[v3] Provide a C++ version of iseqsig

Message ID 20171114132152.22775-1-gabriel@inconstante.eti.br
State New
Headers show
Series
  • [v3] Provide a C++ version of iseqsig
Related show

Commit Message

Gabriel F. T. Gomes Nov. 14, 2017, 1:21 p.m.
Changes since v2:

  - For __cplusplus >= 201103L use __decltype;
  - Otherwise, use __typeof only if compiler defines __GNUC__;
  - Use __MATH_TG for older C++ and not GNU compiler (will fail if
    support for float128 is on).

Changes since v1:

  - Fixed namespace violations in function arguments;
  - Removed additional argument in __iseqsig_type;
  - Templates for __iseqsig_type now have a single parameter, to avoid
    having two different instantiations for different ordering of the
    same pair of types (e.g.: (float, double) and (double, float));
  - iseqsig is now a real-function, instead of a macro.
  - __MATH_EVAL_FMT2 is used whithin typeof and does not evaluate in
    runtime.
  - Added testcase.

-- 8< --
In C++ mode, __MATH_TG cannot be used for defining iseqsig, because
__MATH_TG relies on __builtin_types_compatible_p, which is a C-only
builtin.  This is true when float128 is provided as an ABI-distinct type
from long double.

Moreover, the comparison macros from ISO C take two floating-point
arguments, which need not have the same type.  Choosing what underlying
function to call requires evaluating the formats of the arguments, then
selecting which is wider.  The macro __MATH_EVAL_FMT2 provides this
information, however, only the type of the macro expansion is relevant
(actually evaluating the expression would be incorrect).

This patch provides a C++ version of iseqsig, in which only the type of
__MATH_EVAL_FMT2 (__typeof or __decltype) is used as a template
parameter for __iseqsig_type.  This function calls the appropriate
underlying function.

Tested for powerpc64le and x86_64.

	[BZ #22377]
	* math/Makefile [C++] (tests): Add test for iseqsig.
	* math/math.h [C++] (iseqsig): New implementation, which does
	not rely on __MATH_TG/__builtin_types_compatible_p.
	* math/test-math-iseqsig.cc: New file.
	* sysdeps/powerpc/powerpc64le/Makefile
	(CFLAGS-test-math-iseqsig.cc): New variable.
---
 math/Makefile                        |   2 +-
 math/math.h                          |  72 ++++++++++++++++++++++-
 math/test-math-iseqsig.cc            | 111 +++++++++++++++++++++++++++++++++++
 sysdeps/powerpc/powerpc64le/Makefile |   6 +-
 4 files changed, 187 insertions(+), 4 deletions(-)
 create mode 100644 math/test-math-iseqsig.cc

Comments

Florian Weimer Nov. 14, 2017, 1:35 p.m. | #1
On 11/14/2017 02:21 PM, Gabriel F. T. Gomes wrote:

> diff --git a/math/math.h b/math/math.h
> index 326fd8ebe1..acc909d37f 100644
> --- a/math/math.h
> +++ b/math/math.h
> @@ -1152,8 +1152,76 @@ iszero (__T __val)
>   
>   /* Return X == Y but raising "invalid" and setting errno if X or Y is
>      a NaN.  */
> -# define iseqsig(x, y) \
> -  __MATH_TG (__MATH_EVAL_FMT2 (x, y), __iseqsig, ((x), (y)))
> +# if !defined __cplusplus || (__cplusplus < 201103L && !defined __GNUC__)
> +#  define iseqsig(x, y) \
> +   __MATH_TG (__MATH_EVAL_FMT2 (x, y), __iseqsig, ((x), (y)))
> +# else
> +/* In C++ mode, __MATH_TG cannot be used, because it relies on
> +   __builtin_types_compatible_p, which is a C-only builtin.  Moreover,
> +   the comparison macros from ISO C take two floating-point arguments,
> +   which need not have the same type.  Choosing what underlying function
> +   to call requires evaluating the formats of the arguments, then
> +   selecting which is wider.  The macro __MATH_EVAL_FMT2 provides this
> +   information, however, only the type of the macro expansion is
> +   relevant (actually evaluating the expression would be incorrect).
> +   Thus, the type is used as a template parameter for __iseqsig_type,
> +   which calls the appropriate underlying function.  */

> +template<typename _T1, typename _T2>
> +inline int
> +iseqsig(_T1 __x, _T2 __y) throw()
> +{
> +#  if __cplusplus >= 201103L
> +  typedef __decltype (__MATH_EVAL_FMT2 (__x, __y)) _T3;
> +#  else
> +  typedef __typeof (__MATH_EVAL_FMT2 (__x, __y)) _T3;
> +#  endif
> +  return __iseqsig_type<_T3>::__call(__x, __y);
> +}
> +
> +} /* extern "C++" */
> +# endif /* __cplusplus */

Would these two expressions have the same types, assuming _T1 and _T2 
are the template parameters from the iseqsig definition?

