[FFmpeg-devel] ffmpeg and argument parsing.

Uoti Urpala uoti.urpala
Tue Sep 4 23:59:17 CEST 2007


On Tue, 2007-09-04 at 17:25 -0400, Marc Hoffman wrote:
> I guess our fast float implementation doesn't handle NAN==NAN so this
> fails on our current platform with -mfast-fp -ffast-math.  We do have
> an IEEE library which it does work with but that just requires more
> overall mips.

In GCC -ffast-math explicitly assumes that floating-point arguments and
results are not NaNs or +-Infs, so the d==d check is incompatible with
that option.





More information about the ffmpeg-devel mailing list