[patch, fortran] Fix PR 79956, part two

[patch, fortran] Fix PR 79956, part two

[Thomas Koenig]

[-- Attachment #1: Type: text/plain, Size: 1251 bytes --]

Hello world,

the attached patch fixes another occurence of PR 79956.
In this case, we did

       sdim = GFC_DESCRIPTOR_RANK (source);
..
       for (n = 0; n < sdim; n++)
         {
            sstride[n] = GFC_DESCRIPTOR_STRIDE(source,n)
         }

Now, we know that sdim can never be zero or lower, but
the compiler doesn't, so it warned.  (Why that didn't
happen on x86_64 is an unsolved mystery so far).

So, the attached patch inserts a call to internal_error
when sdim is lower than one.

Regression-tested. OK for trunk?

	Thomas

2017-03-12  Thomas Koenig  <tkoenig@gcc.gnu.org>

	PR libfortran/79956
	* m4/reshape.m4 (reshape_'rtype_ccode`): Add check for sdim < 1 so
	that later assignment from sstride[0] does not cause problems.
	* intrinsic/reshape_generic.c (reshape_internal):  Likweise.
	* generated/reshape_c10.c: Regenerated.
	* generated/reshape_c16.c: Regenerated.
	* generated/reshape_c4.c: Regenerated.
	* generated/reshape_c8.c: Regenerated.
	* generated/reshape_i16.c: Regenerated.
	* generated/reshape_i4.c: Regenerated.
	* generated/reshape_i8.c: Regenerated.
	* generated/reshape_r10.c: Regenerated.
	* generated/reshape_r16.c: Regenerated.
	* generated/reshape_r4.c: Regenerated.
	* generated/reshape_r8.c: Regenerated.

[-- Attachment #2: p4.diff --]
[-- Type: text/x-patch, Size: 7311 bytes --]

Index: generated/reshape_c10.c
===================================================================
--- generated/reshape_c10.c	(Revision 245760)
+++ generated/reshape_c10.c	(Arbeitskopie)
@@ -232,6 +232,12 @@ reshape_c10 (gfc_array_c10 * const restrict ret,
     }
 
   sdim = GFC_DESCRIPTOR_RANK (source);
+
+  /* This is needed so that gcc knows that n is at least one,
+     and that sstride[0] is always initialized.  */
+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");
+
   ssize = 1;
   sempty = 0;
   for (n = 0; n < sdim; n++)
Index: generated/reshape_c16.c
===================================================================
--- generated/reshape_c16.c	(Revision 245760)
+++ generated/reshape_c16.c	(Arbeitskopie)
@@ -232,6 +232,12 @@ reshape_c16 (gfc_array_c16 * const restrict ret,
     }
 
   sdim = GFC_DESCRIPTOR_RANK (source);
+
+  /* This is needed so that gcc knows that n is at least one,
+     and that sstride[0] is always initialized.  */
+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");
+
   ssize = 1;
   sempty = 0;
   for (n = 0; n < sdim; n++)
Index: generated/reshape_c4.c
===================================================================
--- generated/reshape_c4.c	(Revision 245760)
+++ generated/reshape_c4.c	(Arbeitskopie)
@@ -232,6 +232,12 @@ reshape_c4 (gfc_array_c4 * const restrict ret,
     }
 
   sdim = GFC_DESCRIPTOR_RANK (source);
+
+  /* This is needed so that gcc knows that n is at least one,
+     and that sstride[0] is always initialized.  */
+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");
+
   ssize = 1;
   sempty = 0;
   for (n = 0; n < sdim; n++)
Index: generated/reshape_c8.c
===================================================================
--- generated/reshape_c8.c	(Revision 245760)
+++ generated/reshape_c8.c	(Arbeitskopie)
@@ -232,6 +232,12 @@ reshape_c8 (gfc_array_c8 * const restrict ret,
     }
 
   sdim = GFC_DESCRIPTOR_RANK (source);
+
+  /* This is needed so that gcc knows that n is at least one,
+     and that sstride[0] is always initialized.  */
+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");
+
   ssize = 1;
   sempty = 0;
   for (n = 0; n < sdim; n++)
Index: generated/reshape_i16.c
===================================================================
--- generated/reshape_i16.c	(Revision 245760)
+++ generated/reshape_i16.c	(Arbeitskopie)
@@ -232,6 +232,12 @@ reshape_16 (gfc_array_i16 * const restrict ret,
     }
 
