On Wed, Aug 16, 2000 at 10:57:23AM +0200, Matthias Scheler wrote:
> > I haven't look at it in any detail, but if it's broken in the same way as
> > the OSS library, I suggest you go ahead and fix it in the same way.
> 
> That's not so easy. The fix is to use floating point values for the
> intermediary result. And I can't do that of course inside the kernel.

How about using 32-bit integers with 16 bits as integer part and 16 bits
of fraction?  Multiply (shift left) the incoming byte (or 16-bit word,
if necessary) value and store the result.. and convert back when necessary.

  -jm