@matthey

Quote:

The RPi Zero actually offers better value than the RPi Pico for many potential customers.

Cue car analogy…

The pickup truck actually offers better value than the gokart for many potential customers.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@michalsc

Quote:

Not to mention the ahead of time recompilation of binaries from m68k to e.g. arm may be hard since the HUNK format gives only very scarce information about structure of executable

I think the solution to that (in any sort of emulation based future) is an updated toolchain and additional metadata for generated binaries to describe the AOT recompilable sections of the file.

_________________
Doing stupid things for fun...

@Hypex

Quote:

The odd thing is, is that even PPC64 which over is ten years old in the PA6T, and a PPC32 G4 at least 15 years, should be so old and obsolete that a modern x64 should emulate it easily. Perhaps PPC is just a tad too modern in design and features to be so.

I don't find it that odd. It's about expectation management. The fastest 68K parts reached a speed that doesn't even require JIT at all to outclass today. I remember some code I had reached 040 levels of performance even on my 240MHz 603 under interpretation only. This seems to match a rule or thumb that such interpreters tonk along at around 10-20% the performance of native code (some faster, some slower, depending on complexity of the emulated target and implementation strategy).

By contrast, PPC was able to reach the same order of magnitude clock speed and throughout as x64 for much longer. The fastest G5 parts are going to be a challenge because quite frankly, single core performance hasn't increased as dramatically since then as it has since the 90s. Emulators are also less mature. Rosetta is probably the most performant but most Mac users didn't have to depend on it as most actively developed applications for their platform switched to x64 when the OS did.

_________________
Doing stupid things for fun...

@kolla

Quote:

kolla wrote: @matthey Quote: The RPi Zero actually offers better value than the RPi Pico for many potential customers. Cue car analogy… The pickup truck actually offers better value than the gokart for many potential customers.

Did I mention the tank was a tank?

Sold!

_________________
Doing stupid things for fun...

@Karlos

Quote:

I dont really buy the narrative of an Amithlon solution being the inevitable death of AmigaOS development because it would have been limited to 3.x. Very little of OS4 is written in assembler, for good reason. What assembler there is resides in low level routines that manage hardware or accelerate specific functions like copying memory around.

I don't mean it would have been the death of AmigaOS development, but it didn't develop the OS any further. it just provided a means to run it faster by using a faster computer. It's possible H&P or whoever could have taken the OS3.9 sources and recompiled main applications as x86 native. Leaving as little as needed in 68k for it to function. Which would have brought AmigaOS even closer to an x86 port using Amithlon as a platform. But that didn't happen either.

OS4 brought the OS closer to x86 by porting to C. Even though at the time people complained about porting to PPC and not x86. Man that was almost 20 years ago! I argued that porting to PPC brought it closer to x86 since they used portable C code and decoupled it from the Amiga hardware. But of course that doesn't mean it will happen and hasn't yet. Though, OS4 is also a bit too close to PPC, such as with interrupts and the exception context which isn't portable and designed solely for PPC.

Quote:

If Amithlon had been a success and loads of people had gone that route, porting OS4 back to 68K would've been less effort than porting the guts of 3.1 to PPC was originally. I very much doubt anyone on a core i5 or better "modern" evolution of Umilator today running a 68K OS4, with a couple of native accelerations would have particularly missed the PPC. And with cheap commodity hardware for said modern Umilator there's be a much bigger market for OS4/68K than there ever has been for PPC.

That's the thing. The point of OS4 on PPC was to move away from 68K because they stopped production. Moving back makes no sense as the 68K hasn't made a comeback. It needs to be ported to a sustaining CPU that won't disappear ten years later. They should probably port it to little endian first. Like Apple did but keeping the OS and not replacing it. It could still run on PPC but when it's matured enough it can then be ported to a more sustainable CPU. I think they should have ported it to the Itanium. IA64 was what become of the HP-RISC. OS4 could have been running on an Intel but the AmigaOne wouldn't be a PC!

