Re: Several Malta patches developed using QEMU

["Maciej W. Rozycki" <macro%orcam.me.uk@localhost>]

On Wed, 1 Apr 2026, Simon Burge wrote:

> > I'm hoping I didn't break the actual physical hardware with my changes
> > to make QEMU work - but I guess the world of "the physical hardware" for
> > Malta is quite a bit wider than what NetBSD ever supported.

 The Galileo system controller was only used with Atlas/Malta core cards 
for a couple of older LV cores, nominally the 4K/4KE and 5K families, and 
the QED RM5261 CPU.

 There were a couple of card models using the Bonito system controller IP 
from Algorithmics; these were equipped with either the QED RM7061 CPU 
(very rare, no specimen may exit anymore) or the 20Kc LV, and the system 
controller held in an FPGA (the same system controller IP was used with 
numerous own development boards made by Algorithmics).

 All the FPGA (softcore) as well as later LV core cards, such as ones with 
the 24Kc, 74Kc or 74Kf LVs, and in particular all the multiprocessor soft 
CPU cores, such as the interAptiv core, used the various revisions of MTI 
SOCit/ROCit system controller IP.  Later core cards may not work with the 
Atlas board, but all the Galileo and Bonito ones ought to, as should the 
CoreFPGA2 card; I've tried this stuff myself, back in the day.

 I realise that QEMU likely only emulates a very narrow subset of these 
features, and then surely it's not authoritative and may get things wrong 
or omitted, so I think running verification with actual hardware is key.  
I've been bitten with QEMU bugs before.

> > Thank you for offering to help, though!  If I can't find someone with
> > the other core card to run some tests, I guess I'll just commit and
> > hope for the best.
> 
> I will dig though my small collection of Maltas on the weekend and see
> which ones still work.

 Let me know if I could be of any assistance (or if you find any subset of 
your collection redundant and would be happy to part with ;) ).

 With my 5Kc card likely gone for good I'm after a 64-bit core card in 
particular, as it'll bring me back the ability to run proper 64-bit MIPS 
verification in my lab (64-bit DECstation hardware, which I also have, 
sadly suffers from serious processor errata, which limit its ability for 
64-bit operation; it is also severely as opposed to just slightly 
obsolete).

  Maciej