New Dedicated hardware For RO64 - speculations

I’m writing this in anticipation of a soon-to-come 64-bit version of RISC OS which could do with some dedicated hardware to run on it in both desktop- and laptop form factors – I assume there are companies still in the Acorn/RISC OS market capable of designing, manufacturing and marketing something truly special here.

Using ARM’s Neoverse V1 CPU aimed at HPC applications should be a pretty good engine, provided it and its supporting systems don’t go way beyond “consumer” price levels For a good GPU should the inclusion of “VIDC-on-steroids” Immortalis-G925 be suitable, if newly written graphics software can take full advantage of it.

(And – an absolute state-of-the-art Gamer Box could be made from a reduced version of this system, to include a Bluray-drive and plugs for modern joysticks/pedals.
Yeah, game software is an issue here. Nice to see Elite Dangerous on it.)

The latter is naturally dependent on a modern SDK which may take as much work (or more) than designing the new hardware, but if the aim is to remain a separate operating system it’s an absolute must. A very distinct Desktop case is nice.

So what I’ve written here is a proposal, in case RO64 should appear, for suitable new harware to be made in order to deserve the “Platform” designation. If anybody should find better or more suitable combos of (ARM) CPUs and GPUs, do tell.

It’s good to plan for your grandkids.

I’ve had a crappy weekend. Share the psilocybin, mate, I could do with a touch of delirium to unfuzz my mind.

So we share the fact that none of us are close to a field full of “magic mushrooms”.

I’d rather you share what reasonably-priced a bit-HPC hardware RO64 could possibly run on. Speculate.

I’d rather you share what reasonably-priced a bit-HPC hardware RO64 could possibly run on. Speculate.

The magic mushrooms sound more interesting, to be frank.

Bit tricky to run RISC OS on Mr. Semmens.

Use BeebEm – then you can run Magic Mushrooms on RISC OS

Bit tricky to run RISC OS on Mr. Semmens.

So is 64-bit hardware. I’ve got 32-bit hardware that runs it perfectly well though.

On one core out of many. My wish is for RISC OS to follow ARM’s state-of-the-art CPUs and GPUs, taking full advantage of their modern command sets. Don’t know if it’s in ARM Ltd’s interest to fund the conversion, full emulation of 26/32-bit software and the writing of a 64-bit capable SDK so that applications can be written to run fully in a 64-bit environment. Written in C to be compatible with new CPU command sets. Make RISC OS a modern operating system with preemptive multitasking and all the bells and whistles you’d expect in 2024.

On one core out of many.

Indeed. And yet it’s doing everything I want of it, perfectly well, and plenty fast enough.

If I had a sports car capable of 160mph, I’d still drive at or below the speed limit. The Volvo we used to have (that our son has now) is capable of 120mph. On a trip to Germany a few years ago I did drive for a few miles at 100mph, but not for long. 65-70mph felt much more comfortable.

Sounds like you are ‘Peter Pan’

Sounds like you are ‘Peter Pan’

Sveinung, or me? 8~)

Don’t know if it’s in ARM Ltd’s interest to fund the conversion

I don’t think they’d have the slightest interest, but they could afford it out of the petty cash…

If going seriously offtopic.. feast yer orbs @ https://morgan-motor.com/ – their legendary vehicles are even mentioned in Frank Zappa’s tune “Florentine Pogen” – “Deep in the street she drove her ’59 Morgan.” Can you get more classic than that?

Make RISC OS a modern operating system with preemptive multitasking

I suspect that, because of the poll loop, all existing desktop applications, no matter what language they have been written in, will cease to work. But I may be wrong.

If going seriously offtopic..

Not really. Just an extra example of how there’s no point in using more of a technology’s capability than necessary to fulfil the task at hand.

On one core out of many.

Yup. Thus showing there’s so much more to do with existing hardware and software.

Don’t know if it’s in ARM Ltd’s interest to fund the conversion

It’s about as relevant to them as it is to Microsoft porting DOS to x64.

Written in C to be compatible with new CPU command sets.

Didn’t somebody at ARM (or was it one of the RISC-V guys? I forget) do a study showing that most compilers are pretty crap at using the full instruction set of processors, and tend to stick to a conservative core set of instructions.

Can you get more classic than that?

Ten years later in reference, but also a classic…

🎵 I got a sixty-nine Chevy with a 396
Fuelie heads and a Hurst on the floor

will cease to work.

I think the usual bodge is to have the pre-emptive applications running as normal applications, and all of the co-op applications running in a sandbox that looks to the system as one pre-emptive app.

