Could a 68881 Math co-processor go fast?; HwNote07

[Tom Tkacik]

In article <5087 at cbmvax.UUCP> ditto at cbmvax.UUCP (Michael "Ford" Ditto) writes:
>I'd just put in a 68020 at the same time; then you get the coprocessor
>interface and a few other benefits.

Sounds like a good idea, but how do we do it?
>
>Computer System Associates in San Diego makes "daughter board" 68020/881
>upgrades that plug into a 68010 socket.  They come in several different
>shapes, so maybe one would fit in the Unix PC.  There might be some minor
>OS problems.  (And, of course, minor OS problems become major ones without
>source code, but I think it would be doable.)
>-- 
>					-=] Ford [=-
>"The number of Unix installations	(In Real Life:  Mike Ditto)

I think the hardware problem is trivial, as you mentioned, there are
companies who make the needed daughter board.

Unfortunately, without source code, I do not think the problems with the OS
are minor, they are major.  The size of the saved state of a 68020 and 68881
is much larger than that for the 68010.  Each time there is a context switch
the OS thinks it is putting away info for the 68010.  We might be able to
find out where in the OS this is done, and make a slight change to tell it
to store away the extra information, but where would it go.  I do not think
we would be able to increase the structure size used to save this data.

Then again, as Lenny T. pointed out, the compiler has the switches for
generating code for the 68020-68881 combination.  Maybe this was to be
a possible upgrade, and was considered when the OS was designed?  Perhaps
the hooks are all there, only waiting to be found.  Any ideas about where
to start looking?

---
Tom Tkacik
GM Research Labs
Warren MI
{umix, uunet!edsews}!rphroy!megatron!tkacik
{umix, uunet!edsews}!rphroy!tetnix!tet