Re: Just got a Pi1B. What can you actually do with it these days?

In article <jKm*77LeA@news.chiark.greenend.org.uk>,
Theo  <theom+news@chiark.greenend.org.uk> wrote:

Gordon Henderson <gordon+usenet@drogon.net> wrote:

In article <kKm*zLLeA@news.chiark.greenend.org.uk>,
Theo  <theom+news@chiark.greenend.org.uk> wrote:

That was aimed at the OP who was asking the question, not you :-)

Hm. Maybe my threading got mixed up. No matter though.

If you've written your own OS I'm sure it works exactly how you want so no
reason to use something else.

Other than the sheer effort of getting "stuff" going where there is more
lines of verilog (or whatever) in the hardware than the software driver
needed to actually use it - e.g. USB... (or vice-versa, but essentially
things are now too complex to be fully understood by one person now)

for an 0.5MB 8MHz ARM2, so a 512MB 700MHz Pi1 is ample.  It can't use
multiple cores so the Pi1 being single core is no problem.

 
I have a scheme for multiple cores in my own system, (it's capable of
pre-emptive multi-tasking) but not got the energy to implement it over
multiple cores for now.

>
As long as you don't need synchronisation, there's not a lot to running
multiple cores - you just set them going with different PCs.  It's when they
need to talk to each other, or you need them to work on the same data, then
things get more complicated.

It's interestingly complex in my scenario in that there is a virtual
machine written in the native assembly language ('816, RISC-V or ARM32)
- it's like a sort of microcode and that's the level the multi-tasking
happens. Think Inmos Transputer if you know them... The neat thing is
that when I ported it from the '816 onwards it was/is binary compatible
from the output of the compilers point of view.

The one "big lock" needed right now is the memory allocator - it's a flat
memory system with no MMU so all calls to it's equivalent of malloc()
need locking, but that's OK for now and I've been looking at other models
like CSP but time is limited and there's only so much I can do for what's
really a little hobby.

(in the case of RISC OS, a whole load of legacy APIs are not thread safe.
In a custom OS you control everything so you can avoid such problems)
>

This is it in it's original incarnation on a 16Mhz 65C816 CPU demonstrating
multi-tasking:
 
https://www.youtube.com/watch?v=ZL1VI8ezgYc
 
The graphics is via a 115K serial link to a "smart" terminal running on
a Linux desktop. It uses Acorn VDU style commands for graphics, etc.

>
Looks neat.  I see a certain Acorn heritage there too :)

Well - that was the evolution. I started with a 65C02 system and made
it vaguely Apple II compatible, as that's the first 6502 system I used,
but then I wanted a better Basic, so threw that away and wrote an
Acorn MOS compatible OS for it - compatible to the point it will run
some Acornsoft ROMs unchanged including BBC Basic. The MOS has calls
you might expect like *exec, *spool, *. and so on... as well as enough
osbyte, osword, oscli, etc. to make BBC Basic happy.

(does that serial protocol have any similarity to the Tube, by any chance?)

I don't know as I've never done a deep dive into the Tube - but it's good
enough for BBC Basic to run with most of the plot commands working as
they should - although the output resolution isn't scaled to 1280x1024
like the Beebs is but it's whatever the native resolution of the display
happens to be.

So if you type

  PLOT 4,0,0 : PLOT 4,200,200 : PLOT 85,200,0

in BBC Basic running on my '02 or '816 board with my smart serial terminal
it will draw a triangle.

If theory, if you used a temrinal program on a real Beeb and just had it echo
what it got from the serial port to the screen it should "just work"... I've not
tried this yet.

Over recent years I've ported it from the 816 to RISC-V and now to ARM32.
 
The Pi version has native graphics and works well on the Pi 7" screen
thing (or external HDMI). It boots to Basic or BCPL in under 2 seconds.

>
That's the nice thing about not having to boot Linux or systemd...

But I may have to boot a Linux for my campervan project as I'd really like
to use a USB serial line to talk back to the MCUs that form the sensor network.
Running the VM under Linux should be easy as the upper layers (in BCPL) assume
a Posix style interface to the underlying whatever. What's another 20 million
lines of code in a little hobby project anyway...

-Gordon
(Probably the 2nd last person on the planet to code in BCPL)