@Hypex

Quote:

The simple explanation is that they needed to move because the 68K was too slow and not available anymore.

And yet despite two decades of the proliferation of low cost x86 hardware and Apple making pronouncements that it was leaving the architecture, significant investment continued into developing OS4 for PPC.

Look back at the old mailing lists from the early 2000s and you will see that the choice to go with PPC was largely religious. The mere mention of x86 was met with chastisement.

@ne_one

Quote:

the choice to go with PPC was largely religious.

The religion being compatibility/interoperability with legacy 68k code, yes.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@ne_one

Well the x86 was a bit of a hack by that stage and stemmed from the old days. The only reason to use it was because it was common and fast. Why chose a processor where the core design predates the Amiga and C64 both? Perhaps we should ask why was only x86 and PPC considered and not another more suitable CPU? Why not the Intel Itanium for instance which has more in common with where Commodore were planning on heading?

By that stage the x86 had flat memory access. But it was still stuck at 8 registers. The 68K had 16 in total, divided into address and data, as it were. It was only at 64-bit that x86 could match the total 68K register count. So, when porting the API to an x86 ABI, they would have needed to cut the code from having 15 registers tops for parameters down to 7. They would have needed to stack a lot of parameters for larger functions.

The 64-bit x86 was just around the corner, but still a few years away. Not that using an inferior register count mattered, as I think it does, because this is exactly what happened when they went to PPC. The common PPC ABI they chose only allows 8 register parameters. So in the big OS functions, the extra ones have to be put on the stack. Ridiculous really as PPC has 32 registers tops. double what 68K had, four times as much as x86 and double what x86-64 has.

And there is the big one. The big endian problem. Big endian is intrinsically tied to both the hardware and OS design. Aside from the custom chips being 16-bit big endian based the copper executes 32-bit wide big endian code. Now by this stage x86 could swap data around in registers (BSWAP) so likely it would have needed to do that with memory access. Years later x86 would get direct big endian support (MOVEBE) but it was still too early. I think a customised compiler would have been needed that operated the x86 in a virtual big endian mode. And this would have made ASM coding harder except with a virtual x86 big endian assembler.

I look at x86 ASM even now and still cringe. Even though it has more in common with 68K than PPC. And back then there was more of an ASM culture. The PPC was unfriendly, IMHO, but at least it can be understandable. Still it got some fans. I can read it I wouldn't want to write it.

And of course, the Amiga and fans were Motorola. Intel and x86 were diametrically opposed. That was the enemy.

Last edited by Hypex on 23-Jul-2018 at 08:09 AM.
Last edited by Hypex on 23-Jul-2018 at 08:06 AM.

@Hypex

Quote:

Hypex wrote: @ne_one Well the x86 was a bit of a hack by that stage and stemmed from the old days. The only reason to use it was because it was common and fast. Why chose a processor where the core design predates the Amiga and C64 both?

This isn't a good technical reason.
Quote:

Perhaps we should ask why was only x86 and PPC considered and not another more suitable CPU? Why not the Intel Itanium for instance which has more in common with where Commodore were planning on heading?

Because the first Itanium versions sucked a lot: extremely expensive (but this isn't something which hurts post-Amigas, as we know ), drawing a lot of power, and with poor performances.

Other CPUs, yes: possible. But at the time x86 was cheaper, very good performer, and with an extremely wide support.

So the real question should be: why NOT adopting it? In fact they were so much convenient that Apple was (on 2000) about to leave PowerPCs and joining Intel (for more details, look at one of my last message in the 68K Development thread). And Apple was one of the PowerPC consortium founder, eh!
Quote:

By that stage the x86 had flat memory access. But it was still stuck at 8 registers. The 68K had 16 in total, divided into address and data, as it were. It was only at 64-bit that x86 could match the total 68K register count. So, when porting the API to an x86 ABI, they would have needed to cut the code from having 15 registers tops for parameters down to 7. They would have needed to stack a lot of parameters for larger functions.

Who cares? Apple maintained MacOS X for both x86 (which was the PRIMARY platform a the time) and PowerPC. And x86 were very good performer despite the limits that you mentioned.
Quote:

The 64-bit x86 was just around the corner, but still a few years away.