Quote:

Even without any new high end 68K to run on, OS 3.1.4 and 3.2 have still done pretty well.

Yes, though it's more of a side grade after the OS3.9 debacle, and it targets real 68K Amigas so you got the biggest market there.

Quote:

Why though? As an Amiga fan I saw it as a neat solution to a difficult problem. That problem being the discontinuation of 68K hardware. One that didn't require porting to an alien architecture that was always at risk given that the only real consumer was Apple. One that allowed the 68K to continue, even if nothing more than a CLR, but with the advantage of genuine backwards compatibility.

Because for one thing it wasn't Amiga software. A lot of Amiga people went crazy. But it didn't run on an Amiga. It ran on a PC. It was PC software. I was an Amiga person, Motorola Motivated and Intel Outside, that's the team I rooted for.

This just made the PC look good and the Amiga look bad. Again. It also compared with songs like Even better than the real thing by U2. So yeah, it was like a kick in the guts for the Amiga. Even in the Amigas heyday when the hardware was still better than the average PC it still couldn't emulate a PC faster than a PC.

Also, a Mac was closer hardware to what an Amiga was, custom designed with a Motorola CPU. I saw demos of Umilatror, the Mac version, whatever Umi meant. Had it been released I may have bought a Mac!

@Hypex

Quote:

Because for one thing it wasn't Amiga software. A lot of Amiga people went crazy. But it didn't run on an Amiga. It ran on a PC. It was PC software. I was an Amiga person, Motorola Motivated and Intel Outside, that's the team I rooted for

This is where the big disconnect happens for me. It wasn't PC software. It was just software. In the context of something like Umilator, it wasn't even software in the common end user sense, it was more like firmware. You didn't interact with this software. You could, if you chose to or needed to, e.g. updating the kernel for example. Let's be clear about boundaries to. Umilator wasn't the OS as far as the end user was concerned, it was the platform on which to run the OS.

On this note, OS4 didn't run on an Amiga either. It could run on one, but it wasn't really the platform of choice. I ran it on both classic hardware and a G4 A1XE machine and it's clear what the target was. The classic had suddenly become the unexpanded A500 of the day, all the new hotness was for the new hardware.

So looking at the new hardware, what was it? It wasn't a custom designed PPC Amiga. It was a pre-existing PPC evaluation board. It had all the same components you found in a PC around that time, bar one: the CPU. Ok, the Articia too, but it's only job was to interface the PPC to all the vanilla PC buses and hardware. The Articia. It only had one job and as we soon found out it didn't even do that right.

So we arrive at a PC system with an alien CPU that we have to port the operating system to and provide 68K emulation for backwards compatibility. You can argue the pros and cons of big versus little endian, fine. But it wasn't an Amiga and any objective system analysis would put it right next to a commodity PC.

Aside: If the desire was to have custom designed hardware for PPC we should've gone the same route as MorphOS and used PPC Macs. Which were no better at a "it's still a PC with an exotic CPU" but at least was integrated as a complete system and heavily tested before release. And you could find them second hand at reasonable prices.

In principle, exactly the same could've been done with an x86/x64 PC. Like Amithlon it could have even used a big endian memory model if it was going to be a deal breaker for backwards compatibility. Except the Amithlon model was cleaner in my opinion. Just put the entire Amiga software component inside an emulation and manage everything else so that Amiga OS doesn't need to concern itself with such things.

Users of the time had, in my view, some cognitive dissonance. They disliked PCs, but stick a different incompatible CPU into one and suddenly it was an Amiga. Even I drank some of that kool aid at the time, gaslit by the naive idea that if Apple were giving the finger to the x86/x64 world it was all possible. Except they weren't. They were developing OSX on intel machines from day zero. They always had that exit strategy. They were cynical AF about it too, still promoting the superiority of their dual G5 systems on their website on the day of the keynote speech where they announced it was all running on an x64 machine.

