@geen_naam

Quote:

reduces the vintage Amiga to dumb keyboard and mouse dongle. For the sake of calling it an Amiga

You have been very badly misinformed, or someone just lied to you.

@geen_naam

Quote:

I am aware that you have acces to the "chipset". But what is the point when the videocore is vastly superior to AGA.

Tell that e.g. Pinball Dreams, Worms or Lotus when you want to play them. There is a point in using amiga video modes when either the games/software require it or when you want it for some reasons. Believe me, there are PiStorm users not using RTG at all.

Quote:

There is a reason why there's an addon that captures the aga output to display it on the videocore Hdmi output.

This is for those who do want to use both and prefer not to switch permanently between two monitor inputs or do not want to use two monitors.

Quote:

And what is the point of Paula when the audio on the rpi is vastly superior.

Pi does not have audio, what it has (and hardly anyone is using it) are two PWM timers connected through simplest low-pass filter to audio jack. PiStorm does not use it, neither it uses audio pass-through in HDMI.

Quote:

Now who would like to run ethernet in a pcmcia slot when the rpi has its own.

As those who use it.

Quote:

So yes, if there is any use for the a1200 at all, it's just the keyboard and mouse.

You can say exactly the same about nearly every turbo card for the amiga.

@geen_naam

Quote:

geen_naam wrote: Quote: OS4 on 68K is immediately an order of magnitude more compatible: both with existing 68K applications, including many that bang hardware (where the hardware exists in any case) and with available solutions to run it on. You are doing it again! Why would I give 2 f#cks about 68k or chipset compatibility?

Why should anyone give a single f*** what you care about? Do you matter that much, are you of such importance that you can speak for every existing and prospective user?

Let's see. If compatibility with existing 3.x / 68k / native chipset was not wanted, why on earth:

1. Include 68K emulator in the kernel
2. Upgrade the 68K emulator to a JIT implementation for improved performance
3. Release OS4.1 for "Classic" machines complete with support for the native graphics and sound capabilites of those machines?
4. Take time to port and bundle UAE into the distribution to make it even easier to run old HW banging applications and games?

I guess they never got your memo that no f***s should be given here.

Quote:

Which I do not enjoy anyways.

Yeah, f*** anyone else, as long as geen_naam is happy

Quote:

Porting os4 back to 68k is even more stupid then creating a 68k in an FPGA and calling it the faster architecture of all times. When those lightwave benchmarks clearly show that if you account for the difference in clockspeeds, it is barely faster then a 060.

What they show is that if backwards compatibility and performance matter, PPC is not the answer. The answer is then UAE, Amiberry or PiStorm, depending on where you sit on the pure emulation versus physical hardware experience.

They also show that a 68K OS4 port would run just fine on such hardware. You know, if you cared about things like attracting sales, users or additional developers or anything like that.

Quote:

It means no compositing since no classic gfx card supports that... That means no Warp3d nova. Since no classic gfx card supports shaders.

Yes, I hear the sobs of the multitude.

Quote:

What about the PCI subsystem? Or do you think bankswitching your way into the a1200 expansion slot is a great idea.

The reasons most classic users get PCI expansion cards is somewhat obviated by the extra features the PiStorm route offers. And there's no fundamental reason why the additional 3D hardware features of the SoC in the Pi can't be exposed to the guest OS. After all, how do you think the existing RTG solution works?

Quote:

But I am glad that you admit the value of OS4.

Maybe, just maybe there's a reason for that. In any case, I'm not the one strangling it to death by pinning it on vanity hardware that is decades in development and splitting all the IP between various gits that ensure it absolutely cannot be more widely available.


Quote:

So to summarize. You hate powerpc because you cannot afford it. Do you hate Ferrari as well?

I don't hate PowerPC. Again, where on earth do you get that from*?

As for your suggestion that I can't afford it, well again, you'd be wrong. I just don't see it as good value for money because I'm an enthusiast, not an idiot.

Quote:

I cannot imagine a single 68k only binary that I would like to see running on my X5000. I have never felt the need to run that old 68k junk on my X5000 anyways.

*Hmm. Maybe it's projection, judging by that remark. You clearly hate the Amiga.

Quote:

Quote: The other option is OS4 on ARM 32/BE. I am a bit less enthusiastic about this. It does solve a lot of the hardware availability options but it excludes the entirely virtual option. It adds yet another binary format to the meangerie. Literally just hours ago, ARM was the future. And now you are less enhousiastic about ARM.

I have no idea why you should be confused by this, but I will spell it out to avoid any ambiguity:

I am less enthusiastic about ARM native OS4 running on ARM/BE-32 than I am about 68K OS4 running on ARM / Emu68. What's hard to understand here? 68K OS4 will run elsewhere too, like UAE, Vampire, expanded classics.