Unfortunately true, but looking at how long it took to complete the Amiga o.s. port to PowerPC, there's plenty of time to initially go for x86 and then switch to x64.
Quote:

Not that using an inferior register count mattered, as I think it does, because this is exactly what happened when they went to PPC. The common PPC ABI they chose only allows 8 register parameters. So in the big OS functions, the extra ones have to be put on the stack. Ridiculous really as PPC has 32 registers tops. double what 68K had, four times as much as x86 and double what x86-64 has.

If this is really the case then, yes: quite ridiculous. Even the x64 ABI has more spare registers to use to pass parameters and for locals.
Quote:

And there is the big one. The big endian problem. Big endian is intrinsically tied to both the hardware and OS design. Aside from the custom chips being 16-bit big endian based the copper executes 32-bit wide big endian code.

But with a "small detail": the post-Amiga machines had NO Amiga chipset! No Copper! No Blitter! No Paula! No CIAs! ... NOTHING was left from the Amiga hardware!!!

On the contrary, almost all chipsets and peripherals were little-endian because they were... rolling drum... x86 stuff!
Quote:

Now by this stage x86 could swap data around in registers (BSWAP) so likely it would have needed to do that with memory access. Years later x86 would get direct big endian support (MOVEBE) but it was still too early. I think a customised compiler would have been needed that operated the x86 in a virtual big endian mode. And this would have made ASM coding harder except with a virtual x86 big endian assembler.

Well, but the other big question is: WHY?!? When you make a port of an o.s. to a new hardware platform, like OS4 was of the original Amiga o.s., why on Earth you need to care SO MUCH of emulating the old (68K) platform? You run the native o.s. on the original platform, and then the old code on a sandbox with an emulator (UAE was already available and very good). Fertig!
Quote:

I look at x86 ASM even now and still cringe. Even though it has more in common with 68K than PPC. And back then there was more of an ASM culture. The PPC was unfriendly, IMHO, but at least it can be understandable. Still it got some fans. I can read it I wouldn't want to write it.

Well, I still prefer the x86 assembly to the PowerPC one: much easier to understand and use. As I've already stated, I don't like load/store architectures: I'm still a CISC guy.
Quote:

And of course, the Amiga and fans were Motorola. Intel and x86 were diametrically opposed. That was the enemy.

Yup. Religion. I was an Amiga fanatic even after some years from the Commodore bankrupt. But looking at the substantial immobility, I finally realized that the dream was over, and I had to switch to the over-hated PC.

Anyway, fanatism was insane. And I was still too young to understand that it doesn't make sense at all fighting for bits of metal and software. Better to stick to the technical side.

Finally, it's really funny, looking at the thread, how some people "suddenly" completely changed their mind. From being blind fanatics of the post-Amiga religion, to implacable opponents widely promoting the emulation way.

In Italy a very well known and appreciated journalist (which died some years ago) was used to say: "Il tempo è galantuomo".

@cdimauro

Quote:

Quote: Not that using an inferior register count mattered, as I think it does, because this is exactly what happened when they went to PPC. The common PPC ABI they chose only allows 8 register parameters. So in the big OS functions, the extra ones have to be put on the stack. Ridiculous really as PPC has 32 registers tops. double what 68K had, four times as much as x86 and double what x86-64 has. If this is really the case then, yes: quite ridiculous. Even the x64 ABI has more spare registers to use to pass parameters and for locals.

Pffft, I think 8 parameters on PowerPC is way too much. In the x64 ABI you have "only" 6. How many parameters are in typical function call? I think it is very often less than 6. Some Amiga graphics.library calls have 9-10 parameters and I think they are cumbersome.

On PowerPC registers r0 and r3-r12 are reserved for called function while registers r14-r31 are reserved for callee. One could shift this border but I believe designers tried to find optimal balance.

