pb Gnat 3.12 WinNT with float_io.put and OpenGL

pb Gnat 3.12 WinNT with float_io.put and OpenGL

[Nicolas Brunot]

With Gnat 3.12 compiler under WinNT 4, we have a problem with
float_io.put for
a floating type digit 15
(problem with ATI graphic card, not with others graphic cards)

the result of float_io.put is correct until the application creates an
OpenGL context.
After that, every call to float_io.put returns absurd results if EXP=0
and FORE/=DEFAULT_FORE
Did somebody already encounter that problem or have an explanation for
the effect
of calling OpenGL (with ATI driver, not others graphic cards) on the
gnat executable ?

Thanks

Nicolas BRUNOT
n.brunot@cadwin.com

Re: pb Gnat 3.12 WinNT with float_io.put and OpenGL

[John Merryweather Cooper]

I suspect that the OpenGL libraries you are using do not segregate the MMX
and Floating Pointing Processor states.  When OpenGL uses MMX as part of its
acceleration, it is corrupting the Floating Point state.  YMMV.

jmc
======================
"Nicolas Brunot" <n.brunot@cadwin.com> wrote in message
news:38C516A8.A6EBC01D@cadwin.com...
> With Gnat 3.12 compiler under WinNT 4, we have a problem with
> float_io.put for
> a floating type digit 15
> (problem with ATI graphic card, not with others graphic cards)
>
> the result of float_io.put is correct until the application creates an
> OpenGL context.
> After that, every call to float_io.put returns absurd results if EXP=0
> and FORE/=DEFAULT_FORE
> Did somebody already encounter that problem or have an explanation for
> the effect
> of calling OpenGL (with ATI driver, not others graphic cards) on the
> gnat executable ?
>
> Thanks
>
> Nicolas BRUNOT
> n.brunot@cadwin.com
>

Re: pb Gnat 3.12 WinNT with float_io.put and OpenGL

[Nick Roberts]

Ah, but never mind the bugs, look just how FAST it goes!

:-)

--
Nick Roberts
http://www.adapower.com/lab/adaos