Sound for Squeak/Scratch

Well-written software survived the change to Risc PC hardware without problems, but nothing could anticipate 32bit compatibility.

I think a lot of effort went into making the RiscPC “look” like a traditional Archimedes-class machine to most software.

The big thing with 32 bit was …. The API at the time had you preserve flags, hence MOVS PC, R14 and LDMFD R13!, {....,PC}^ did this; plus all manner of TEQP and similar to mess with the PSR in-situ. There was a clearly defined selection of what is read/written when so that you could modify status without altering PC by writing stuff to R15 (and the ARM would mask out the bits that shouldn’t be touched).
Accordingly, all of this was likely to blow up in your face when the status and PC separated into different registers. Any messing with R15 now will affect only the PC (status is in CPSR/SPSR).

As you can imagine, this is insurmountable. You can either fake it (Aemulor) or you can rewrite all of the code to work in a safe manner. While the majority of alterations are fairly simple, that times several thousand (plus extensive debugging for RISC OS isn’t too fault tolerant) would make a simple project into a personal vendetta. To think that RISC OS appears to have been 32bitted twice (Castle and later ROLtd), it makes my head spin. But probably nowhere near as spinny-pukey as the people that actually did it must have felt at times…

I think it’s already somewhat too long as it is, so I’d rather not add yet more qualifiers and exceptions to cover all possible bases.

I would argue that it is important information to mention that some support exists – as it makes the following statement a little less true:

Almost all old software breaks and has to be updated. A hefty proportion is left behind, as the ‘market’ is now too small to be worthy of the term.

The owning companies of both forks of RISC OS are effectively dead and very little software is still maintained.

Very negative, and not entirely true.

ROLtd. Dead. Anybody that says otherwise, please provide me with some content on the site that isn’t years old.

Castle. Sort-of dead. They didn’t die because RISC OS failed, it was the RoHS requirement of the Iyonix that killed it. I do not imagine we would see an Iyonix 2 as we have entered the era of the SoC. Maybe the Iyonix2 would be a lot like the ARMini? Either way, pre-packaged plug’n’play RISC OS computers are available, it’s just many of us here are handy enough with a screwdriver and reference manuals to do it the hard way.

One could wonder is Castle isn’t on to a useful thing here. They can collect licences while enthusiasts do the work. The trade-off, of course, is full source code access. There is nothing to stop us building our own custom versions, porting it to eccentric purposes, and as long as we don’t commercialise then I don’t think Castle gets much of a look-in. So it is an arrangement with the ability to benefit everybody.
Plus, obviously, ports brought to hardware that the enthusiasts like. A company by definition needs to make profit – so companies are only going to support what they think will make money and not cost money. Well, an enthusiast doesn’t have this – it is more a technical challenge, a “can it be done?” rather than “how much will it cost?”.
End result, everybody benefits.

David Pilling is still around. Okay, there’s not been much work on OvationPro in recent times, but I suspect that is because of limitations in RISC OS more than anything else (for instance my Japanese lyrics docs written with the Windows version, but unusable under RISC OS due to the lack of coherent Unicode support at this time).

In fact, take a look at the stuff on the NutPi. Sure, it is far from the glory-days of RISC OS when Arcade and Digital Databank (where is John Stonier these days anyway?) would have screenfuls of new files each time I logged in… but this does not mean the scene is dead. We’re more like on tick-over.

I would suggest that the notable exceptions are included (otherwise it is only a partial history), and please consider making it sound a little less pessimistic, eh? Like I said, we’re still around, so is RISC OS.

Oh, and:

Introduces an entirely new, 64-bit instruction set which would necessitate the complete rewriting of RISC OS and all its software from the ground up.

Technically true, but everybody is in the same boat. Sure, it will be easier to recompile Linux from C than to port RISC OS, however note that absolutely EVERYTHING will need to be rebuilt. And possibly debugged if the change triggers any “size effects”.
Assuming, of course, that we’re even interested in 64 bit ARM at this time. There’s a huge codebase and installation of 32 bit devices (more than a few PVRs, pick up practically any Android device…) so I don’t see the 32 bit realm dying out tomorrow.

Most of your complaints I had already anticipated when I was writing it. There will always be a trade-off between readability and scrupulous accuracy (witness the existence of the word ‘legalese’), and I deliberately veered considerably on the side of the former. It’s not an in-depth article, it’s a reference sheet.

hi everyone, so Chris Evans got in touch about all this stuff… I’m kind of surprised my old code from back in the day is still interesting, but the basics don’t change I guess.

Anyway… feel free to use my code.. I don’t actually have an Acorn any more, and I don’t have a Pi. I have no idea how to update my website, I’m surprised it’s still up to be honest! I think a guy called Rich Goodwin runs the hosting?

You should feel free to use my code if you need it… I’ve a feeling I wrote the versions on my web site, and presumably Andrew Sellors asked for a 32 bit version (vague recollections….) possibly after I forgot how to update my own web site. Anyway I probably sent him a version I was playing with and he compiled it for 32 bit? What’s on my site isn’t necessarily what was on my Risc PC.

I do vaguely remember burning all my interesting Acorn stuff onto a CD before I sold my Risc PC, I wonder if I have it after literally 8 house moves, and I wonder if the CD still works after 10 years. I can look at some point..!

And I saw a load of stuff about poll words, I don’t know what poll words are for now… but from memory at the time I was only using them to make sure the foreground GUI app I made could load the next block off the disk. I only used it on my old RiscStation to play music, since the machine was too slow to actually play MP3s.

Yay! Hellooooooo John.

I’m the guy that has used your code and it works pretty damn well. Apart from a somewhat wrong usage of the module-provided poll word and the doc getting the damn flag setting wrong, I had no great problems making it do the job I needed.

It would be really nice to get hold of the source for the ‘final’ version since the source package I managed to find was for a single module instead of the split modules. I’d rather like to try to persuade people that they should become part of the standard pile of modules, so the latest source and your permission would be useful.

Okay I’ll take a look later… I assume you’ve already tried A Sellors and had no joy? He might have the source of the later version.

Urm. Shuffles feet in embarrassment. Looks anywhere else.

I was going to try Andy but made an attempt to get a message to you via your website which obviously didn’t work. Forgot the next bit. I plead that thing that makes you forget stuff.

Did you get anywhere with this? Just wanted to mention that I actually moved house yet again, and looked thorough all my random stuff, and didn’t find a CD of my stuff. I did find a CD marked RiscOS software, do not distribute ever, but nothing that looked like a backup of all my stuff. So I suppose it’s gone for ever! However maybe one day I’ll get a PI and rewrite it, only 10 years since I did any assembly code!

Wow, well hi there John. Nope, nothing happened on this. I think I at some point found your code in the CVS tree but can’t remember with any certainty. I haven’t actually had any time to do any RISC OS work since, ooh, last march I guess. Oh – wait – I did get to demo it in Phoenix to a bunch of Smalltalk developers. They were quite impressed with the idea of a $35 Smalltalk machine that benchmarked at 30 Dorado, 30 years after the Dorado.