> From: Paul Mather <paul@gromit.dlib.vt.edu>
>
> Recently, I reported problems using fpsetmask() on NetBSD/alpha 1.4L,
> in that it didn't appear to work.  Well, to cut a long story short, I
> did a bit more investigating, and am now wondering if the reason it
> doesn't work is that it is "not implemented yet."
>
> Can someone who knows please let me know if fpsetmask is actually
> implemented under NetBSD/alpha, and, if not, if there is any way to
> use IEEE 754 mathematics semantics on that port?
>
> (As far as I can gather, fpsetmask is implemented as an alpha system
> call, but that the system call itself is not implemented:

Very good analysis. I fixed up the library stubs and sysarch stub a while
back so that programs compiled on 1.4 and later would eventually start
doing the right thing when the kernel FP code was integrated.

The main issue is high quality emulation; this is required on alpha
for some IEEE fine points. Also, the alpha doesn't completely implement
the result state bits; these require more kernel magick.

Now, we have some high quality FP emulation SW from UCB that we want to
use, but I haven't integrated it yet.

I'll try to get to it sometime before the next millenium... :-)

        ross