Instead of all this nonsense, today we could've been running an evolved, high end 68K AmigaOS on your own choice of commodity hardware, with excellent performance and compatibility. Who cares if it's virtualised if the user experience is right? When you run a legacy 68K application under emulation in OS4, you don't think "ooh, stinky underperforming throwback." I used 68K drivers on OS4 for the blizzard for things like the PCMCIA NIC I had at the time. Still do, come to think of it. Did that mean it wasn't a full PPC OS? How much of it has to be native to matter? My contention is that taken to it's logical conclusion, absolutely none of it needs to be. Would it be a lesser platform for not having a truly native, precompiled operating system? If performance is the metric by which you ought to judge such things, clearly not.

And the key advantage of staying with the 68K ISA in my view is that it would allow applications, and even the OS itself to run on a much bigger variety of hardware, as I've mentioned previously.

And without bashing OS4 (or MorphOS which in my view is equally dead ended on PPC for now), how has it evolved in any way shape or form that was contingent on being PPC native? The biggest features of PPC over 68K have not been realised. There's no SMP and no 64 bit. There's limited memory protection but that could have been realised regardless. All the eye candy, compositing and what not critically depend on hardware features other than the CPU. I contend that there's no significant feature that couldn't run on a real 68K (albeit at a performance loss due to the lack of assembler optimisation, increased abstraction and shiny things) if the code were backported. And if it could run on a real 68K, it could run on JIT, which means it could be running right now, at a performance that would probably leave all but the most carefully hand tuned PPC code behind on x64 today.

With native code passthrough (which was a feature of Amithlon), the most demanding data processing tasks could have libraries and devices and datatypes that make use of fully native code. So today you'd likely have AVX optimised datatypes and what not. 68K Warp3D and OpenGL? Those could both be library stubs running on top of a host provided native Vulkan implementation that you don't have to write from scratch yourself. Why not? Imagination is the limit.

Who knows how Amiga OS itself could've evolved on a platform like this?

We aren't victims of the platform's demise, the platform is a victim of our own obdurate intolerance of a hardware architecture we used to feel superior to and weren't mature enough to just get over as it got better. You said it yourself, instead of embracing it as a path out of obsolescence, you saw it as a kick in the guts and stuck with an equally alien architecture (unless endianness is the only measure of similarity) with a much less certain future. And all with the added, crippling irony of using everything from that architecture except the one thing that was future proof by virtue of the sheer scale of it's installed user base: the CPU.

Let's say, as if by magic, I presented tomorrow a complete working solution based on the principles I've described bere. A Umilator inspired solution, on an updated kernel, complete with extensive hardware support, RTG, 3D, AHI, USB, SANA etc out of the box, running a bang up to date JIT, shipping with native SSE/AVX optimised datatypes, etc. Free software. Provide your own OS. The only caveat being that it runs 3.x with JIT PPC emulation in the works. Would you pass that up for the advantages of existing PPC? And if so, can you enumerate what those advantages are?

Last edited by Karlos on 03-May-2022 at 12:35 PM.

_________________
Doing stupid things for fun...

@Karlos

As I understand it you prefer a kind of amithlon solution booting in 68k environment. Then there is still the question how you want to emulate the amiga hardware (expecially chipset) because people want to use the old stuff (expecially games) on it.

Modern to me means: 64bit, full MP, SMP support, USB3, full 3D support and so on

Full memory protection is in a environment that is used seriously not a can but a must because of security. On 68k this breaks most software without adding really new. I do not see it as realistic. Amithlon was a great project at its time but today the situation is different.

@OlafS25

Emulating the legacy hardware is a nice to have, not a must have. After all, it's not as if any NG platform provides this outside of UAE (aside from OS4 and maybe MOS on the original hardware). And equally, with the possible exception of AROS, none of the NG platforms offer the other features you list either, but Linux does and a lot of that can be exposed to the amiga environment through thin layer drivers/libraries making use of native code entry points.

As for security etc. You can't modernise AmigaOS as regards security, SMP, 64 bit etc without transforming it into something totally unrecognisable and being a second fiddle to every other mainstream OS (many of which are free). I'd actually rather keep the original/current UX running in a more secure sandboxed environment than a half-arsed hack that in all probability would fail to achieve everything on your list anyway.