   __MATH_EVAL_FMT2 (__x, __y)
   __MATH_EVAL_FMT2 (_T1 (), _T2 ())

I believe the second expression would be safe to evaluate, so it could 
be used to select a suitable inline function.  This would then work with 
any C++ version.

But the current approach is okay as well.

> diff --git a/math/test-math-iseqsig.cc b/math/test-math-iseqsig.cc
> new file mode 100644
> index 0000000000..0316340638
> --- /dev/null
> +++ b/math/test-math-iseqsig.cc

> +static bool errors;

Maybe use support_record_failure from <support/check.h> instead?

> +static void
> +check (int actual, int expected, const char *actual_expr, int line)
> +{
> +  if (actual != expected)
> +    {
> +      errors = true;
> +      printf ("%s:%d: error: %s\n", __FILE__, line, actual_expr);
> +      printf ("%s:%d:   expected: %d\n", __FILE__, line, expected);
> +      printf ("%s:%d:   actual: %d\n", __FILE__, line, actual);
> +    }
> +}
> +
> +#define CHECK(actual, expected) \
> +  check ((actual), (expected), #actual, __LINE__)

I think my memory protection key patches contain a more general 
implementation of this called TEST_COMPARE.  Adhemerval requested just a 
minor documentation change, so I could commit that separately if you 
want to use it.

Thanks,
Florian
Joseph Myers Nov. 14, 2017, 4:40 p.m. | #2
On Tue, 14 Nov 2017, Florian Weimer wrote:

> Would these two expressions have the same types, assuming _T1 and _T2 are the
> template parameters from the iseqsig definition?
> 
>   __MATH_EVAL_FMT2 (__x, __y)
>   __MATH_EVAL_FMT2 (_T1 (), _T2 ())

Yes, they should always have the same type.
Gabriel F. T. Gomes Nov. 17, 2017, 5:25 p.m. | #3
On Tue, 14 Nov 2017, Florian Weimer wrote:

>On 11/14/2017 02:21 PM, Gabriel F. T. Gomes wrote:
>
>> +template<typename _T1, typename _T2>
>> +inline int
>> +iseqsig(_T1 __x, _T2 __y) throw()
>> +{
>> +#  if __cplusplus >= 201103L
>> +  typedef __decltype (__MATH_EVAL_FMT2 (__x, __y)) _T3;
>> +#  else
>> +  typedef __typeof (__MATH_EVAL_FMT2 (__x, __y)) _T3;
>> +#  endif
>> +  return __iseqsig_type<_T3>::__call(__x, __y);
>> +}
>
>Would these two expressions have the same types, assuming _T1 and _T2 
>are the template parameters from the iseqsig definition?
>
>   __MATH_EVAL_FMT2 (__x, __y)
>   __MATH_EVAL_FMT2 (_T1 (), _T2 ())
>
>I believe the second expression would be safe to evaluate, so it could 
>be used to select a suitable inline function.  This would then work with 
>any C++ version.

Do you mean that I should go back to the first version of the patch (that
worked with function overloading and had a third parameter)?

For instance:

template <typename _T1, typename _T2> inline int
__iseqsig_type (float, _T1 x, _T2 y)
{
  return __iseqsigf (x, y);
}

/* ...  Likewise for other types.  */

template <typename _T1, typename _T2> inline int
iseqsig (_T1 __x, _T2 __y)
{
  _T1 __w = _T1 ();
  _T2 __z = _T2 ();
  return __iseqsig_type (__MATH_EVAL_FMT2 (__w, __z), __x, __y);
}

I am asking it, because on the second and third versions of this patch,
the selection of the underlying function is through a template parameter,
so I think we need to use typeof or decltype (at least I think that the
template parameter requires it).

>Maybe use support_record_failure from <support/check.h> instead?

I already adapted locally to use this, but I'll wait for your answer to my
question above, before sending a new version.

>> +static void
>> +check (int actual, int expected, const char *actual_expr, int line)
>> +{
>> +  if (actual != expected)
>> +    {
>> +      errors = true;
>> +      printf ("%s:%d: error: %s\n", __FILE__, line, actual_expr);
>> +      printf ("%s:%d:   expected: %d\n", __FILE__, line, expected);
>> +      printf ("%s:%d:   actual: %d\n", __FILE__, line, actual);
>> +    }
>> +}
>> +
>> +#define CHECK(actual, expected) \
>> +  check ((actual), (expected), #actual, __LINE__)  
>
>I think my memory protection key patches contain a more general 
>implementation of this called TEST_COMPARE.  Adhemerval requested just a 
>minor documentation change, so I could commit that separately if you 
>want to use it.