Quote:

Please list these NEXT GENERATION arm options.

Not following your tangent too well, but I'm not that interested in "next generation", that was you, remember? You are the one that purchased a 64-bit, SMP capable machine to run AmigaOS on.

I accept the fundamental limitations of AmigaOS as a 32-bit, single CPU offering. You can't change that without breaking it in a very fundamental way. Add to that the utterly glacial pace of development and even if there were a roadmap, I fully expect to be dead from old age before it happens to AmigaOS.

AROS has made real, tangible inroads into 64-bit and SMP. It's by far the more "next generation" descendant of the Amiga. Perhaps you should look at porting the x64 64-bit SMP fork to your X5000.

Last edited by Karlos on 15-Feb-2024 at 07:31 PM.

_________________
Doing stupid things for fun...

@geen_naam

Quote:

There is a reason why there's an addon that captures the aga output to display it on the videocore Hdmi output. Clearly RTG is the main target as it should.

BTW. This very argument clearly shows that people still do want/need to use chipset. Why would they ever need to ask for such solution or impatiently wait for it if they would use Amiga only as a keyboard/mouse dongle for PiStorm? In that case the RGB passthrough for HDMI would be of no use for them.

@geen_naam

Quote:

Nice, so if I would create an a1200 expansion slot Usb dongle, plug it into the USB port of my pc and run a 68k emulator on windows which talks with the Amiga chipset while running amigaos3.1. Then I receive your blessing to call it a real Amiga

You don't need my blessing. You are running on an Amiga 1200. One that happens to have an extremely unusual CPU expansion. Still an Amiga.

Choice of interfacing aside, it's not completely stupid idea either. I've used a few development systems for various CPU and DSP over the years where a socketed CPU/DSP could be replaced with a hardware interface to a PC running a simulator of the silicon running in situ for debugging and evaluation purposes. I think I still have one of them somewhere.

Given your experience working with ASIC, I assumed you'd be familiar with stuff like that?

_________________
Doing stupid things for fun...

@geen_naam

You call your PPC system amiga too, don´t you?

It is fun for people, please simply accept it.

"Sure, if you strip down windows 11 to the level of windows 3.11 you can run it on a 286 just fine"

Hopefully you do not want to compare your AmigaOS on PPC with Win 11

Last edited by OlafS25 on 15-Feb-2024 at 08:31 PM.

@geen_naam

I think you missed the point. You made a silly joke, but in practise described an ICE (In Circuit Emulator) which is, or was a fairly common thing.

Quote:

Apparently you do, otherwise you wouldn't try that hard to convince me that everything is better then amigaos4.1 on PPC

I'm not trying to convince you of anything. I'm stating simple facts. There are better operating systems for 64-bit multi core PPC and there are better targets for 32-bit, non SMP OS4. You don't have a good intersection, you just have the only one currently in town.

Last edited by Karlos on 15-Feb-2024 at 08:51 PM.
Last edited by Karlos on 15-Feb-2024 at 08:47 PM.
Last edited by Karlos on 15-Feb-2024 at 08:47 PM.

_________________
Doing stupid things for fun...

@pixie

Quote:

So that goes on explaining the burden that UAE has on Pi...

and this why emu68k that use the hardware mmu, just maps memory is a faster solution. But there also point emu68k is more like hypervisor, sure does tick all boxes, but does tick a lot of the boxer for a hypervisor.

Emulators like WinUAE/EUAE is also memory bound because it has refreshed the display, at high rate, as resolution increases the framebuffer increases. There is bus speed, and latency that also plays into this.

Last edited by NutsAboutAmiga on 15-Feb-2024 at 08:51 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

The hability that pistorm has to access the chipset gives it the edge IMO regarding winuae, because despite winuae being a lot faster when it comes to emulate the chipset I didn't manage to get the smoothness I felt when using Amiga. Now, it may be to use LCD I don't know, I have quite a nice setup, and use g-sync and yet it doesn't seem to be enough.

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

geen_naam Quote:

Geez, you figured out RISC. Cracking complex instructions into multiple simpel ones. This is exactly what is happening for CISC nowadays on an execution level. But instead they call them uops. Powerpc also uses uops but often a single instruction leads to a single uop.