_________________
Amiga Developer
Amiga 500, Efika, Mac Mini and PowerBook

@cdimauro

Why? Because large chunks of OS4 (and MorphOS alike) was initially developed on 68k Amiga with and without PowerUP cards, that's where the developers were. Heck, quite a few OS4 developers are still on that hardware.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@Hypex

Quote:

And of course, the Amiga and fans were Motorola. Intel and x86 were diametrically opposed. That was the enemy.

Bingo... and the developers bought in to this as well.

Despite all of the incessant arguments about compatibility, the OS could have - and should have - been made hardware agnostic.

Long before 4.0 was completed, PPC was known to be a dead end.

@kolla

Quote:

Why? Because large chunks of OS4 (and MorphOS alike) was initially developed on 68k Amiga with and without PowerUP cards, that's where the developers were. Heck, quite a few OS4 developers are still on that hardware.

Here's an interesting concept that we've learned in the last few decades:

Avoid tightly coupling the software to the hardware.

It amazes me that Apple made the transition to x86 as the AmigaOS was in the process of being ported to PPC and yet the whole exercise was conveniently ignored.

"But the performance will be compromised... and the applications on FF#47 won't work... and Intel sux!"

Argh.

@ne_one

Quote:

Long before 4.0 was completed, PPC was known to be a dead end.

Development of OS4 started in 2001, long before PPC was known to be a dead end.

@pavlor

Quote:

Development of OS4 started in 2001, long before PPC was known to be a dead end.

In fact the PPC's heyday was yet to come! By 2006 all three of the mainstream games consoles were using the PPC. However, Apple announcing the transition to Intel the previous year was a sign that the golden years may not last forever! The PS3 is a great machine and I think the Cell is the ONLY reason that The Last of Us was possible in 2013 on a 7 year old console! The likes of that increase in quality as the generation progressed will never be seen again. Hell we might not even get clear cut games console generations again!

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

@ne_one

Quote:

It amazes me that Apple made the transition to x86

Apple never did that transition, NeXT did.

Under Apple the transition was _from_ x86 _to_ PowerPC, and once it was finished... presto, OSX.

In fact, Apple ported several operating systems to their PowerPC systems.
* MacOS Classic, from 68k
* Linux - remember mkLinux?
* NeXTStep, aka Rhapsody, eventually OSX, from x86 (and others)
* A/UX was on its way too, but became redundant once they bought NeXT

Last edited by kolla on 27-Jul-2018 at 07:54 PM.
Last edited by kolla on 27-Jul-2018 at 07:52 PM.
Last edited by kolla on 27-Jul-2018 at 07:45 PM.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

Quote:

Apple never did that transition, NeXT did

No. NeXT started on their own 68K platforms, then ported NeXTStep to X86, PA-RISC and SPARC-based platforms as they moved from a hardware centered company to an operating system vendor.

When Apple purchase NeXT, they used NeXTStep as the basis for OSX (porting everything to PPC), but its likely that X86/X64 variants were developed concurrently in order to keep their options open to switch to an alternative ISA.

And its quite likely that the initial PPC versions of OSX owe some of their structure to the 68K versions, as both a big-endian platform.

The X86 project no doubt started at NeXT, but was continued at Apple, leading to the final decision to make the jump in ISAs once then IBM 970FX failed to top 3 GHz.

So both companies deserve credit for this transition.

@JimIgou

Quote:

JimIgou wrote: @kolla Quote: Apple never did that transition, NeXT did No. NeXT started on their own 68K platforms, then ported NeXTStep to X86, PA-RISC and SPARC-based platforms as they moved from a hardware centered company to an operating system vendor. When Apple purchase NeXT, they used NeXTStep as the basis for OSX (porting everything to PPC), but its likely that X86/X64 variants were developed concurrently in order to keep their options open to switch to an alternative ISA. And its quite likely that the initial PPC versions of OSX owe some of their structure to the 68K versions, as both a big-endian platform. The X86 project no doubt started at NeXT, but was continued at Apple, leading to the final decision to make the jump in ISAs once then IBM 970FX failed to top 3 GHz. So both companies deserve credit for this transition.

So your initial "No." is really a "Yes." :) All Apple had to do in the small time frame that was "OSX on PPC" was to make sure it continued to work on x86. Not exactly super-tricky.

