On Nov 18,  9:54am, "Erik E. Fair" (Time Keeper) wrote:
> Subject: Re: A few general interest questions...
>
> [ plain text
>   Encoded with "quoted-printable" ] :
At 7:30 -0800 11/18/96, Gordon W. Ross wrote:
> >
> >There is almost exactly zero FP code in plain NetBSD.
>
> If that were really true, the folks working on the mac68k port wouldn't be
> tearing their hair out over the currently inadequate FPU emulation support.
> Might be nice to put together a list of programs that do use FP sometime,
> and perhaps even eliminiate some use of floating point when it's not
> absolutely necessary...

That would be very nice for two reasons. One is FPU emulation speed. The other
is the number of programs you can get up and running before you have FPU
emulation running in any new ports to FPU-less architectures. In many of the
statistics generating programs (df, *stat and friends) floating point is used
for calculating percentages. This could easily be done in a lot of cases using
fixed-point integer arithmetic. These programs are particularly useful when
your kernel is full of bugs and you want to know what conditions (such as onset
of swapping for instance) are causing it to panic.

Cheers

Rob Black

PS I have cross-posted this to port-arm32 because I think there might be quite
a bit of interest there.

--