The definition of a uop varies but there is a difference between RISC and CISC for the types of instructions and uops executed. Breaking down powerful CISC instructions into many simple dependent RISC instructions reduces performance. CISC ISAs practically bundle multiple RISC instructions into a single instruction which often can be executed with single cycle throughput. It is entirely possible to execute powerful instructions like "add.l #100000,myscore" in a single cycle without breaking them down into uops which wastes power. CISC ISAs like the 68k often allow a single instruction to include a variable length encoded instruction allowing large immediates and displacements, complex addressing modes and reg-mem accesses. Even an old 68060 CPU core can superscalar execute multiple instructions in a single cycle using all of these advantages. The 68060 does not break down the most common instructions into simpler uop like instructions either. It has arbitrary instruction length limitations to save power so it does break up large instructions. With x86(-64) CPU cores, similar limitations were removed to turn x86(-64) cores into the memory munching monsters they are today. Yes, x86(-64) CPU cores quickly moved to uoped OoO execution but the in-order Pentium was not nearly as good as the 68060, partially due to the less orthogonal ISA. Most of the highest performance CPU cores have moved to OoO execution yet x86(-64) cores retain an advantage by executing more powerful CISC uops. CISC ISAs can have a code density advantage as well.

geen_naam Quote:

And what is the whole asm nonsense anyways. Nobody uses asm anymore in application. Except for very rare cases. I did some Analog Devices SHARC assembly once. Very readable asm (math like). My SW colleague's replaced the code with the c equivalent because of readability and maintainability reasons. And the compiler was good enough to handle fast SIMD code. So ASM readability is really a 68k relic. Probably the new "68k is superior to Zen 5 because our assembly language is better readable....." Unbelievable....

The lowest level code of every major CPU architecture is not important because you don't see it?
The assembler code generated by compilers as posted in this thread is not important? PPC CPUs received optimized code without assembler coding that allowed it to compete with x86(-64) assembler coding optimizations? Debugging below source level is no longer necessary?

Saying assembler is no longer important is like saying the atomic level of chemistry is not important because humans are more likely to alter the molecular level. Unbelievable!

geen_naam Quote:

Therefore it is amusing that people are still defending the chipset as if it is something special today when the need to bypass them was clearly there over 20 years ago.

The Amiga custom chipset has reasonable performance in FPGA using modern memory for chip memory. You attack the 68k AC/V4 hardware because of low CPU performance/clock but the PiStorm solution because of slow chipset. A 68k device could use a PiStorm for the CPU and FPGA for the chipset. We could look at the strengths and weaknesses of different solutions between PPC, PiStorm and FPGA. Agami posted the following in a different thread.

Agami Quote:

Apollo V4SA is a lot more affordable than an A1222 and it also comes with a fast 68080 + FPU + SAGA + AMMX + Maggie 3D + RTG, and an open source ApolloOS based on AROS 68k. But they would rather spend 2x - 3x on the last PowerPC machine (sans FPU) to support AmigaOS 4.

CPU integer support and performance
1. PPC
2. PiStorm
3. FPGA/V4

CPU FPU support and performance
1. PPC (A1222 drops to #2 with soft float)
2. PiStorm
3. FPGA/V4

CPU SIMD unit support and performance
1. FPGA/V4 (primitive but standard)
2. PPC (most PPC hardware used has no SIMD unit)
3. PiStorm (no 68k standard)

CPU MMU support and performance
1. PPC
2. FPGA/V4 (poorly documented MPU only)
3. PiStorm (not implemented?)

3D support and performance
1. PPC
2. PiStorm (can use with classic bus board and 3D gfx card)
3. FPGA/V4 (primitive 3D support in limited affordable FPGA)

chipset support and performance
1. FPGA/V4
2. PiStorm (classic Amiga chipset)
3. PPC (no compatibility except chipset emulation)

future hardware availability and viability
1. FPGA/V4
2. PiStorm
3. PPC

affordability
1. PiStorm
2. FPGA/V4
3. PPC

PPC still does well in CPU performance categories because an ASIC with native code is used. A FPGA CPU is less competitive than emulation for performance but the chipset in FPGA is significantly better than chipset emulation. A 68k device could use an emulated 68k CPU and FPGA chipset to reduce some disadvantages but it wouldn't solve others. PPC is worse than it looks because there isn't enough of a market at current prices and the future doesn't look good. Compatibility and affordability are more important for many Amiga users than performance. A semi-modern 68k SoC ASIC could be more competitive in all categories but too many people think the current hardware situation is good. AmigaOS 4 is not usable as is on most 68k hardware which is why parts of it are being backported instead.

@green_naam

Once again, with feeling and for the absolute avoidance of doubt. Neither AmigaOS4 or MorphOS are "next generation" since there is precisely zero difference in the sum total of their core capabilities over 3.x/68K. They do not scale to take advantage of any generational advances that have happened since the 68K, namely multiprocessor and 64-bit.

That is not to decry the incremental improvements made in peripheral features, many of which are indeed very cool, it's just setting some realistic definitions and boundaries.

It's comical to hear you condemn 3.x as ancient with no apparent sense of irony.

I ask you again. Why do you call me a hater? The only person throwing scorn and hatred here is you.

_________________
Doing stupid things for fun...

@geen_naam

Quote:

Everybody who buys a RPI to run vintage 68k stuff is an idiot

But nobody provided a way to interface x86/amd64 to native Amiga chipset yet.