Oh, and you know what else is/was big-endian? Sparc. And PA-RISC.

Last edited by kolla on 28-Jul-2018 at 07:12 PM.
Last edited by kolla on 28-Jul-2018 at 07:12 PM.
Last edited by kolla on 28-Jul-2018 at 07:10 PM.
Last edited by kolla on 28-Jul-2018 at 07:07 PM.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@JimIgou

IIRC it is because the G5 failed as a processor for laptop.

@cdimauro

Quote:

This isn't a good technical reason.

It would be if the core CPU technology still wasn't up to the same stage as the other CPU.

Quote:

So the real question should be: why NOT adopting it? In fact they were so much convenient that Apple was (on 2000) about to leave PowerPCs and joining Intel (for more details, look at one of my last message in the 68K Development thread). And Apple was one of the PowerPC consortium founder, eh!

This is comparing apples with oranges here. I think we also have to consider that at that time Apple was also dumping the real MacOS and about to replace it with OS X. Those are worlds apart and OS3 has more in common with OS4.

I'm surprised Apple was about to leave so early. They pushed for PowerPC. Why did they bother in the frst place? A computer with PowerPC slapped on the front that wasn't even a powerful PC was damn confusing. Now on Intel it would make sense.

If there is anything Apple prove it's that you can't make up excuses to replace one CPU with a completely different type because you will lose compatibility. For years we heard the argument that x86 had to stay because people needed to run DOS or whatever and people were afraid. So what! Who cares! That's total BS! Look at Apple. They ported MacOS from 68K to PPC, then replaced OS with OSX, then replaced PPC with x86, and all through they maintained compatibility. Even OS4 proves you can replace CPU and retain compatibility to a good extent. So Intel certainly could have come up with a replacement like they were doing. And they certainly could have allowed DOS applications to run. No different endian excuses here.

Quote:

Who cares? Apple maintained MacOS X for both x86 (which was the PRIMARY platform a the time) and PowerPC. And x86 were very good performer despite the limits that you mentioned.

I bet it is based on a standard ABI. Or did Apple invent their own? I wonder where the Amiga ABI came from. Was that based ona standard or did they invent it?

Quote:

Unfortunately true, but looking at how long it took to complete the Amiga o.s. port to PowerPC, there's plenty of time to initially go for x86 and then switch to x64.

They wouldn't have been able to port it. They would have had to completely rewrite it. Which would have taken longer. By which stage doing x64 first would make more sense. But don't you remember "we don't have a license for x86?"

Quote:

If this is really the case then, yes: quite ridiculous. Even the x64 ABI has more spare registers to use to pass parameters and for locals.

It's getting worse. Damn SystemV ABI. Should have done their own.

Quote:

But with a "small detail": the post-Amiga machines had NO Amiga chipset! No Copper! No Blitter! No Paula! No CIAs! ... NOTHING was left from the Amiga hardware!!!

They didn't apart from bitplanes being entrenched in the graphic library API. However, back then a lot of hardware like video cards still supported big endian modes. But the CIA chips don't matter here as they are litttle endian, coming from the C64 era, and have a strange memory map that wastes space with one regster every 256 bytes.

Quote:

Well, but the other big question is: WHY?!? When you make a port of an o.s. to a new hardware platform, like OS4 was of the original Amiga o.s., why on Earth you need to care SO MUCH of emulating the old (68K) platform? You run the native o.s. on the original platform, and then the old code on a sandbox with an emulator (UAE was already available and very good). Fertig!

As you have pointed out there, with 68K, not only is the core OS structures built around big endian but 68K as well. Big reason is compatibility and the real Amiga feel. If OS4 didn't transparently support 68k apps it would have been dead ages ago. Though for years there were Amiga people who never knew this about OS4. Also, parts of OS4 remained in 68K. So UAE wouldn't have cut it.

Quote:

Well, I still prefer the x86 assembly to the PowerPC one: much easier to understand and use. As I've already stated, I don't like load/store architectures: I'm still a CISC guy.

The load/store has more in common with a Commodore C64.

