[MPlayer-dev-eng] float vs. double

[Sidik Isani]

On Sun, Sep 22, 2002 at 01:32:31PM +0200, Arpi wrote:
> Hi,
> 
> >   It seems like rounding errors in some of the time sync calculations
> >   accumulate after a while.  They might be less if "double" was used.
> >   And aren't most of our processors as fast at dealing with a double?
> >   I guess it doesn't matter since pts_correction eventually picks up
> >   the slack, but it uses float too, so it might get less and less
> >   precise toward the end of a long movie.  Just curious if there's a
> >   reason a float is better?
> 
> i didn't notice such behaviour...

  But I think it is happening:  Maybe it is why the sh_audio->timer
  even needs to be corrected by the "pts_correction" code in the
  first place.  I have not tried changing all the timestamp variables
  to double's to see if it changes anything.  I was only wondering.
  It works fine for me too.

Thanks,

- Sidik