I can wait for your changes to go in in their own timing, then I'd gladly
write a clean up to use them (here and in other similar tests).
Florian Weimer Nov. 18, 2017, 12:18 p.m. | #4
On 11/17/2017 06:25 PM, Gabriel F. T. Gomes wrote:
>> Would these two expressions have the same types, assuming _T1 and _T2
>> are the template parameters from the iseqsig definition?
>>
>>    __MATH_EVAL_FMT2 (__x, __y)
>>    __MATH_EVAL_FMT2 (_T1 (), _T2 ())
>>
>> I believe the second expression would be safe to evaluate, so it could
>> be used to select a suitable inline function.  This would then work with
>> any C++ version.

> Do you mean that I should go back to the first version of the patch (that
> worked with function overloading and had a third parameter)?

I do not have a strong preference here.  I'm fine with the use of 
decltype/__typeof as well.  I just wanted to point out this other 
solution in case we need it at some point.

Thanks,
Florian

Patch

diff --git a/math/Makefile b/math/Makefile
index b2bd3d3bcc..de4106b0ff 100644
--- a/math/Makefile
+++ b/math/Makefile
@@ -215,7 +215,7 @@  tests-static += atest-exp atest-sincos atest-exp2
 
 ifneq (,$(CXX))
 tests += test-math-isinff test-math-iszero test-math-issignaling \
-	 test-math-iscanonical test-math-cxx11
+	 test-math-iscanonical test-math-cxx11 test-math-iseqsig
 endif
 
 ifneq (no,$(PERL))
diff --git a/math/math.h b/math/math.h
index 326fd8ebe1..acc909d37f 100644
--- a/math/math.h
+++ b/math/math.h
@@ -1152,8 +1152,76 @@  iszero (__T __val)
 
 /* Return X == Y but raising "invalid" and setting errno if X or Y is
    a NaN.  */
-# define iseqsig(x, y) \
-  __MATH_TG (__MATH_EVAL_FMT2 (x, y), __iseqsig, ((x), (y)))
+# if !defined __cplusplus || (__cplusplus < 201103L && !defined __GNUC__)
+#  define iseqsig(x, y) \
+   __MATH_TG (__MATH_EVAL_FMT2 (x, y), __iseqsig, ((x), (y)))
+# else
+/* In C++ mode, __MATH_TG cannot be used, because it relies on
+   __builtin_types_compatible_p, which is a C-only builtin.  Moreover,
+   the comparison macros from ISO C take two floating-point arguments,
+   which need not have the same type.  Choosing what underlying function
+   to call requires evaluating the formats of the arguments, then
+   selecting which is wider.  The macro __MATH_EVAL_FMT2 provides this
+   information, however, only the type of the macro expansion is
+   relevant (actually evaluating the expression would be incorrect).
+   Thus, the type is used as a template parameter for __iseqsig_type,
+   which calls the appropriate underlying function.  */
+extern "C++" {
+template<typename> struct __iseqsig_type;
+
+template<> struct __iseqsig_type<float>
+{
+  static int __call(float __x, float __y) throw()
+  {
+    return __iseqsigf (__x, __y);
+  }
+};
+
+template<> struct __iseqsig_type<double>
+{
+  static int __call(double __x, double __y) throw()
+  {
+    return __iseqsig (__x, __y);
+  }
+};
+
+template<> struct __iseqsig_type<long double>
+{
+  static int __call(double __x, double __y) throw()
+  {
+#  ifndef __NO_LONG_DOUBLE_MATH
+    return __iseqsigl (__x, __y);
+#  else
+    return __iseqsig (__x, __y);
+#  endif
+  }
+};
+
+#  if __HAVE_DISTINCT_FLOAT128
+template<> struct __iseqsig_type<_Float128>
+{
+  static int __call(_Float128 __x, _Float128 __y) throw()
+  {
+    return __iseqsigf128 (__x, __y);
+  }
+};
+#  endif
+
+template<typename _T1, typename _T2>
+inline int
+iseqsig(_T1 __x, _T2 __y) throw()
+{
+#  if __cplusplus >= 201103L
+  typedef __decltype (__MATH_EVAL_FMT2 (__x, __y)) _T3;
+#  else
+  typedef __typeof (__MATH_EVAL_FMT2 (__x, __y)) _T3;
+#  endif
+  return __iseqsig_type<_T3>::__call(__x, __y);
+}
+
+} /* extern "C++" */
+# endif /* __cplusplus */
+
 #endif
 
