Have you seen Fuzix? It's essentially exactly this: a V7-ish Unix ported to
run on very, very small machines. It'll run happily on a variety of Z80,
6809, 68000, 8086 architectures etc. On a Z80 with a basic MMU, like my
NC200 laptop, you get about 55kB of userspace per process, which is enough
to run some compilers.

https://github.com/EtchedPixels/FUZIX

In single-tasking mode it'll run on really absurdly small devices: I had it
running, happily, on an MSP430 development board with 66kB of RAM, with a
25kB userspace (after severe hacking; it's since bitrotted).
http://cowlark.com/2015-10-27-fuzix/

It makes ELKS look big.

On Tue, 31 Jul 2018 at 19:40 Paul Osmialowski <[email protected]> wrote:

> > Note this message came out of subj: ELKS Git repository up on
> SourceForge + patches applied
>
> > Hello All,
>
> > JB>Additionally, I wish to continue discussing options for making ELKS
> less 8086-specific, and moving to a compiler that can make ELKS more
> easily portable.
>
> > The last discussion about ELKS possible way of porting/evolution was
> going on here many years ago. As I remember, I've made short conclusion
> for myself out of all those quite intensive message "ping pong" and
> brainstorming. The conclusion was similar to: "The 8088/86/286 memory
> access models make programming for those processors very similar to many
> 8-bit chips and PDP-11. The 64kb limit is very strong and serious barrier
> that is not possible to overcome without serious performance penalties
> (e.g. FORTH or an virtual 32bit RISC CPU). The whole industry already
> found the way out by switching to a new hardware with 24/32/64bit address
> space, so most sources around the world is not possible to compile using
> 64kb code/data segments. If you ever run any virtual machine on 88/86/286
> CPU with direct memory access to few MB of RAM then you would run ucLinux
> instead of ELKS. So, the only possible way for the ELKS evolution is to
> support EXE (i.e. multi segments executable) and probably support other
> systems with 16bit address space limit (including those 8-bit and
> PDP-11)."
>
> > Thanks,
>
> > Andrey
>
> ELKS on 8-bit machines? How cool idea is that?! This reminds me a good
> hardware candidate for it, Zilog's eZ80 microcontroller with 24-bit
> address space, e.g. eZ80F920120MOD board has 512kB extenal RAM which
> should be enough for ELKS. Some time ago I managed to boot Nuttx on it and
> telnet into it over TCP/IP (and Nuttx can run on various non-MMU
> architectures too starting from 8-bit Z80 through 8086/8088 machines as
> well as 32-bit ARMv7m MCUs e.g. Cortex-M4). In theory, old Z80-based MSX
> machines were also well stuffed for their times (enough for ELKS), and
> non-free modern-era OS'es like SYMBOS can prove good old Z80 is capable to
> run multi-tasking OS (well, actually Nuttx also proved that).
> --
> To unsubscribe from this list: send the line "unsubscribe linux-8086" in
> the body of a message to [email protected]
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

Reply via email to