Chuck McManis wrote:

> You are exactly right, some VAX chips do not have all of the floating point
> instructions implemented, and in particular POLYD. The SOC has a different
> set implemented than NVAX and so you're running into problems.

Yeah.  I found Bob Supnik's chart of processor features at
http://simh.trailing-edge.com/docs/vax_proc.txt and found it quite handy as a
reference for what processor can do what.  Could/should something like this be
somewhere on the NetBSD pages?
 
> There is a #define in the kernel config file that talks about emulating
> floating point. Have you tried building a kernel for the 4000/90 with this
> define switched on (or off I can't remember the default off hand)

GENERIC kernels do not have NO_INSN_EMULATE defined.  I turned it on in the
4k/90 kernel out of optimism - "The NVAX is fast, therefore it must have all the
cool features!"  Obviously, according to the chart I mentioned above, it does
not have decimal floating-point, nor does it have a hardware emulated mode like
the SOC.  I question DEC's logic in this design, but I suppose it's far too late
to do anything except gripe pointlessly.

I'm recompiling the kernel for the 4k/90 now with NO_INSN_EMULATE turned off, so
hopefully this will solve the issue.

-- 
Henry Bent
hbent@cs.oberlin.edu