Quote:

Yup. Religion. I was an Amiga fanatic even after some years from the Commodore bankrupt. But looking at the substantial immobility, I finally realized that the dream was over, and I had to switch to the over-hated PC.

So having a favourite and being a fan is religious? Hmmm. For years in my country the battle has raged on between Holden and Ford cars, which are like AMD and Intel really, but I never recall them being called religious. It could be compared to football and religious zeal for one team. Well, Motorola was part of what an Amiga was, I think they were a good combination.

However, I thought up the following conversation:
A: "I favour this God. He's the one I believe in."

B: "That's your favourite God!? You're religious! I'm not religious. I favour whatever God is popular at the time."



Quote:

Anyway, fanatism was insane. And I was still too young to understand that it doesn't make sense at all fighting for bits of metal and software. Better to stick to the technical side.

The technics would be why I was also a fan of the Amiga.

Quote:

Finally, it's really funny, looking at the thread, how some people "suddenly" completely changed their mind. From being blind fanatics of the post-Amiga religion, to implacable opponents widely promoting the emulation way.

That is funny. I suppose they have kept it classic thst way. But also used it to propel the Amiga forward onto faster hardware.

Quote:

In Italy a very well known and appreciated journalist (which died some years ago) was used to say: "Il tempo è galantuomo".

@ne_one

Quote:

It amazes me that Apple made the transition to x86 as the AmigaOS was in the process of being ported to PPC and yet the whole exercise was conveniently ignored.

And like I said before, this is like comparing apples with oranges, as Apple didn't port MacOS to x86. They dumped MacOS. But they dumped 68K first. On x86 they were on OS X.

If OS4 wasn't a port of OS3 but a completely new AmigaOS based on a modern reimplmentation of AmigaOS ideas, then yes, you could make that argument. But it wasn't, OS4 is more closer to OS3, than they are apart.

We could also almost argue. But Windows went from x86 to x64. So why can't AmigaOS go from 68k to x64?

Last edited by Hypex on 31-Jul-2018 at 01:58 AM.

@itix

Quote:

itix wrote: @cdimauro Quote: If this is really the case then, yes: quite ridiculous. Even the x64 ABI has more spare registers to use to pass parameters and for locals. Pffft, I think 8 parameters on PowerPC is way too much. In the x64 ABI you have "only" 6.

True. When I replied I was thinking to all registers (even SIMD ones), but the free GPRs are only 6 (5 on Windows) to pass parameters.
Quote:

How many parameters are in typical function call? I think it is very often less than 6.

True again, but see below.
Quote:

Some Amiga graphics.library calls have 9-10 parameters and I think they are cumbersome.

Frankly speaking I prefer to pass as many parameters as possible in registers. The altenatives are:
- pushing them to the stack;
- fill a structure and pass it.

Both have pros and cons, but I think that using the registers is the best one (and fastest).

Is there any reason why you don't like it?
Quote:

On PowerPC registers r0 and r3-r12 are reserved for called function while registers r14-r31 are reserved for callee. One could shift this border but I believe designers tried to find optimal balance.

I think that this decision makes sense: quite balanced for most scenarios.

@kolla

Quote:

kolla wrote: @cdimauro Why? Because large chunks of OS4 (and MorphOS alike) was initially developed on 68k Amiga with and without PowerUP cards, that's where the developers were. Heck, quite a few OS4 developers are still on that hardware.

AFAIK OS4 development started (on late 2001) using the Amiga o.s. 3.1 sources, plus some parts coming from 3.5/3.9.

So, the PowerUP cards usage started from there (not before) only because, at that time, it was decided to port the Amiga o.s. to a PowerPC platform.

A PC would have been used otherwise, if the decision was about porting to x86, of course.

@pavlor

Quote:

pavlor wrote: @ne_one Quote: Long before 4.0 was completed, PPC was known to be a dead end. Development of OS4 started in 2001, long before PPC was known to be a dead end.

Not a dead end, but they already had lost competitiveness and convenience (thus the decision of Apple to abandon them in favour of Intel. Postponed only because an IBM manager offered the G5 to Jobs).

