> It was my understanding from reading the Sun 3 FAQ archive from Germany
> that the 68882 is pin for pin compatible with the 68881, and about 50%
> faster in some fractions of its operations, at the same clock rate (the
> pages on hacking a 3/60 recommended replacing the '81 FPU with an '82). How
> does the Sun FPA compare against the 68882?

This all sounds absolutely correct.  The Motorola databook sez that
the 68882 is pin compatible with the 68881 and "provides in excess of
1.5 times the performance of the 68881".  I have several 68882 chips, and
could give this a try sometime soon and report on the results.  I had
some concerns that there might be some kernel level software details
that were different (like the contents of what Frestore and Fsave deal
with), but ultimately this may not matter if it is handled like a magic
cookie.

As far as the FPA versus the 68881/2 they are utterly different in
instruction set (not to mention implementation).  I was running under
SunOS yesterday playing with the FPA, I will try to do some more of this
and maybe run some simple benchmarks (this will be more instructive if
I actually manage to get my FPA to *work-at-all* !!!!! )

Based on what I have read so far, I would expect:

	68881	0.2 Mflops
	68882	0.3 Mflops
	FPA	1.0 Mflops

Benchmarking is a great sport, I'll let you know when the scores are in.

	Tom Trebisky			Steward Observatory
	ttrebisky@as.arizona.edu	University of Arizona
	http://kofa.as.arizona.edu/	Tucson, Arizona 85721
					(520) 621-5135