Hi,

I'll second John's answer.

Basically it works fine regardless of mask level/what ever.
The softfloat build doesn't generate any F instructions
whihc means that traps aren't messed up.

My experience with a LC040 that was supposed to be good was
that only some traps failed, randomly, at odd times.  It was
a major annoyance.

The stock NetBSD with fpu emulator either works or doesn't.
If it works ls -l say will work over and over.  If it doesn't
ls -l probably won't work on a directory of more than a 
few files.

You'll also get cores booting as well.

I guess it's time for me to write some docs, right :-) ?

Apple's SANE is quite similar to our softfloat solution, ie,
you never generate F instructions rather

double d = 0;
d += 10;
would generate calls to SANE just like gcc in the NetBSD
softlfloat build calls addxxxx to do the above addition.

cheers

bruce


"E. Seth Miller" <esmiller@umich.edu> wrote:
> Date: Thu, 13 Nov 2003 15:35:54 -0500 (EST)
> From: "E. Seth Miller" <esmiller@umich.edu>
> Subject: Current level of LC040 support?
> To: port-mac68k@NetBSD.org
> 
> Greetings, all-
> This is probably of greatest relevance to Bruce ONeel, but I'd love to
> hear from anybody with input.  What's the current status of LC040 support
> for NetBSD/mac68k?  I've seen the -msoft-float build of 1.6.1, but haven't
> had the chance to try it out yet.  However, I have a few questions about
> it.
> 
> First, is this going to work with any LC040, or only the non-buggy (mask
> revision 2E71M and later) ones?  My impression is that, due to an errata
> on the early LC040s, -msoft-float is insufficient to allow machines
> equipped with those processors to run NetBSD.  Is this no longer the case?
> 
> Second, assuming that it's still the case, does anybody have information
> on what exactly fails?  Most importantly, is there a quick way to test if
> a system has a good or bad mask revision?  Last I knew, the easiest way to
> test was to run SoftwareFPU or just try NetBSD.  Is there any better
> method out there now?  Assuming there isn't, does anybody have some leads
> I could follow to figure out what exactly is failing and a way to easily
> test it?
> 
> Of course, if I've missed something, by all means, let me know.  I've
> looked through the list archives and didn't see an answer, but if it was
> there, feel free to send me on my way with a thwack upside the head.
> The documentation, though, seems decidedly out of date with the current
> state of things, so it's hard to know what's been done.  Hopefully,
> somebody can fill me in!
> 
> 	-Seth Miller