 __END_DECLS
diff --git a/math/test-math-iseqsig.cc b/math/test-math-iseqsig.cc
new file mode 100644
index 0000000000..0316340638
--- /dev/null
+++ b/math/test-math-iseqsig.cc
@@ -0,0 +1,111 @@ 
+/* Test for the C++ implementation of iseqsig.
+   Copyright (C) 2017 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C 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
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#define _GNU_SOURCE 1
+#include <math.h>
+#include <stdio.h>
+
+#include <limits>
+
+/* There is no NaN for _Float128 in std::numeric_limits.
+   Include ieee754_float128.h and use the bitfields in the union
+   ieee854_float128.ieee_nan to build a NaN.  */
+#if __HAVE_DISTINCT_FLOAT128
+# include <ieee754_float128.h>
+#endif
+
+static bool errors;
+
+static void
+check (int actual, int expected, const char *actual_expr, int line)
+{
+  if (actual != expected)
+    {
+      errors = true;
+      printf ("%s:%d: error: %s\n", __FILE__, line, actual_expr);
+      printf ("%s:%d:   expected: %d\n", __FILE__, line, expected);
+      printf ("%s:%d:   actual: %d\n", __FILE__, line, actual);
+    }
+}
+
+#define CHECK(actual, expected) \
+  check ((actual), (expected), #actual, __LINE__)
+
+template <class T1, class T2>
+static void
+check_type ()
+{
+  T1 t1 = 0;
+  T2 t2 = 0;
+  CHECK (iseqsig (t1, t2), 1);
+
+  t2 = 1;
+  CHECK (iseqsig (t1, t2), 0);
+
+  if (std::numeric_limits<T1>::has_quiet_NaN
+      && std::numeric_limits<T2>::has_quiet_NaN)
+    {
+      CHECK (iseqsig (std::numeric_limits<T1>::quiet_NaN (), t2), 0);
+      CHECK (iseqsig (t1, std::numeric_limits<T2>::quiet_NaN ()), 0);
+      CHECK (iseqsig (std::numeric_limits<T1>::quiet_NaN (),
+		      std::numeric_limits<T2>::quiet_NaN ()), 0);
+    }
+}
+
+#if __HAVE_DISTINCT_FLOAT128
+static void
+check_float128 ()
+{
+  ieee854_float128 q1, q2, q3_nan;
+
+  q1.d = 0;
+  q2.d = 1;
+  q3_nan.ieee_nan.negative = 0;
+  q3_nan.ieee_nan.exponent = 0x7FFF;
+  q3_nan.ieee_nan.quiet_nan = 1;
+  q3_nan.ieee_nan.mantissa0 = 0x0000;
+  q3_nan.ieee_nan.mantissa1 = 0x00000000;
+  q3_nan.ieee_nan.mantissa2 = 0x00000000;
+  q3_nan.ieee_nan.mantissa3 = 0x00000000;
+
+  CHECK (iseqsig (q1.d, q1.d), 1);
+  CHECK (iseqsig (q1.d, q2.d), 0);
+  CHECK (iseqsig (q1.d, q3_nan.d), 0);
+  CHECK (iseqsig (q3_nan.d, q3_nan.d), 0);
+}
+#endif
+
+static int
+do_test (void)
+{
+  check_type<float, float> ();
+  check_type<float, double> ();
+  check_type<float, long double> ();
+  check_type<double, float> ();
+  check_type<double, double> ();
+  check_type<double, long double> ();
+  check_type<long double, float> ();
+  check_type<long double, double> ();
+  check_type<long double, long double> ();
+#if __HAVE_DISTINCT_FLOAT128
+  check_float128 ();
+#endif
+  return errors;
+}
+
+#include <support/test-driver.c>
diff --git a/sysdeps/powerpc/powerpc64le/Makefile b/sysdeps/powerpc/powerpc64le/Makefile
index f554a791b7..9b8bd138cd 100644
--- a/sysdeps/powerpc/powerpc64le/Makefile
+++ b/sysdeps/powerpc/powerpc64le/Makefile
@@ -17,9 +17,13 @@  $(foreach suf,$(all-object-suffixes),$(objpfx)test-float128%$(suf)): CFLAGS += -
 $(foreach suf,$(all-object-suffixes),$(objpfx)test-ifloat128%$(suf)): CFLAGS += -mfloat128
 CFLAGS-libm-test-support-float128.c += -mfloat128
 CFLAGS-test-math-iscanonical.cc += -mfloat128
+CFLAGS-test-math-iseqsig.cc += -mfloat128
 CFLAGS-test-math-issignaling.cc += -mfloat128
 CFLAGS-test-math-iszero.cc += -mfloat128
-$(objpfx)test-float128% $(objpfx)test-ifloat128% $(objpfx)test-math-iszero: \
+$(foreach test, \
+	  test-float128% test-ifloat128% test-math-iscanonical \
+	  test-math-iseqsig test-math-issignaling test-math-iszero, \
+	  $(objpfx)$(test)): \
   gnulib-tests += $(f128-loader-link)
 endif