/16 01:26, Jan Hubicka wrote:
> > > > > > > > > > Would excluding weak symbols (I believe I can check
> > > > > > > > > > DECL_WEAK
> > > > > > > > > > for this)
> > > > > > > > > > good
> > > > > > > > > > enough. Or looking for acceptable subset would work?
> > > > > > > > > 
> > > > > > > > > I think we should add a symtab helper to tell if
> > > > > > > > > address_nonzero_p
> > > > > > > > > (if
> > > > > > > > > that doesn't aleady exist).
> > > > > > > > 
> > > > > > > > We have node->nonzero_address()
> > > > > > > 
> > > > > > > Thanks for the pointer. Here is an attempt. Does this look OK?
> > > > > > > 
> > > > > > > +	  if (TREE_CODE (arg) == ADDR_EXPR)
> > > > > > > +	    {
> > > > > > > +	      /* See if the AADR_EXPR is nonnull.  */
> > > > > > > +	      varpool_node *node = NULL;
> > > > > > > +	      tree base = TREE_OPERAND (arg, 0);
> > > > > > > +	      base = get_base_address (base);
> > > > > > > +
> > > > > > > +	      if (decl_address_ip_invariant_p (base)
> > > > > > > +		  || !is_global_var (base))
> > > > > > > +		{
> > > > > > > +		  /* If the symbol address is local or
> > > > > > > +		     constant.  */
> > > > > > 
> > > > > > "constant" doesn't matter.  You want
> > > > > > 
> > > > > >   if (TREE_CODE (base) == CONST_DECL
> > > > > > || TREE_CODE (base) == PARM_DECL
> > > > > > || TREE_CODE (base) == RESULT_DECL)
> > > > > >  addr_nonzero = true;
> > > > > >   else if (VAR_P (base))
> > > > Better to check decl_in_symtab_p (decl)
> > > > > >     addr_nonzero = ! is_global_var (base) || (varpool_node::get
> > > > > > (base)->nonzero_address ());
> > > > and symtab_node::get.
> > > > 
> > > > I wonder if we can't unify the logic with tree_expr_nonzero_warnv_p
> > > > and corresponding vrp code?
> > > Are you saying that we should export tree_expr_nonzero_warnv_p and use
> > > here
> > > with the logic from above added to tree_expr_nonzero_warnv_p as shown in
> > > the
> > > attached patch?
> > 
> > tree_single_nonzero_warnv_p already handles this correctly via
> > maybe_nonzero_address.  And tree_single_nonzero_warnv_p is already
> > exported.
> Indeed. maybe_nonzero_address does look at node->nonzero_address () but we
> still seems to miss TREE_CODE (base) == PARM_DECL and TREE_CODE (base) ==
> RESULT_DECL.

Should be handled by

        /* Function local objects are never NULL.  */
        if (DECL_P (base)
            && (DECL_CONTEXT (base)
                && TREE_CODE (DECL_CONTEXT (base)) == FUNCTION_DECL
                && auto_var_in_fn_p (base, DECL_CONTEXT (base))))
          return true;

> Does the attached patch looks OK if no regressions?

+         else if (tree_single_nonzero_warnv_p (arg, &strict_overflow))
+           {
+             if (!strict_overflow)
+               addr_nonzero = true;

you don't need to care for strict_overflow here.

The patch is ok with that removed and the fold-const.c changes dropped.

Richard.

> Thanks,
> Kugan
> 
> 
> > 
> > Richard.
> > 
> > > Thanks,
> > > Kugan
> > > 
> > > > 
> > > > Otherwise the patch looks fine to me.
> > > > Honza
> > > > 
> > > 
> > 
> 

-- 
Richard Biener <rguenther@suse.de>
SUSE LINUX GmbH, GF: Felix Imendoerffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nuernberg)