[Ironpython-users] IronPython float_info values (was: IronPython always in 32-bit mode on 64bit machine even when running ipy64.exe)

[Jeff Hardy]

For me, using BitConverter.IsLittleEndian is clearer:

  private static readonly double EPSILON =
System.BitConverter.Int64BitsToDouble(
    System.BitConverter.IsLittleEndian ? 0x3cb0000000000000 :
0x000000000000b03c);

  private static readonly double MIN = System.BitConverter.Int64BitsToDouble(
    System.BitConverter.IsLittleEndian ? 0x0010000000000000 :
0x0000000000001000);

I don't think half-endian machines are worth worrying about. :)

- Jeff


On Fri, May 10, 2013 at 1:41 AM, Markus Schaber <m.schaber at codesys.com> wrote:
> Hi,
>
>
>
> Maybe we'd use some conditional initializiation, checking for the properties
> of the number, like the following:
>
>
>
>         // The definition of Epsion and MinValue differ between .NET and
> Python/C,
>
>         // so we need to define our own values.
>
>         // Approx. -2.2204460492503131e-16 - the byte order reversed value
> is approx.
>
>         // 2.2290265677773679e-319, so this code delivers the correct value
> independently
>
>         // of the byte order of the machine.
>
>         private static readonly double EPSILON = Math.Max(
>
>             System.BitConverter.Int64BitsToDouble(0x3cb0000000000000),
>
>             System.BitConverter.Int64BitsToDouble(0x000000000000b03c));
>
>         // Approx. 2.2250738585072014e-308 - the byte-order reversed value
> is approx.
>
>         // 2.0236928853657458e-320, so this code delviers the correct value
> independently
>
>         // of the byte order of the machine.
>
>         private static readonly double MIN = Math.Max(
>
>             System.BitConverter.Int64BitsToDouble(0x0010000000000000),
>
>             System.BitConverter.Int64BitsToDouble(0x0000000000001000));
>
>
>
>         public static floatinfo float_info = new floatinfo(Double.MaxValue,
> 1024, 308, MIN, -1021, -307, 15, 53, EPSILON, 2, 1);
>
>
>
> This trick relies on the fact that for the given values, the reversed byte
> order produces an zero-Exponent, which are denormalized numbers, and in both
> cases, the wanted number is a normalized one which (by definition) is larger
> than the denormalized one.
>
>
>
> This code is as elegant as checking for the endianness explicitly, and we
> are even safe for machines where the integer and float endianness are
> different (see
> https://en.wikipedia.org/wiki/Endianness#Floating-point_and_endianness –
> although I doubt that those hardware is relevant for our user base…) J
>
>
>
>
>
>
>
> Best regards
>
> Markus Schaber
>
> CODESYS® a trademark of 3S-Smart Software Solutions GmbH
>
> Inspiring Automation Solutions
>
> ________________________________
>
> 3S-Smart Software Solutions GmbH
> Dipl.-Inf. Markus Schaber | Product Development Core Technology
> Memminger Str. 151 | 87439 Kempten | Germany
> Tel. +49-831-54031-979 | Fax +49-831-54031-50
>
> E-Mail: m.schaber at codesys.com | Web: codesys.com | CODESYS store:
> store.codesys.com
> CODESYS forum: forum.codesys.com
>
> Managing Directors: Dipl.Inf. Dieter Hess, Dipl.Inf. Manfred Werner | Trade
> register: Kempten HRB 6186 | Tax ID No.: DE 167014915
>
> Von: Slide [mailto:slide.o.mix at gmail.com]
> Gesendet: Mittwoch, 8. Mai 2013 22:44
> An: Jeff Hardy
> Cc: Markus Schaber; Curt Hagenlocher; ironpython-users at python.org
>
>
> Betreff: Re: [Ironpython-users] IronPython float_info values (was:
> IronPython always in 32-bit mode on 64bit machine even when running
> ipy64.exe)
>
>
>
> Unless Microsoft really wants pain and anguish, they are likely running in
> little endian mode. I've never run across an ARM setup that actually runs in
> big endian (nor do I want to), so I'm pretty sure its little endian. I'd be
> more than willing to test an app on my phone since Windows Phone is crap
> anyway, so you couldn't damage it any worse than it is now :-)
>
>
>
> Alex
>
>
>
> On Wed, May 8, 2013 at 12:54 PM, Jeff Hardy <jdhardy at gmail.com> wrote:
>
> On Wed, May 8, 2013 at 9:15 AM, Markus Schaber <m.schaber at codesys.com>
> wrote:
>> Hi,
>>
>>
>>
>
>> The main problem for now is that I can't test whatever solution we find,
>> because I don't have access to any hardware with non-intel Byte Order and
>> a
>> Microsoft .NET. (I don't want to rely on Mono for this test…)
>
> Is ARM big-endian? If so, any Windows 8 ARM tablet (i.e. Surface RT)
> or Windows Phone should do. Heck, even the emulator might do. It
> shouldn't be too hard to whip up an app that prints those two values,
> and I think Alex has a Windows Phone if he'd be willing to test it.
>
> If ARM is little-endian (or Windows/ARM runs in little-endian mode)
> then I don't think it matters; we'd have to try Mono on a PowerPC or
> something like that.
>
> - Jeff
>
>
>
>
>
> --
> Website: http://earl-of-code.com
>
>
> _______________________________________________
> Ironpython-users mailing list
> Ironpython-users at python.org
> http://mail.python.org/mailman/listinfo/ironpython-users
>