RPi5 out

Falcon: Whoosh. Nice.

Dodo: Assuming that’s still an RPCEmu-Interpreter?

What about running RISC OS in a web browser.

Tried that (admittedly on a somewhat pedestrian Wintel laptop – 2.5GHz Core i3), but not encouraging: processor speed was one-twentieth of a RPC (5%). Most of the other benchmarks were in single-figure percentages, though Memory was 729%, nearly as fast as a Pi 1!

Dodo: Assuming that’s still an RPCEmu-Interpreter?

It is. Fair do’s though, it is almost three time faster on the 2400MHz RPi5 than on the 1800MHz RPi400, at about half the speed of an RPi1.

Anyway to find out the performance under aarch64 on the Pi5? Would be interesting to compare to native aarch32…

The Dodo was a flightless bird, with the combination of being slow to move and tasting very good = extinct.

Which is what RISC OS will become unless it can recruit thousands of new users, something which didn’t happen with the Raspberry Pi SBCs so new 64bit Arm-based desktop and laptop computers with 64-bit RISC OS possibly running from ROM, able to emulate all legacy software and with a user friendly software development kit, could attract new users, professionals and hobbyists alike.

Be honest, RISC OS need such a revival to survive in the long run.

Be honest, RISC OS need such a revival to survive in the long run.

Be realistic, RISC OS is what it is, not a magical future that saves humanity.

Which is what RISC OS will become unless it can recruit thousands of new users, something which didn’t happen with the Raspberry Pi SBCs

When RPi 1 happened, RISC OS was much better equipped than its competition for some tasks because of its low memory footprint and RPi 1 (especially Model A) lack of RAM. Now, more than 10 years later, the hardware is much more powerful, and the competition is even further ahead. RISC OS has not catched up in any meaningful way, but has fallen behind even further. Where would you expect those new users to come from? The only use case for RISC OS is “I always used it and will therefore continue to use it because migrating somewhere else is something I am not interested to do.” And has been for a long time now.

Be honest, RISC OS need such a revival to survive in the long run.

RISC OS survives in the same way it survived the last 30 years: largely irrelevant for IT, but cherished by its ever-diminishing user base. A revival is very unlikely to happen, that is the honest truth. Because all the ingredients for a revival beyond the “retro scene” are just not there – even if we had a perfectly working 64bit RISC OS with perfect emulation layer for all the old apps.

The general move of the ARM/SoC/SBC world to Aarch64 puts RISC OS into the same state of play as it already was in 1997: the CPUs it can run on are getting expensive and/or slow compared to the competition. So the future will (again) be emulated, because emulation will soon outperform native systems.

Coming up with a decent dynamic JIT emulation for the CPU as well as a high-performance portable emulation base layer for the few pieces of hardware that RISC OS actually has drivers for (networking, video, audio, FS) is a much better bet to future-proof RISC OS than actually trying to convert RISC OS into a native Aarch64 OS.

The Pi 4B is going to be on sale for many years to come, and I’d rather we made full use of that before worrying about 64 bit only processors which can be catered for by emulators, in the medium term.

For example:-

1. Support for WiFi – no one expects to have to use wired networking these days
2. Support for USB 3 – not as important by would allow faster external storage
3. Some use of the other 3 cores, mainly as it pains me to have so much CPU power sitting idle.

not a magical future that saves humanity.

<spits tea across the room> 👍

+ what Druck said.

I hope that wasn’t good tea you just wasted.

I used RISC OS daily until 1995, and in all the time I did, I saw RISC OS based computers in exactly 4 places, the school computer room, homes of people I was at school with, some radio station that used to use one for the DJ and Steve Furbers office, unplugged on top of a filing cabinet.

I still love RISC OS, but I love it for what it is as well as what it reminds me of, even though I only use it a little compared to my pre-PC days.

I see little difference between a physical machine and an emulated one, as long as both run the software I want.
I am happy with the portable machine I made out of a Raspberry Pi 3A, a Elesar WiFi HAT and a Motorola lapdock, it’s perfect for my needs.

Would I like to see RISC OS rise up and conquer the world? Sure, but that needed to happen 30 years ago.

