Re: SMP 68k (Re: New pmap module for m68k)

[John Paul Adrian Glaubitz <glaubitz%physik.fu-berlin.de@localhost>]

Hi Romain,

On Mon, 2026-10-19 at 17:07 +0200, Romain Dolbeau wrote:
> Le jeu. 13 nov. 2025 à 10:05, Romain Dolbeau <romain%dolbeau.org@localhost> a écrit :
> > I'd love to see this happen, but building hardware for it is a bit of
> > a pain. Having emulation working as a starting point would be great
> > (for buildhost and test SW before creating HW).
> 
> Well, barely a year later and the world has changed a lot for this
> kind of stuff - whether we like it or not- because of LLMs (I'm not
> using the i-word to describe it).
> 
> ```
> NetBSD/virt68k (virt68k) (constty)
> 
> login: root
> Oct 19 14:54:44 virt68k login: ROOT LOGIN (root) on tty constty
> Last login: Mon Oct 19 11:52:23 2026 on constty
> Copyright (c) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003,
>     2004, 2005, 2006, 2007, 2008, 2009, 2010, 2011, 2012, 2013,
>     2014, 2015, 2016, 2017, 2018, 2019, 2020, 2021, 2022, 2023,
>     2024, 2025, 2026
>     The NetBSD Foundation, Inc.  All rights reserved.
> Copyright (c) 1982, 1986, 1989, 1991, 1993
>     The Regents of the University of California.  All rights reserved.
> 
> NetBSD 11.0 (GENERIC.MP) #15: Sun Oct 18 12:50:46 CEST 2026
> 
> Welcome to NetBWe recommend that you create a non-root account and use
> su(1) for root access.
> virt68k# sysctl hw.ncpu
> hw.ncpu = 8
> virt68k# sysctl hw.model
> hw.model = Qemu 11.1.2: MC68040+MMU+FPU, 4k+4k on-chip physical I/D caches
> virt68k#
> ```
> 
> Took claude code a day to get a first version (of both qemu & netbsd),
> and about another one to "optimize" the pmap code, moving away from
> one big lock and making "make -j" relatively efficient to self-host a
> full kernel rebuild. Whether that would translate to actual hardware
> is a wide open question. But now that software is much less of an
> issue, I need to get around to actually building my SMP MC680x0
> homegrown... (which is basically two CPUs and a big FPGA to handle
> everything else)
> 
> I don't think AI-generated code is welcome in NetBSD, so I'll stop
> here, but if anyone wants to discuss this off-list they're welcome.

I think the QEMU guys (unfortunately) currently don't accept any AI
contributions, but at least the Linux kernel will gladly accept such
contributions if you were to provide Linux support.

I would be interested in the QEMU patches either way.

Adrian

-- 
 .''`.  John Paul Adrian Glaubitz
: :' :  Debian Developer
`. `'   Physicist
  `-    GPG: 62FF 8A75 84E0 2956 9546  0006 7426 3B37 F5B5 F913