1. Forum
  2. Usenet
  3. COMP.STD.C

  • short float

    From Dave Brower@21:1/5 to Philipp Klaus Krause on Wed Dec 9 11:21:29 2020
    On Monday, October 28, 2019 at 3:58:08 AM UTC-7, Philipp Klaus Krause wrote:
    While the idea of a short float was viewed favorably by the comitee a
    few years ago, in London, the authors never followed up with a detailed proposal.

    I'd like to see short float in C, not just because a lot of hardware
    supports it, but also as a relatively efficient floating-point type for devices that don't have hardware support for floating-point at all:

    http://www.colecovision.eu/stuff/proposal-short-float.html

    Just stumbled on this. If there are OpenGL unsigned floats, why not have 'unsigned short float' as a thing, with 'short float' as a signed type?

    (Yes, I know it's a year later).

    -dB

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Keith Thompson@21:1/5 to Dave Brower on Wed Dec 9 13:09:51 2020
    Dave Brower <[email protected]> writes:
    On Monday, October 28, 2019 at 3:58:08 AM UTC-7, Philipp Klaus Krause wrote:
    While the idea of a short float was viewed favorably by the comitee a
    few years ago, in London, the authors never followed up with a detailed
    proposal.

    I'd like to see short float in C, not just because a lot of hardware
    supports it, but also as a relatively efficient floating-point type for
    devices that don't have hardware support for floating-point at all:

    http://www.colecovision.eu/stuff/proposal-short-float.html

    Just stumbled on this. If there are OpenGL unsigned floats, why not
    have 'unsigned short float' as a thing, with 'short float' as a signed
    type?

    (Yes, I know it's a year later).

    This:
    https://www.khronos.org/opengl/wiki/Data_Type_(GLSL)
    does not indicate that OpenGL has unsigned floats.

    I found a reference to 10- and 11-bit unsigned floating-point types, but
    I get the impression that they're not language-defined types.
    https://www.oreilly.com/library/view/opengl-programming-guide/9780132748445/app07lev1sec3.html
    C would not be able to support a type whose size is not a multiple of
    CHAR_BIT (bit fields notwithstanding).

    --
    Keith Thompson (The_Other_Keith) [email protected]
    Working, but not speaking, for Philips Healthcare
    void Void(void) { Void(); } /* The recursive call of the void */

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Philipp Klaus Krause@21:1/5 to All on Sat Dec 12 14:49:47 2020
    The proposal is dead anyway. Many, including Nvidia, now see float,
    double and long double as mistakes that can't be undone, but shouldn't
    be repeated either.
    They'd prefer all floating-point formats to be defined precisely down to
    the last bit in representation and semantics.

    Philipp

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From jacob navia@21:1/5 to All on Sun Dec 13 13:42:34 2020
    Le 12/12/2020 à 14:49, Philipp Klaus Krause a écrit :
    The proposal is dead anyway. Many, including Nvidia, now see float,
    double and long double as mistakes that can't be undone, but shouldn't
    be repeated either.
    They'd prefer all floating-point formats to be defined precisely down to
    the last bit in representation and semantics.

    Philipp


    IEEE754 is dead also?

    Can you point me to some written text by a comptetent authority that
    says what you are saying?

    Thanks

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)