If this is a 486DX, there should be a floating point unit built-in.
In your original dmesg, I didn't see an npx device, but I should have.  is
it missing from your configuration?  I ran NetBSD successfully on a 486DX
for years, and it did floating point just fine.  Is there something special
about your 486?
-thanks
-Brian
On Feb 25,  2:56pm, Pierre Dubuc wrote:
} Subject: Re: Status of math_emulate
} -----BEGIN PGP SIGNED MESSAGE-----
} Hash: SHA1
} 
} Replying to message <Pine.NEB.4.64.0702250743360.4723@localhost.localhost>
} 
} >
} > As an experiment, I'll build a kernel without it and give it a try.
} >
} 
} No go. A bunch of daemons failed saying stuff about "no floating" 
} available or something. On my 486, the MATH_EMULATE is absolutely 
} mandatory for running even the basic stuff.
} 
} Google told me 0xd9fa is the square root function. Looking in 
} src/sys/arch/i386/i386/math_emulate.c, I do find a mention of this 
} operation code, along with a bunch of others, leading to the "not 
} implemented" message. Running a diff between the version in NetBSD-3.1 
} (1.28) and the most recent CVS version (1.32) shows a bunch of changes, 
} but nothing that would improve on my problem.
} 
} So I guess I'm on my own. I'm far from being a kernel developer, so 
} I'm still open to other options, but upgrading the hardware is not one
} of them ;-)
} 
} - -- 
} Pierre Dubuc
} pldubuc@yahoo.ca
} -----BEGIN PGP SIGNATURE-----
} Version: GnuPG v1.4.6 (NetBSD)
} 
} iEYEARECAAYFAkXh6fkACgkQaMA1iJ2qYmIRJQCfTlMJeneB9cKEDthLDbaEI3si
} PUkAnA8nqL0VlzFeBZxvjb/1LfHHvcj2
} =B/mz
} -----END PGP SIGNATURE-----
} 
>-- End of excerpt from Pierre Dubuc