@JimIgou

Quote:

JimIgou wrote: When Apple purchase NeXT, they used NeXTStep as the basis for OSX (porting everything to PPC), but its likely that X86/X64 variants were developed concurrently in order to keep their options open to switch to an alternative ISA.

x86 was the main platform.
Quote:

And its quite likely that the initial PPC versions of OSX owe some of their structure to the 68K versions, as both a big-endian platform.

Very very unlikely. NeXTStep was a platform-agnostic o.s., and binding it to a specific architecture would have been a regression.

BTW, there was absolutely no reason for such move: MacOS and NextStep/OS X had completely different APIs. In fact, MacOS apps ran in a sandbox.

The Amiga o.s. story is totally different: they decided to just port the old o.s., without fixing any of the many fallacies. Of course, the same applies to MorphOS and AROS. For this reason labelling them as "NG" (Next/New Generation) is inappropriate: it's still the same old stuff; recycled.
Quote:

The X86 project no doubt started at NeXT, but was continued at Apple, leading to the final decision to make the jump in ISAs once then IBM 970FX failed to top 3 GHz. So both companies deserve credit for this transition.

The decision was made already on 2000. It was postponed only because G5 came. See above.

@Hypex

Quote:

Hypex wrote: It would be if the core CPU technology still wasn't up to the same stage as the other CPU.

Who cares? 80286 in protected was already quite different compared to 8086. 80386 was FAR AWAY from the 8086, even in the so called real mode ("equivalent" to how 8086 worked).

Would you have used the real mode for porting any o.s., working exactly like 8086, or would you have written it to take advantage of all changes which the 80386 (and successors) introduced?

That's the (real) point.
Quote:

This is comparing apples with oranges here.

Absolutely no. We were talking about porting an o.s. to another platform, even completely different from the original one.

Something which AROS already did it, and MANY years before that there was a decision to port the Amiga o.s. 3.1 to PowerPC (aka OS4).
Quote:

I think we also have to consider that at that time Apple was also dumping the real MacOS and about to replace it with OS X. Those are worlds apart and OS3 has more in common with OS4.

Yes, they are totally different, since OS4 was just a port of the o.s. to PowerPC.

Nevertheless, AROS is the answer here.
Quote:

I'm surprised Apple was about to leave so early. They pushed for PowerPC. Why did they bother in the frst place? A computer with PowerPC slapped on the front that wasn't even a powerful PC was damn confusing. Now on Intel it would make sense.

Ask Jobs, not me.

Anyway, at the time PCs offered better power consumption and performances. PowerPCs rapidly lost their "RISC advantage" because processors were using tents of millions of transistors, hence the so called "x86 tax" was reduced a lot. Plus, Intel and AMD adopted aggressive CPU designs which were RISC-only domain before. Therefore the RISC "philosophy", which worked quite well with few transistors and costly memories, wasn't superior anymore. Whereas the CISC designs, which allowed better code density (which positively impacts performances at all levels) and more "useful work" to be computed with a single instruction, gained more advantage.
Quote:

If there is anything Apple prove it's that you can't make up excuses to replace one CPU with a completely different type because you will lose compatibility. For years we heard the argument that x86 had to stay because people needed to run DOS or whatever and people were afraid. So what! Who cares! That's total BS! Look at Apple. They ported MacOS from 68K to PPC, then replaced OS with OSX, then replaced PPC with x86, and all through they maintained compatibility.

Apple:
- had only a tiny market share;
- the o.s. APIs allows an easier port to completely different hardware platforms.

So, it was much easier for Apple.

Amiga was in a completely different shape. But, still, AROS was and is there...
Quote:

Even OS4 proves you can replace CPU and retain compatibility to a good extent.

OS4 NEEDS backwards compatibility because the port was never completed, and it still uses/runs some 68K code.
Quote:

So Intel certainly could have come up with a replacement like they were doing. And they certainly could have allowed DOS applications to run. No different endian excuses here.

On the contrary: Intel (and Microsoft) NEEDED to stay backward compatible due to the very large software base. There are still PCs which run DOS applications, either directly or using a virtualization layer or a virtual machine.
Quote:

I bet it is based on a standard ABI. Or did Apple invent their own? I wonder where the Amiga ABI came from. Was that based ona standard or did they invent it?

