[Mesa-dev] Mesa (master): Use signbit() in IS_NEGATIVE and DIFFERENT_SIGNS

[Brian Paul]

On 09/24/2012 12:43 PM, Matt Turner wrote:
> On Mon, Sep 24, 2012 at 11:02 AM, Brian Paul<brianp at vmware.com>  wrote:
>> On 09/24/2012 10:49 AM, Matt Turner wrote:
>>>
>>> Module: Mesa
>>> Branch: master
>>> Commit: 0f3ba405eada72e1ab4371948315b28608903927
>>> URL:
>>> http://cgit.freedesktop.org/mesa/mesa/commit/?id=0f3ba405eada72e1ab4371948315b28608903927
>>>
>>> Author: Matt Turner<mattst88 at gmail.com>
>>> Date:   Fri Sep 14 16:04:40 2012 -0700
>>>
>>> Use signbit() in IS_NEGATIVE and DIFFERENT_SIGNS
>>>
>>> signbit() appears to be available everywhere (even MSVC according to
>>> MSDN), so let's use it instead of open-coding some messy and confusing
>>> bit twiddling macros.
>>>
>>> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=54805
>>> Reviewed-by: Paul Berry<stereotype441 at gmail.com>
>>> Suggested-by: Ian Romanick<ian.d.romanick at intel.com>
>>>
>>> ---
>>>
>>>    configure.ac           |    7 +++++++
>>>    src/mesa/main/macros.h |   21 ++-------------------
>>>    2 files changed, 9 insertions(+), 19 deletions(-)
>>>
>>> diff --git a/configure.ac b/configure.ac
>>> index 4193496..cb65467 100644
>>> --- a/configure.ac
>>> +++ b/configure.ac
>>> @@ -499,6 +499,13 @@ AC_SUBST([DLOPEN_LIBS])
>>>    dnl See if posix_memalign is available
>>>    AC_CHECK_FUNC([posix_memalign], [DEFINES="$DEFINES
>>> -DHAVE_POSIX_MEMALIGN"])
>>>
>>> +dnl signbit() is a macro in glibc's math.h, so AC_CHECK_FUNC fails. To
>>> handle
>>> +dnl this, use AC_CHECK_DECLS and fallback to AC_CHECK_FUNC in case it
>>> fails.
>>> +AC_CHECK_DECLS([signbit],[],
>>> +               AC_CHECK_FUNC([signbit],[],
>>> +                             AC_MSG_ERROR([could not find signbit()])),
>>> +               [#include<math.h>])
>>> +
>>>    dnl SELinux awareness.
>>>    AC_ARG_ENABLE([selinux],
>>>        [AS_HELP_STRING([--enable-selinux],
>>> diff --git a/src/mesa/main/macros.h b/src/mesa/main/macros.h
>>> index 04d59d7..7b7fd1b 100644
>>> --- a/src/mesa/main/macros.h
>>> +++ b/src/mesa/main/macros.h
>>> @@ -693,31 +693,14 @@ NORMALIZE_3FV(GLfloat v[3])
>>>    static inline GLboolean
>>>    IS_NEGATIVE(float x)
>>>    {
>>> -#if defined(USE_IEEE)
>>> -   fi_type fi;
>>> -   fi.f = x;
>>> -   return fi.i<   0;
>>> -#else
>>> -   return x<   0.0F;
>>> -#endif
>>> +   return signbit(x) != 0;
>>>    }
>>>
>>> -
>>>    /** Test two floats have opposite signs */
>>>    static inline GLboolean
>>>    DIFFERENT_SIGNS(GLfloat x, GLfloat y)
>>>    {
>>> -#if defined(USE_IEEE)
>>> -   fi_type xfi, yfi;
>>> -   xfi.f = x;
>>> -   yfi.f = y;
>>> -   return !!((xfi.i ^ yfi.i)&   (1u<<   31));
>>> -#else
>>> -   /* Could just use (x*y<0) except for the flatshading requirements.
>>> -    * Maybe there's a better way?
>>> -    */
>>> -   return ((x) * (y)<= 0.0F&&   (x) - (y) != 0.0F);
>>> -#endif
>>> +   return signbit(x) != signbit(y);
>>>    }
>>>
>>>
>>
>> Looks like we don't have signbit() on Windows.  We build with scons there so
>> the autoconf check isn't applicable.  I'll post a patch in a bit.
>>
>> -Brian
>
> MSDN claims that Windows does have signbit():
> http://msdn.microsoft.com/en-us/library/hh308342.aspx

I saw that too.  The page mentions "amp_math.h" but I don't have that 
header (which is supposed to be part of their Accelerated Massive 
Parallelism package).  Anyway, I also grepped all the MSVC header 
files for signbit and didn't find it.

-Brian