It’ll majorly screw with the Wimp_SendMessage system, but to be honest that’s possibly the least of our problems…

more of a technology’s capability than necessary to fulfil the task at hand.

But Bigger! Better! Shiny! Shiny! Shiny!

Well, it works for Apple…

Daisy, Daisy,
Give me your answer, do!

Better by half. Clive is good at cryptic stuff

Clive is good at cryptic stuff

Not that good – I’ve absolutely no idea what almost half of what’s going on is all about.

It’s about as relevant to them as it is to Microsoft porting DOS to x64.

And that’s why we have DOSBox, to do the things neither Microsoft nor game devs have any interest in doing.

I’ve absolutely no idea what almost half of what’s going on is all about.

Neither does the OP it would seem…

Not that good

I know it was from before our time, but the next line is?

Make RISC OS a modern operating system with preemptive multitasking and all the bells and whistles you’d expect in 2024.

Or…

Write an emulation layer that mirrors the existing API. Run said emulation layer on an AArch64 port of a unix-like system on a Pi5. Run RISC OS apps on an X (or nowadays Wayland) desktop.

The advantage is that the applications will multitask properly (as in pre-emptive). But you’ll also be able to run native *nix apps alongside. Such as Firefox.

My background using RISC OS in a more or less professional way is within Desktop Publishing, using !ArtWorks, Photodesk, Ovation Pro and Impression Publisher to make a book, an exhibition catalog, a raft of flyers for a rave party and later, CD inlays and CD-plate graphics plus advertising material for these, jobs for the CDROM Dept. of Oslo Uni, all written to film from PostScript files and sent to the printers. Those were the days..

With Linux graphics software I simply cannot achieve the results I did in RISC OS, like graded transparencies in color transitions on the same vector object (!ArtWorks) and so on. This is why I miss working with RISC OS – for its sheer productivity and ease of use. With Martin Wuerthner’s PostScript 3 driver one could achive some pretty stunning results, but my currently cramped living quarters and crappy economy I cannot set up the Risc PC nor the very-upgraded Archimedes 310 anywhere – believe me, I wish I could. For the sake of productivity and good results.

I’m not 100% certain, but I suspect that all the RISC OS titles I’ve mentioned above has been written in Assembler possibly mixed in with BBC Basic which make them hard to port to a 64-bit system so full sandboxed emulation is necessary to run them. Not being a programmer I can’t evaluate at what speed this would be – if at the level of ARM3 it would be fully usable, closer to StrongArm would be spectacular. I’d like some comments on this, if you please.

has been written in Assembler possibly mixed in with BBC Basic

Impression – devious assembly using the BASIC assembler, custom linking/fixing up. There’s a reason this is still a work in progress so many years later.

Artworks – that was CC too wasn’t it? So probably more devious assembly.

Photodesk – app core in BASIC, with assembler for speed. Complex add-on system with code mostly in assembler but some parts (like the JPEG loader) in C.

OvationPro – written in C. That’s why there’s also a Windows version. People might have said David’s work was “slow” being written in C rather than painstakingly hand optimised assembler… But remember, it wasn’t the wabbit that won the race in the end.

I simply cannot achieve the results I did in RISC OS

What did Xara become? Or did Corel kill it off? I ask because didn’t the rewrite of ArtWorks for Windows eventually become Xara? So it would be logical to assume it could do much the same.

my currently cramped living quarters and crappy economy

Try a Pi. Cheap, small, and would absolutely run rings around your RPC.

nor the very-upgraded Archimedes 310

My 310 has Arthur 1.20 inside. 😂

so full sandboxed emulation is necessary to run them

If that’s not necessary anyway, then this mythical new OS has been designed entirely wrongly.

if at the level of ARM3 it would be fully usable, closer to StrongArm would be spectacular

Meanwhile a regular Pi 3/4 with only one core is p**s-your-pants fast compared to the likes of a StrongARM machine.

Give yourself a pat on the back, because going from ~1.2GHz native to around 230MHz emulated is not only “a retrograde step”, it’s also a massive self own.

So can we maybe just work with and improve the current hardware that does run RISC OS rather than chase these 64 bit ¹ delusions that will only create more problems than solutions?

¹ That you seem singularity obsessed with 64 bit means you are definitely doing it wrong. Any future rewrite of the OS must be designed to be processor agnostic so it can, in theory, be widely ported rather than being directly targetted at specific versions of specific processors, which is, if you recall, one of the reasons we’re in this situation in the first place.