   sdim = GFC_DESCRIPTOR_RANK (source);
+
+  /* This is needed so that gcc knows that n is at least one,
+     and that sstride[0] is always initialized.  */
+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");
+
   ssize = 1;
   sempty = 0;
   for (n = 0; n < sdim; n++)
Index: generated/reshape_i4.c
===================================================================
--- generated/reshape_i4.c	(Revision 245760)
+++ generated/reshape_i4.c	(Arbeitskopie)
@@ -232,6 +232,12 @@ reshape_4 (gfc_array_i4 * const restrict ret,
     }
 
   sdim = GFC_DESCRIPTOR_RANK (source);
+
+  /* This is needed so that gcc knows that n is at least one,
+     and that sstride[0] is always initialized.  */
+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");
+
   ssize = 1;
   sempty = 0;
   for (n = 0; n < sdim; n++)
Index: generated/reshape_i8.c
===================================================================
--- generated/reshape_i8.c	(Revision 245760)
+++ generated/reshape_i8.c	(Arbeitskopie)
@@ -232,6 +232,12 @@ reshape_8 (gfc_array_i8 * const restrict ret,
     }
 
   sdim = GFC_DESCRIPTOR_RANK (source);
+
+  /* This is needed so that gcc knows that n is at least one,
+     and that sstride[0] is always initialized.  */
+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");
+
   ssize = 1;
   sempty = 0;
   for (n = 0; n < sdim; n++)
Index: generated/reshape_r10.c
===================================================================
--- generated/reshape_r10.c	(Revision 245760)
+++ generated/reshape_r10.c	(Arbeitskopie)
@@ -232,6 +232,12 @@ reshape_r10 (gfc_array_r10 * const restrict ret,
     }
 
   sdim = GFC_DESCRIPTOR_RANK (source);
+
+  /* This is needed so that gcc knows that n is at least one,
+     and that sstride[0] is always initialized.  */
+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");
+
   ssize = 1;
   sempty = 0;
   for (n = 0; n < sdim; n++)
Index: generated/reshape_r16.c
===================================================================
--- generated/reshape_r16.c	(Revision 245760)
+++ generated/reshape_r16.c	(Arbeitskopie)
@@ -232,6 +232,12 @@ reshape_r16 (gfc_array_r16 * const restrict ret,
     }
 
   sdim = GFC_DESCRIPTOR_RANK (source);
+
+  /* This is needed so that gcc knows that n is at least one,
+     and that sstride[0] is always initialized.  */
+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");
+
   ssize = 1;
   sempty = 0;
   for (n = 0; n < sdim; n++)
Index: generated/reshape_r4.c
===================================================================
--- generated/reshape_r4.c	(Revision 245760)
+++ generated/reshape_r4.c	(Arbeitskopie)
@@ -232,6 +232,12 @@ reshape_r4 (gfc_array_r4 * const restrict ret,
     }
 
   sdim = GFC_DESCRIPTOR_RANK (source);
+
+  /* This is needed so that gcc knows that n is at least one,
+     and that sstride[0] is always initialized.  */
+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");
+
   ssize = 1;
   sempty = 0;
   for (n = 0; n < sdim; n++)
Index: generated/reshape_r8.c
===================================================================
--- generated/reshape_r8.c	(Revision 245760)
+++ generated/reshape_r8.c	(Arbeitskopie)
@@ -232,6 +232,12 @@ reshape_r8 (gfc_array_r8 * const restrict ret,
     }
 
   sdim = GFC_DESCRIPTOR_RANK (source);
+
+  /* This is needed so that gcc knows that n is at least one,
+     and that sstride[0] is always initialized.  */
+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");
+
   ssize = 1;
   sempty = 0;
   for (n = 0; n < sdim; n++)
Index: intrinsics/reshape_generic.c
===================================================================
--- intrinsics/reshape_generic.c	(Revision 245760)
+++ intrinsics/reshape_generic.c	(Arbeitskopie)
@@ -158,6 +158,12 @@ reshape_internal (parray *ret, parray *source, sha
 
       source_extent = 1;
       sdim = GFC_DESCRIPTOR_RANK (source);
+
+      /* This is needed so that gcc knows that n is at least one,
+	 and that sstride[0] is always initialized.  */
+      if (sdim < 1)
+	internal_error (NULL, "Source array cannot be scalar");
+
       for (n = 0; n < sdim; n++)
 	{
 	  index_type se;
Index: m4/reshape.m4
===================================================================
--- m4/reshape.m4	(Revision 245760)
+++ m4/reshape.m4	(Arbeitskopie)
@@ -236,6 +236,12 @@ reshape_'rtype_ccode` ('rtype` * const restrict re
     }
 
   sdim = GFC_DESCRIPTOR_RANK (source);
+
+  /* This is needed so that gcc knows that n is at least one,
+     and that sstride[0] is always initialized.  */
+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");
+
   ssize = 1;
   sempty = 0;
   for (n = 0; n < sdim; n++)