Apple had new APIs (and ABIs).

Whereas the Amiga o.s. APIs (and ABIs) are mostly the same.
Quote:

They wouldn't have been able to port it. They would have had to completely rewrite it. Which would have taken longer. By which stage doing x64 first would make more sense. But don't you remember "we don't have a license for x86?"

They had/have the license to port it to any architecture.

For the rest... AROS.
Quote:

It's getting worse. Damn SystemV ABI. Should have done their own.

Clarified. It was my mistake.
Quote:

They didn't apart from bitplanes being entrenched in the graphic library API.

I was talking about the machines, so the hardware: hence there are no bitplanes supported in hardware (unless you like EGA and most VGA graphic modes).
Quote:

However, back then a lot of hardware like video cards still supported big endian modes.

Such video cards came from PCs...
Quote:

But the CIA chips don't matter here as they are litttle endian, coming from the C64 era, and have a strange memory map that wastes space with one regster every 256 bytes.

Well, they aren't really little-endian, since they have only byte-accessible registers.
Quote:

As you have pointed out there, with 68K, not only is the core OS structures built around big endian but 68K as well.

Again, AROS proves the contrary here.
Quote:

Big reason is compatibility and the real Amiga feel. If OS4 didn't transparently support 68k apps it would have been dead ages ago. Though for years there were Amiga people who never knew this about OS4. Also, parts of OS4 remained in 68K. So UAE wouldn't have cut it.

That was the point: a never finished port. Otherwise following the AROS approach you can have old 68K applications running on a (UAE-based) sandbox.
Quote:

The load/store has more in common with a Commodore C64.

Well, I guess that you never (low-level) programmed a Commodore 64:

ADC ($44),Y

Do it with a RISC, and count how many instructions are needed.
Quote:

So having a favourite and being a fan is religious?

Being fanatic is religious.
Quote:

Hmmm. For years in my country the battle has raged on between Holden and Ford cars, which are like AMD and Intel really, but I never recall them being called religious. It could be compared to football and religious zeal for one team. Well, Motorola was part of what an Amiga was, I think they were a good combination. However, I thought up the following conversation: A: "I favour this God. He's the one I believe in." B: "That's your favourite God!? You're religious! I'm not religious. I favour whatever God is popular at the time."

https://www.youtube.com/watch?v=eWG-nHuuCRc
Quote:

The technics would be why I was also a fan of the Amiga.

Me too. Until I seriously technically analyzed all facts.
Quote:

Hypex wrote: @ne_one Quote: It amazes me that Apple made the transition to x86 as the AmigaOS was in the process of being ported to PPC and yet the whole exercise was conveniently ignored. And like I said before, this is like comparing apples with oranges, as Apple didn't port MacOS to x86. They dumped MacOS. But they dumped 68K first. On x86 they were on OS X. If OS4 wasn't a port of OS3 but a completely new AmigaOS based on a modern reimplmentation of AmigaOS ideas, then yes, you could make that argument. But it wasn't, OS4 is more closer to OS3, than they are apart.

Again: AROS.
Quote:

We could also almost argue. But Windows went from x86 to x64. So why can't AmigaOS go from 68k to x64?

Because the Amiga o.s. was bad by design. I've already talked about it the past here. If you make a search you'll find my posts with much better (and many) technical facts about the argument.

P.S. Sorry but the quote tool didn't worked for the first message, so I had to manually copy & paste your writings, losing all smiles (and a lot time. Damn!). The forum is ageing, badly, like the Amiga o.s....

@cdimauro

Quote:

We could also almost argue. But Windows went from x86 to x64. So why can't AmigaOS go from 68k to x64?

Quote:

Because the Amiga o.s. was bad by design. I've already talked about it the past here.

By 1985 standards, AmigaOS has quite advanced design in comparison to then mainstream PC market (eg. DOS/Win, Mac). Note Windows that went from x86 to x64 was not the "classic" Windows, but NT - entirely different technology.

If AmigaOS has to follow example of Windows, we need something comparable with NT: a solid basis to build upon. Impossible task with current resources.