[LispM] Re: Outstanding Meroko bugs and a question of my own

[nyef at lisphacker.com]

Robert Swindells writes: 

> nyef wrote: 
> 
>>Probably this weekend, or maybe Friday. It's obvious to me at this point 
>>that I'm not going to get another test to pass just yet unless it's the 
>>I-Memory test, and that'll probably take some dumb luck. 
> 
> Whenever you are ready. 
> 
> I will play with usim first.

Whichever. I'll still try to put together a release. For backup purposes if 
nothing else. ^_- 

>>On the upside, it now passes all of EXPT save the I-Memory, Event Posting, 
>>and Memory Cycle Abort tests. On the downside, the 8th set of VMA tests are 
>>scary, and implied un-fun things about the MMU. I -really- wish the MMU were 
>>better documented, as the solution I hit upon is fairly shaky feeling. 
> 
> Is there anything you can learn from Hummingbird that would help ?

Probably not, seeing as Hummingbird has a completely different MMU design, a 
cache setup, and the scary microinstruction pipelineing stuff. 

>>> I am going to get back into lisp and lispm hacking full time from
>>> today onwards.
> 
>>Cool. You still have an mX board, right? I'm thinking that maybe after we 
>>get Raven emulated properly we go on to Hummingbird. 
> 
> I have still got the mx, I never got it to boot though. It was
> probably most valuable as a way for me to get a complete Hummingbird
> software tree.

Mmm... We're still missing the on-chip ROM, which apparently includes some 
basic self-tests. I do know what the load format for the board ROMs is, 
though, so hopefully that should suffice to get started. 

> I would like to see Hummingbird emulated as well. 
> 
> A longer term plan is to try and do a Hummingbird replica in VHDL. 
> 
> How good is the disassembler in Nevermore now ?

I haven't messed with it in ages, so I don't know. And it's not quite a 
priority, if you know what I mean. 

> One advantage that usim has now is that the microcode source is
> available.  This would make it easier to define some "back door"
> instructions that call down to the underlying operating system in a
> similar way to how guest operating systems under VMWare and MOL work.

Eh. The mX already has a setup for that. Micronet, remember? Or we could 
create a NuBus device for the host interface and ping it directly with 
%nubus-write and friends. 

> Robert Swindells

 --Alastair Bridgewater