_________________
Doing stupid things for fun...

@Karlos

then you end in niche again with something not interesting outside the amiga hardcore bubble. It would be something like Amithlon 2.0 but for die hards chipset is a must because it is a retro bubble. I can only say "go" if you want but I personal do not see the potential audience for it.

Last edited by OlafS25 on 03-May-2022 at 11:49 AM.

@OlafS25

You could still run other operating systems on it. Multiboot isn't exactly a new idea.

As for chipset emulation, I'm all for it, as long as it can be done in a manner with some genuine asynchronous implementation that allows the costlier aspects to be offloaded from the core dealing with the JIT. This is certainly a more complex proposition than UAE and may not be feasible with total compatibility. In which case, you could provide both models. Fastest possible 68k experience with no chipset emulation or a less performant one with chipset emulation implemented in a similar vein to UAE. Something you decide with a reboot intercept. The latter ought to be at least as performant as UAE in any case, the former even moreso.

_________________
Doing stupid things for fun...

@OlafS25

Quote:

Full memory protection is in a environment that is used seriously not a can but a must because of security. On 68k this breaks most software without adding really new. I do not see it as realistic.

OS4 provides private memory allocation for that purpose but as long as all the messaging systems and everything else still rely on public memory for IPC, it's no actual protection beyond a subset of illegal accesses. The main benefit of private memory is that you can also page it knowing that it should only impact the process that owns it. I'm not sure what the state of the art is for memory protection among other offshoots, but I'd be surprised if any of them had "full" memory protection. So it's hardly a 68K specific peoblem as far as I can see.

_________________
Doing stupid things for fun...

@Karlos

where did I write it is a specific problem of 68k? It is a problem that was inherited by all successors to stay compatible to old binaries (AmigaOS and MorphOS). Aros also took over the limitations. But it is critical today if you want to use something serious. So either you develop a toy platform or a professional platform. And for toying people already have a number of options. I personal do not see the market and the need for it. But perhaps I am wrong, who knows. Go for it and then show it

@OlafS25

It seemed to be implicit from the comment "on 68k this breaks...". Show it, you say. If only kidnapping were a legal means to an end... Where is Bernie these days anyway? Just asking, for a friend, you know.

Last edited by Karlos on 03-May-2022 at 12:41 PM.

_________________
Doing stupid things for fun...

@Karlos

Playing second fiddle isn’t bad, certainly better than not even having a seat in the orchestra. Or even worse, playing viola.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

The majority of Amiga users today apparently have no clue not care about how the OS works under “the hood” and really just care about “the gui”, and so it should be rather easy to “bring them over” to a modern incarnation of AmigaOS, a graphical environment on top of *ix (or whatever), one that really resembles AmigaOS as we know it, with screens (draggable), screen bars, icons with tool types, somewhat reproduced file systems structure etc. and everyone would be happy. Except of course, all amiga programmers, who easily can say “this has nothing to do with Amiga!”.

Last edited by kolla on 03-May-2022 at 10:26 PM.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

Quote:

kolla wrote: @Karlos Playing second fiddle isn’t bad, certainly better than not even having a seat in the orchestra. Or even worse, playing viola.

I played triangle for a local orchestral reggae fusion group. I just sat at the back and ting.

Last edited by Karlos on 03-May-2022 at 10:36 PM.

_________________
Doing stupid things for fun...

@Karlos

If you want a recular and steady job, try cow bells in a salsa band :)

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

Any band. All music sounds better with cowbell.

_________________
Doing stupid things for fun...

@Karlos

Quote:

Any band. All music sounds better with cowbell.

Well, that's idea fow a killer app, CowPlayer, a music player that picks up the rythm (well, try to) of the music, and more or less (an adjustable parameter) descretely mixes in a cowbell. For the bovine experience

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

It would need to play increasingly off beat and more loudly in Finkle mode.

_________________
Doing stupid things for fun...