@MagicSN

Quote:

MagicSN wrote: @NutsAboutAmiga I think that new software appearing faster benefits anyone, not just me. For example to get GemRB running on 68k will take quite some work. If that game appears faster, doesn't that benefit any Amiga fans who like RPGs ?

No, it doesn't.

If I wanted to play an Infinity Engine game such as Baldur's Gate I could do it right now on one of my retro PCs.

Owners of PPC 'Amigas' might want to run PC games on their machines as some kind of justification for spending all that money on them. But many Amiga fans are not interested in turning their machines into PC clones and want to see new actual Amiga games, even if they take a bit longer to arrive.

As for PiStorm, I personally would rather see support for native ARM code. That makes sense for 'moving forward', whereas emulating PPC just perpetuates a mistake.

Not saying it shouldn't be done, but it definitely won't benefit all Amiga fans. Many of us are not at all interested in PPC, and some of us would prefer it didn't exist. It has already fractured the community and we don't want more of it!

Quote:

bhabbott wrote: If I wanted to play an Infinity Engine game such as Baldur's Gate I could do it right now on one of my retro PCs.

If I want to do XY I could do it better on my PC than on any Amiga(OID) but if we don't want to do as much as possible on our proprietary computers, we can shut them away and switch to PC.

@bhabbott

"Back in the day" a lot of the Amiga's software base was also on PC. Not just that, but as an Amiga enthusiast, if I can play a game on my Amiga where the experience isn't compromised to a point it suffers vs. playing the same game on another platform, especially if its a game I've not played before then I'm happy to play it on Amiga.
I'd hazard a guess I'm in the majority with that.
Many of these games, even on pc require emulation or other workarounds to get running as well. If I'm going to play a retro game on a retro machine then why not my favorite retro machine?
The whole "why would I play a game on Amiga when I can play it on PC" sentiment can often be inverted, but it holds as much weight whether its wos/ppc emulation or ARM native.

@all

Putting aside the who should do. should not do, could have been done, should have been done, does anything Amiga even make sense bitchfest for a moment.

What Michal Schulz has built and is building is very impressive. I'd love to see more people creating stuff for the Amiga.

Hans

_________________
Join the Kea Campus - upgrade your skills; support my work; enjoy the Amiga corner.
https://keasigmadelta.com/ - see more of my work

@OneTimer1

Yes I am always surprised if amiga user start to talk about what makes sense rational

@bhabbott

Michal wants that WarpUP software works on it. Does it really make sense on a commercial scale? No. It is simply because he wants to do it. His time, his decision.

I think this is a great idea, it brings the possibility of bring the PPC users back to a single hardware platform. PPC was a good idea at the time but it's fallen way behind, so now it's expensive and slow.

I'm just wondering what CPU you need for the PPC emulation to exceed the speed of the PPC hardware.

The choice is now x86 or Arm, but Arm is pulling ahead in single threaded performance, that's specifically Apple at the moment but early benchmarks of Qualcomm's X2 Elite are not far behind.

_________________
Whyzzat?

@bhabbott

Quote:

As for PiStorm, I personally would rather see support for native ARM code. That makes sense for 'moving forward'

Reading Michal Schulz's blog that Hammer linked, he seems to be debating what sort of Arm to support. I don't know why though, M series (microcontrollers) are not likely to be very interesting, R series are probably too exotic. The only one's interesting are the A series and given the latest ones have dropped 32 bit and Thumb2, the only one left is Aarch 64, and probably little endian because that's what everyone else is doing.

_________________
Whyzzat?

@minator

Quote:

I'm just wondering what CPU you need for the PPC emulation to exceed the speed of the PPC hardware.

QEMU on my laptop (AMD Ryzen 7 8845HS 3.8 GHz) gives performance a little bit above G4 1266 MHz in Pegasos 2 I have on my desk (except AltiVec of course). Not bad.

@minator

And that alone would mean that there is no chance to let ZZ9000 benefit from that.

What I definitely try to avoid is a situation where you have Emu68-only Amiga binaries.

@michalsc

Quote:

What I definitely try to avoid is a situation where you have Emu68-only Amiga binaries.

there are already few WarpOS programs that only works in special case, like only on with 15bit graphic modes. or some where stuck between hardware banging, and next generation hardware.

Last edited by NutsAboutAmiga on 30-Oct-2025 at 05:11 PM.
Last edited by NutsAboutAmiga on 30-Oct-2025 at 05:09 PM.

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

@NutsAboutAmiga

There is, however, a difference - on one hand some badly written software requiring e.g. a 15bpp mode and failing otherwise; on the other hand creating new ABI which is completely foreign for Amiga in every form, leading to a situation where you could see in future a requester saying "this software requires Emu68 version x.y. to run".

@minator

Quote:

minator wrote: I'm just wondering what CPU you need for the PPC emulation to exceed the speed of the PPC hardware.

I don't think a PPC emulation on a RasPi can beat an old A1/Pegasos, but it might be a good replacement for Phase5 PowerUp cards, especially with RTG.

@minator

Quote:

minator wrote: ... he seems to be debating what sort of Arm to support.

I believe this depends also on the boards he could use, It doesn't matter what CPU you have selected, if you have to build your own motherboard it will end in a price range we knew from PPC.

And if those systems are to big to be integrated into an existing Amiga your should try and switch to a stand alone system ... with or without a ECS/AGA emulation in soft- or more expensive in hardware.