For ARM this exists, like PiStorm.

And THAT is the point - emulation of the entire Amiga system vs emulation of pretty much just CPU architecture.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@geen_naam

Quote:

it reduces the vintage Amiga to dumb keyboard and mouse dongle. For the sake of calling it an Amiga.

No, not just dumb keyoboard and mouse dongle - alse the actual Amiga chipset, and _THAT_ does matter a lot, cause that's where the vast majority of the really useful and cool Amiga software is.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@geen_naam

Quote:

geen_naam wrote: @Karlos Quote: There are better operating systems for 64-bit multi core PPC and there are better targets for 32-bit, non SMP OS4 That is an opinion. Not a fact. Problem is that people like you can't tell the difference anymore.

I am not sure I am the one with the issue here.

An OS's reason for existing is to manage the resources of the hardware and make it available to applications the user wishes to run.

So yes, objectively, there are better operating systems for 64-bit multi core PPC. Linux, BSD, older iterations of OSX. They all do that job better than OS4 does.

Equally, since multi core / 64 bit is the standard for new hardware, you aren't going to find an ideal hardware match for your 32-bit single core supporting OS, but you might objectively prioritise availability and cost for a given performance level.

ARM is an objectively better target, even if you wanted to "go native", since you can run big endian 32-bit still, on hardware that's still widely available.

The only room for subjectivity here is in what you "feel" is better. There's no argument against the obdurate attachment to PowerPC because it's not rational.

Quote:

And an ICE is a (real-time) hardware debugging facility.... Allthough I cannot rule out that the term ICE has been used for multiple solution.

The In Circuit Emulators I have worked with I'm the past are pin compatible mocks for real processor/DSP that allow a running simulation of the chip in the physical target hardware. What you described as a joke just reminded me of it. Fun times.

Last edited by Karlos on 15-Feb-2024 at 09:51 PM.

_________________
Doing stupid things for fun...

@Hammer & Karlos

Nice comparison of 68k Emu performance.

Comparing the X5000 to Amiberry is giving a nearly similar performance per clock ratio, (some better some worse)

If you have the performance of a 2GHz host system it might not matter if 100MHz worth of performance is wasted on endian swapping, this might have been different on a 300MHz PowerUp card.

And because those applications and speed test are running in their own loops it doesn't matter much if there is something lost on system calls.

Applications working on old 68k filesystem might lose more, but HDFs on modern SD-Cards could still be faster than old SCSI drives.

@Karlos


Quote:

ou are correct. This is why I say that 68K is the only way forwards.

To me it seems as if 68k is the only way backwards, native code would be more efficient, than the best emulation.

There are 2 major problems when you want to want to improve something from the old Amiga:

68k: We need an Open Source OS that can be steadily improved, maybe AROS68k?
Native: We need applications that are ported to modern platforms.
---

BTW.: I'm using Amiberry but JIT is crashing, is there any advice to get it work with JIT?

@geen_naam

Context is everything. If you want to draw parallels as poor as that, I'm at a loss where to begin. Obviously a word can be used in multiple different contexts that have domain specific meanings.

But fine, if you want to call zero progress on the fundamentals "next generation", you're welcome to.

Last edited by Karlos on 15-Feb-2024 at 10:31 PM.
Last edited by Karlos on 15-Feb-2024 at 10:31 PM.

_________________
Doing stupid things for fun...

@geen_naam

to agree to not agree sounds best idea

@Thread

What about using an FPGA to swap Endian in real time..

CPU (LE)---FPGA--Amiga (BE).

Comple the OS for (LE) CPU , dont care what type.

It's just a thought, no need to burn me..

Wol.

_________________
It is my conviction that killing under the cloak of war is nothing but an act of murder.~Albert Einstein

@geen_naam

Absolutely. Ultimately we are arguing over semantics so there's always ample room for disagreement.

I do have a question though..

Suppose, just for a moment, the devil ice skated to work, disturbing a flock of flying pigs along the way and a complete 68K port of OS4, complete with support for compositing and 3D (on Pi/UAE class systems) able to run legacy 68K software directly and able to run backported PPC titles, etc.

Even less plausibly let's say all this without getting caught in some bullshit web of litigation.

How well do you think it would sell?

_________________
Doing stupid things for fun...

@geen_naam

I think we do not need the argument. I understand Karlos thinks NG as a current platform aiming to compete with modern platforms, what basically propably was the idea when developers started to develop MorphOS and AmigaOS on PPC. That obviously failed.

But of course it was a improvement compared to what existed before so you can call it next generation.

So it is a discussion about formulations and definitions.

In a sense of course PiStorm or V4 is NG too, expecially V4 with its improved chipset implementation.

so we all have NG now and can be happy with it