I’d rather just see it continue to live in any form necessary.

I see little difference between a physical machine and an emulated one, as long as both run the software I want.

Agreed – as long as the emulation runs fast enough. If I could have RISCOS running in a good sized window (up to full screen, 3840×2160) on my M1 Mac Mini, as fast as it runs on my Pi4 or nearly so, I’d be very happy. (Although the Pi4 probably uses significantly less electricity…)

@Clive: Isn’t that already possible ? Asahi Linux for M1 + RISC OS Linux port ? I think Chris had it running on a M1 a while ago…

I believe it possibly is, but I’ve not been following closely. If you could point me to an idiot’s guide to how to set it all up, I’ll give it a whirl. Is the performance anywhere near as good as the Pi4’s?

I think Chris or somebody capable has to provide some idiots guide. I don’t have an Apple system or run Risc OS on Linux. From reading about Asahi Lunux it seems at a very good level, even GPU driver is there after being reverse engineered. That driver seems even more compatible to the GL standard than the Apple one :-)

If the current companies who tweak RISC OS had the money/skills to do a 64-bit version in C (to make it both portable and maintainable) they’d done so years ago.

Guess that development died with Acorn Computers, who were capable of doing it – remove the dependence on first-generation Arm CPU’s, enabling long-term survival.

The currently dwindling RISC OS market doesn’t quite attract the sort of funding necessary to perform this crucial renewal of both Operating System and its SDK. Which is a bloody shame, because of the rabid performance and productivity it’d have supporting Arm V9 and beyond architectures.

Now that the first Raspberry Pi that can’t run RISC OS natively has appeared, is this an opportune moment for the platform’s remaining entities – ROD, ROOL, Elesar, R-Comp, RISCOSBITS et al – to get together with any interested developers to discuss and hopefully agree a route to the future that makes efficient use of the platform’s scarce resources? The next Wakefield show in April might offer an opportunity, since the major players will be there already. What we don’t need is the usual RISC OS syndrome of simultaneous reinvention of the wheel in two or three different places, all mutually incompatible.

If the current companies who tweak RISC OS had the money/skills to do a 64-bit version in C (to make it both portable and maintainable) they’d done so years ago.

Do we know how much money and which skills are needed? I have not yet seen a convincing “big picture RISC OS 64bit” target. Maybe it exists, and only typical RISCOS secrecy prevents a broader discussion? Same old recipe for failure.

It’s said to take around 20 man-years to do a full 64-bit conversion, in C, total wages at around a million UKP/year. Plus a new SDK must be developed, full legacy software emulation, office space rented – don’t think the currently shrinking RISC OS marketplace is capable of attracting that sort of funding, sadly.

Skills? Deep knowledge of early Arm assembler and modern C.

It’s said

By whom?

to take around 20 man-years to do a full 64-bit conversion, in C

Based upon what metrics? And will this be “translating assembler to C” or the rather more sensible approach of “write something in C to do what the assembler version does”.

Also, will one be retaining the existent API? If so, why?

total wages at around a million UKP/year.

Uh-huh. Median programmer salaries of the few that still dabble with assembler and do OS level stuff are around $100K/year leftpondian way. Obviously more in Silly Valley and less in the Shenandoah Valley, but it’s far from the million you casually drop here.

Thirty five hour work week, forty seven weeks per year. That’s about 1,645 hours per year. With a salary of a million, you’re expecting a little over £600 per hour.

To put this into comparison: The average computer programmer salary in the United Kingdom is £31,037 per year or £15.92 per hour. Entry level positions start at £26,325 per year while most experienced workers make up to £41,509 per year.
and: The average developer salary in the United Kingdom is £50,000 per year or £25.64 per hour. Entry level positions start at £40,000 per year while most experienced workers make up to £75,000 per year.
[https://uk.talent.com/salary?job=developer]

So clearly you should call yourself a “developer” and not a “programmer”. But in any rate, that figure you have given is entirely plucked out of your posterior quarters.