---

@bhabbott

Quote:

If I wanted to play an Infinity Engine game such as Baldur's Gate I could do it right now on one of my retro PCs. Owners of PPC 'Amigas' might want to run PC games on their machines as some kind of justification for spending all that money on them. But many Amiga fans are not interested in turning their machines into PC clones and want to see new actual Amiga games, even if they take a bit longer to arrive. As for PiStorm, I personally would rather see support for native ARM code. That makes sense for 'moving forward', whereas emulating PPC just perpetuates a mistake.

PowerPC was IBM's vision for replacing the Intel x86-powered PC, hence why there's "PC" in PowerPC.

IBM attempted to replace the 16-bit 286 PC with a 32-bit RISC-based RT PC (RISC Technology Personal Computer) in 1986.

IBM attempted to replace many clock cycle CISC x86 PC CPUs with hardwired RISC-based PowerPC.

IBM attempted to replace the IA-32 (32-bit x86) PC with the 64-bit PowerPC 970.

Something to consider about IBM vs US DOJ's antitrust investigation and unbundled software (e.g. DOS) with IBM PC https://truthonthemarket.com/2020/02/03/the-ghosts-of-antitrust-past-part-2-ibm/

_________________

@OneTimer1

Quote:

OneTimer1 wrote: I don't think a PPC emulation on a RasPi can beat an old A1/Pegasos, but it might be a good replacement for Phase5 PowerUp cards, especially with RTG.

The work in progress 3600 mips PowerPC emulation (EmuPPC) by RPi 4's ARM Cortex A72 would beat some old A1/Pegasos, especially on Ativec 128-bit SIMD (to AArch64 NEON 128-bit SIMD) and superior memory bandwidth.

PPC G3 (7xx), PPC e500 V2, PPC 440/460, and PPC e5500 don't have 128-bit SIMD.

On the same RPi hardware, EmuPPC is already faster than Emu68 on early benchmarks.

From https://www.patreon.com/posts/one-more-thing-141985279
Rendering a Buddhabrot fractal takes 14.6 seconds on PowerPC, compared to 22.9 seconds on M68k. Other tests, such as Dhrystone 2.1, confirm a roughly 2× performance increase.

When compared to complex 68K, PowerPC's RISC code stream is already optimized and clean for other RISC-based CPUs like AArch64.

The Amiga can have two virtual processors ISAs (one CISC, one RISC), hence the Amiga can float on any CPU platform. The Amiga doesn't need the Tao Group's Virtual Processor IP.

Last edited by Hammer on 31-Oct-2025 at 02:41 AM.
Last edited by Hammer on 31-Oct-2025 at 12:55 AM.
Last edited by Hammer on 31-Oct-2025 at 12:46 AM.

_________________

@pavlor

On the X86, QEMU's PPC module is not the fastest PPC emulator when compared to RPCS3.

RPCS3's PPC emulation doesn't emulate the low-level PPC MMU, RPCS3 instead uses high-level emulation methods to translate system calls e.g. RPCS3 intercepts the sys_memory_allocate function and returns a valid pointer from the host computer's memory. RPCS3 leverages the host operating system's own memory management functions, which are optimized for the PC hardware. RPCS3's high-level PPC MMU emulation is closer to native performance.

Last edited by Hammer on 31-Oct-2025 at 01:13 AM.

_________________

Where’s PiStorm for Quantum computers? Much better performance!

_________________
Sent from my Quantum Computer.

Read his notes all: https://www.patreon.com/posts/one-more-thing-141985279

Why not use one unused core for this ? Especially when you read he got such fast results with his experiments...

"he idea was born a few years ago and kept coming back to me. The reasoning is simple: it’s a RISC, I have a RISC -- they might even be similar. Then there are the manuals I’ve read over and over again, every time thinking, “Yeah, that might be doable.�

So it happened.

I took the core parts of Emu68: the translator, caches, and main JIT loop, and started to experiment.

First opcode… easy.

Second one… whoa, still easy.

…

About a week before Amiga40, I already had around 80–90% of all integer instructions running.

That was way too easy. The entire JIT -- translator, main loop, and helper functions -- fit within roughly 7,000 lines of code. Yes, I wrote that mess in a single C++ file. Now I’m paying the price for that and cleaning it up. My fault, my extra work. :)

Once everything was stitched together, I wrote a small test tool.

Nothing spectacular -- it just draws a color pattern on the screen.

But since I’ve been doing this for a long time, counting every executed PPC opcode and measuring time, I decided to benchmark it.

Hmm… not bad for a first attempt. About 1400 MIPS (not a benchmark score, just total instructions divided by runtime) with an average performance of roughly 0.7 PPC instructions executed per AArch64 CPU cycle.

Wait -- that’s with data being written pixel-by-pixel to a 16bpp screen without data cache.

So I reworked the test to write the same garbage into cached memory, still using 16-bit writes.

Then I started the test.
BOOOM

The numbers looked right — but somehow didn’t feel right."

Last edited by kriz on 31-Oct-2025 at 05:54 AM.

@minator

there is no such possibility as "bring the PPC users back to a single hardware platform"
PPC users already have decent x86/arm hardware running windows/android
there are no reasons to switch to emulator either 68k or ppc
if you want to switch to x86/arm you have to made aros
no more than 15 years behind windows/android
at least win7 level
so start working on aros