Subject: fpu emu progress report (not much yet) and some request of help
To: None <macbsd-development@sun-lamp.cs.berkeley.edu>
From: Ken Nakata <kxn3796@hertz.njit.edu>
List: macbsd-development
Date: 09/09/1994 10:13:40
I will soon be writing actuall code, thanks to Chris Demetriou, Chris
Torek, Brad Grantham, and other people who have replied to my
questions.

Consequently, I will also soon be in need of

	1) a source tree which compiles and builds a kernel at least
	as stable as BETA-1 kernel, and
	2) some way to tell gas and/or gcc the fpu coprocessor id.

Of course, I need (1) in order to build kernels with my fpu emulator
(or fpue for short... do you like this name?) code.  Does -current
tree compile?

And with (2), I can generate f-line traps on my SE/30 which has an fpu
so that I can test fpue without an fpu-less machine.  I thought of
assigning fpue to a-line trap and patching gas so that it would
generate a-line instructions instead of f-line (true fp)
instructions.  But I guess changing the coprocessor id should be much
easier than that.

By the way, I've thrown away my previous idea of calling fpsp, because
emulating 68040 fpu UNIMP stack frame seems non-trivial.  And I've
come to realization that evaluation of transcendental functions was
not as bad as I thought before, as my study has progressed.

I appreciate any help or comments including but not limited to 8-)
constructive criticism.

Ken

------------------------------------------------------------------------------