Plus, this is RISC OS. Zuckerberg ain’t gonna drop a few billion to buy it, and the other guy is fast losing cash over his utter inability to run a social networking site. So no sugar daddies to pay ridiculously inflated salaries.

full legacy software emulation

Might as well just do that for the entire OS and be done with it. Minimum derangement, maximum compatibility.

office space rented

2023 and no WFH? I think you’ll find it hard to attract any talent if you’re expecting them to schlep to some random building and waste hours of their lives in commutes. Covid taught us all one thing – being “at the office” is only of use to shitty managers who don’t trust their staff and wield their presence like a weapon. If you want to get actual work done, leave the office, and those sorts of managers, behind.

Skills? Deep knowledge of early Arm assembler and modern C.

And… Operating system design, perhaps? Or are we only going to clone RISC OS in as much as is necessary to get it running on 64 bit machines? If so, how d’you plan to work around the swathes of existing software that expects a 32 bit world? If the answer is emulation, then what’s the point of creating a new OS (those twenty man years and kerching!)?

I can fully get behind those looking at providing emulation solutions such as QEMU or whatever. But I can’t help but feel that those who think of porting the OS to 64 bit (in whatever language)…….probably haven’t understood the problem.

typical RISCOS secrecy prevents a broader discussion?

I hope not, as “secrecy” is how we appear to have ended up with two stacks. Now that the OS is open source, it’s time to start being open about where it’s going in the future. Secrecy only serves to obfuscate these days.

The next Wakefield show in April might offer an opportunity, since the major players will be there already

Why wait till then when you have ALL the major players at the MUG RISC OS Xmas Market event on 9th December

Or indeed most likely at the South West show in February 24

I agree this would take some co-ordination of effort and a roadmap of how to get there whilst maximising the current hardware we have in the meantime.

But isn’t that roadmap already there on this very site just awaiting the magic dust to conjure up some much needed developers.

The person who suggested it’d take 20 man-years is a much respected and longtime RISC OS programmer who mentioned it in a phone conversation. I don’t believe this person would appreciate me mentioning something said in confidence.

The word Dreckly springs to mind :-)

This person… called Mike and lives on a farm in the arse end of Kansas by any chance? I have a feeling I’m hitting the old déjà vu here…

So… we’re talking about a 35-year-old operating system that still (7 years after the Pi3 was released) doesn’t support Wi-Fi. And will only run on the first core of a multicore CPU. (So basically anything more than a Pi Zero or Pi 1 is wasted.)

There is an operating system out there that uses all the features and capabilities of the Pi. It’s called Raspbian.

So here’s a radical thought. Write an emulation layer for Raspbian that allows RO software to run.

If developed further, this layer could be dropped on top of piCore (a TinyCore build for the Pi).

As for the window manager – it’s probably easier to write a new WM for X (or modify an existing one) to look and behave like the RISC OS desktop than to attempt to kludge the Wimp into running on AArch64, or support proper pre-emption, VM, memory protection, privilege separation… etc etc.

The advantage there would be that with a RISC OS-alike WM for X11, you could run Firefox and it would appear to be running native in RO.

An emulation layer is, I feel, going to be the key to keeping RO alive. Like how 64-bit Windows can run Win32 apps using the WOW64 (Windows-On-Windows) subsystem. Even then, when Windows 7 x64 came out (and possibly in 64-bit Vista too), M$ removed the 16-bit runtime from 64-bit versions of Windows. Which means that your old legacy Windows 3.1 apps will no longer run on Win64.

That’s how they did things in the Wintel world. Might be worth seeing how things are done on the Dark Side.

I have long thought the same.
What I like about RISC OS is the way the GUI works.
I imagine this needs a rewrite of the window manager on linux/freebsd.
Thats going to take a bit of work.
Then it needs as many RISC OS SWI calls as possible veneered into the host os.
Then you build a basic interpreter for that environment.
Then you might have an environment where apps written in C might be recompiled for the host os while still thinking its RISC OS, & things written in BASIC might stand some chance of running.
The desktop & core c library can fork RISC OS like processes with extra sockets for handling RISC OS things that have no unix equivalent and cant just be veneered. (I recently saw a core dump on my mac which suggests that’s what they do too)
I made a start on this last bit one time before I had children…