SMS2 in FPGA

[Tinyfpga]

From: SOLVED !! smsq/e on rom
Post by Peter » Sat Jun 14, 2025 3:41 pm

Tinyfpga wrote: Sat Jun 14, 2025 2:13 pm
It runs on the Atari ST core, but I would love to see SMS2 running on your FPGA boards.

I did successfully try open source PC/AT and Z80 implementations on my boards, so Atari would also not be rocket science.
Just glad, that Markus warned about the mass storage issue. A good reason not to invest my m

Another thing is: My boards have no external CPU to easily swap FPGA cores. They were designed for QL-only use.
The newer ones allow FPGA field updates, but those happen under SMSQ/E with the update software running on the FPGA CPU.
I would not want to also write such software for Atari, so someone can go back to the QL-style core.

I felt that the above should form the basis for new topic.

My statement starting "but I..." was meant as a rehtorical statement, not as a kind of suggestion that Peter or anyone else might want to put SMS2 into his FPGA boards and certainly not in an Atari core.
The reason that SMS2 was written to run on an Atari is that it was ,at that time, the simplest,off-the-shelf, 68000 based computer available. An unmodified QL was just not msystem that would run instantly from an unmodified machine.

SMS2 and SMSQE are remarkably similar I would guess that anything that can run SMSQE natively would be capable of running SMS2 in the same core. SMS2 is currently smaller and probably simpler than SMSQE because it does not need to support an interpreter as its user interface.

The question one might well ask is, what would be the point?
My answer is as follows:- I still believe that one day, I will be able to find someone willing to finish the Stella project. No user interface was ever (as far as I know) finished for Stella. Finishing the Stella kernel would be relatively simple but creating a new user interface and IDE for it would be too much work in the first instance.
It would make no sense using the QL like user interface of SMSQE but the much simpler SMS2 interface might be suitable system for accessing Stella.

Stella, according to Tony Tebby, represents the zenith of his operating system designs.

As a matter of interest where do Qimi's/Qzero's/Q68's cores come from and how are they implemented? Do they have to be re-implemented after power loss?

[Peter]

My answer is as follows:- I still believe that one day, I will be able to find someone willing to finish the Stella project.

Is it open source?

Stella, according to Tony Tebby, represents the zenith of his operating system designs.

But shackled to the 68K CPU line. Which was already in decline when he still worked on Stella and is dead now. If Tony Tebby wrote it in C, totally different story.
In my humble opinion, the last little door of opportunity was to exploit the compact size for FPGA implementations without external memory.
That was a decade ago. With small size becoming less and less important, I see this door closed.

As a matter of interest where do Qimi's/Qzero's/Q68's cores come from and how are they implemented?

Not sure what you mean. Except the CPU, they are from me, mainly implemented using Verilog and schematics.

Do they have to be re-implemented after power loss?

Again not sure what you mean. The Q68 FPGA is flash-based, instant-on after powerup.
QIMSI and Qzero use SRAM based FPGA and load the logic from SPI Flash during powerup.

[Derek_Stewart]

The question one might well ask is, what would be the point?
My answer is as follows:- I still believe that one day, I will be able to find someone willing to finish the Stella project. No user interface was ever (as far as I know) finished for Stella. Finishing the Stella kernel would be relatively simple but creating a new user interface and IDE for it would be too much work in the first instance.
It would make no sense using the QL like user interface of SMSQE but the much simpler SMS2 interface might be suitable system for accessing Stella.

Stella, according to Tony Tebby, represents the zenith of his operating system designs.

If the Stella source exists, why not release it under a Open Source Licence, there might programmers that are interested in looking at the code and extending it.

[Tinyfpga]

Derek wrote:-
"If the Stella source exists, why not release it under a Open Source Licence, there might programmers that are interested in looking at the code and extending it."

Some time ago I posted a screenshot showing a copy of the Stella source code files, to see if anyone would notice. It seems as though no one looks at my screenshots. Stella does not need extending.