Re: [patch, fortran] Fix PR 79956, part two

[Jerry DeLisle]

On 03/12/2017 02:27 PM, Thomas Koenig wrote:
> Hello world,
>
> the attached patch fixes another occurence of PR 79956.
> In this case, we did
>
>       sdim = GFC_DESCRIPTOR_RANK (source);
> ..
>       for (n = 0; n < sdim; n++)
>         {
>            sstride[n] = GFC_DESCRIPTOR_STRIDE(source,n)
>         }
>
> Now, we know that sdim can never be zero or lower, but
> the compiler doesn't, so it warned.  (Why that didn't
> happen on x86_64 is an unsolved mystery so far).
>
> So, the attached patch inserts a call to internal_error
> when sdim is lower than one.
>

Can this condition:

+  if (sdim < 1)
+    internal_error (NULL, "Source array cannot be scalar");

Ever exist in reality?

Does adding this statement add a small runtime performance penalty or is it 
optimized out?

Jerry

Re: [patch, fortran] Fix PR 79956, part two

[Thomas Koenig]

Hi Jerry,

> +  if (sdim < 1)
> +    internal_error (NULL, "Source array cannot be scalar");
>
> Ever exist in reality?

I think it can only occur when memory corruption has happened.

> Does adding this statement add a small runtime performance penalty or is
> it optimized out?

I think it is still tested at runtime.  Sometimes it is hard to read the
generated assembly...

Regards

	Thomas

Re: [patch, fortran] Fix PR 79956, part two

[Steve Kargl]

On Mon, Mar 13, 2017 at 12:10:05AM +0100, Thomas Koenig wrote:
> Hi Jerry,
> 
> > +  if (sdim < 1)
> > +    internal_error (NULL, "Source array cannot be scalar");
> >
> > Ever exist in reality?
> 
> I think it can only occur when memory corruption has happened.
> 
> > Does adding this statement add a small runtime performance penalty or is
> > it optimized out?
> 
> I think it is still tested at runtime.  Sometimes it is hard to read the
> generated assembly...
> 

Please, don't add a  performance penalty to gfortran
to silence a silly gcc warning.
-- 
Steve
20161221 https://www.youtube.com/watch?v=IbCHE-hONow

Re: [patch, fortran] Fix PR 79956, part two

[Richard Biener]

On Mon, Mar 13, 2017 at 2:57 AM, Steve Kargl
<sgk@troutmask.apl.washington.edu> wrote:
> On Mon, Mar 13, 2017 at 12:10:05AM +0100, Thomas Koenig wrote:
>> Hi Jerry,
>>
>> > +  if (sdim < 1)
>> > +    internal_error (NULL, "Source array cannot be scalar");
>> >
>> > Ever exist in reality?
>>
>> I think it can only occur when memory corruption has happened.
>>
>> > Does adding this statement add a small runtime performance penalty or is
>> > it optimized out?
>>
>> I think it is still tested at runtime.  Sometimes it is hard to read the
>> generated assembly...
>>
>
> Please, don't add a  performance penalty to gfortran
> to silence a silly gcc warning.

Note it helps GCC in optimizing the following loops - GCC then can assume
the loops are always executed, saving a loop header copy and avoding the
threading code to separate the case of sdim == 0 (that just generates extra
dead code as well).

The "silly" warning tells you GCC has gone amok thinking those cases
can get special-casing for speed.

Bounding input values is always a good idea.  If you think they cannot happen
at runtime a possibility is to use sth like gcc_assert in system.h:

#define gcc_assert(EXPR)                                                \
  ((void)(__builtin_expect (!(EXPR), 0) ? __builtin_unreachable (), 0 : 0))

this does not result in any extra code generation but still allows GCC
to constrain
the values you assert (as a bonus you can build the library with a special flag
turning __builtin_unreachable () into a runtime error).

Richard.

> --
> Steve
> 20161221 https://www.youtube.com/watch?v=IbCHE-hONow