In any event, first things first.
Stella exists anly as a kernel. It does not have a user interface and at the moment the only user environment that would make sense is the one from SMS2. Stella was developed in SMS2 and was written in 68000 assembler:- more of which in my next post.

To resurrect the project one would want to see SMS2 running natively in something better than an Atari, an Atari in FPGA or an Atari emulator. The Q68 would be a good start

[RalfR]

Some time ago I posted a screenshot showing a copy of the Stella source code files, to see if anyone would notice. It seems as though no one looks at my screenshots. Stella does not need extending.

Link?

[Tinyfpga]

I will find the post Later on. I am busy at the moment. I have posted many screen shots.

[Tinyfpga]

My reply to Peter's post of Sat Jun 14:-

I see no reason why it shouldn't end up being open source. Its name might need changing because its now a widely used name in the domain of computing.

I remember your reply to and TT's post on a QL users group in 2017, where you wrote:-
"So nothing would be better to continue development of the Stella than the 68K architecture." (I have attached that "conversation below. It includes Arnoud's post)

I remember TT saying that C is not a suitable language to express his OS ideas. In particular the work involved in hand checking the object code for errors would be greater than simply rewriting the OS in assembler for another processor.

In reality the only other processor suitable for Stella would be an ARM processor so the code would only have to be written once more.

I am not convinced C would make a system more portable. If one were to port Stella to ARM you would then be faced with an ever changing hardware environment needing complex drivers for I/O subsystems which may or may not be closed source.

It was difficult enough writing drivers for Atari sub systems let alone the kind of stuff one finds in today's myriad of ARM SoCs.

Your argument that 68000 in FPGA is perfect for the task is, in my opinion, still true today.
It is easier to chose a 68000 in FPGA than choosing which ARM system to use.

How, one asks, do others manage to use ARM systems with such ease, well, am the answer lies in a company called Linaro, and I quote from their website:-

"Arm Solutions at Lightspeed
Linaro empowers rapid product deployment within the dynamic Arm ecosystem. Our cutting-edge solutions and collaborative platform facilitate the swift development, testing, and delivery of Arm-based innovations, enabling businesses to stay ahead in today's competitive technology landscape."

I can't imagine Linaro being interested in rapidly developing a Stella system.

Stella has no commercial future but as a fascinating alternative to the status quo it has a lot going for it. Being shackled to 68000 or being shackled to ARM, what's the difference?

Stella discussion.pdf

(7.27 KiB) Downloaded 9 times

[Tinyfpga]

Peter wrote:- Not sure what you mean.

Your earlier post about your FPGA boards confused me. I know who wrote the code. I wanted to know from what storage device the FPGA implemetation code came from.

You half answered my questions in the next sentence as:-
The Q68 FPGA comes fom internal flash
QIMSI an Qzero comes from external SPI flash.
Presumably one or two of the small ICs to the left of the FPGA.

I assume code has to be loaded into flash memory from an external source and
that the code has to reloaded into the the FPGAs on power on?

[XorA]

I spent 8 years working at Linaro, I can assure you they would have no interest in such an obscure OS

Also the reason ARM is everywhere is their heavy investment in Linux and Zephyr OSes!

[Tinyfpga]

I need no assurance. I am intimately familiar with their lack of interest.

A while back I had a personal connection to ARM's board of directors. I used this connection to convince ARM to look at TTs work. The were given a demonstration and in a nutshell their response was "What kind of nonsense is this"

I could be equally rude about Linux and Zephyr but I can't argue that they are obscure or unsuccessful.

Factoids (approx figures from internet 2024):-
Intel sold 2 billion things for $53 billion. Arm sold more than 30 billion items for $3.63 billion. I use products from both companies. Intel is currently unprofitable and ARM generates a small (In computer company terms) profit. Linaro is a non-profit company with sales of $33 million.

As an experiment I am going to buy a few shares in both Arm and Intel. I will then follow their fortunes over the next few years.