Anyone intrested in Amithlon type PS3 Amiga emulator?

[Trev]

@Piru

The last time I looked at Amithlon, it was running on a Celeron 333 with 128KB level 2 cache--certainly not the world's fastest processor.

It seems to me that an Amiga-like hypervisor (on top of the PS3 hypervisor) designed to run across multiple SPEs would be fast enough.

Written on top of a smart parallel API, it could be ported to both PS3 and Xbox 360.

It's really too bad Amithlon can't be separated from whatever IP made it so controversial--unless it already is. I've never looked at the internals.

[Trev]

@Piru

Last time I looked Amithlon, it didn't run on PS3. In fact all you have is UAE - without JIT.

Obviously. In terms of functionality, UAE would be just as capable. Some of Amithlon's features (native Linux video drivers) would be superfluous; others (x86 native code) would be cool (PowerPC native code with WarpUP and PowerUP compatibility).

There's nothing wrong with dreaming. ;-) As I recall, a little dream thread like this (and a lot of work, of course) spawned the Minimig.

And who'd write that? Amiga emulation doesn't multithread anyway.

I don't know, Santa Claus. There's no reason a proper design couldn't take advantage of multiple execution units to implement the asynchronous parts of the system or even offload emulation tasks like recompilation. At that point, though, you'd have a hypervisor with a platform scheduler running behind the operating system scheduler.

Funny thought: virtualization has existed in mainframes and minicomputers for decades; however, it's still near impossible to convince business system owners that virtualization is safe and effective on distributed systems, despite what VMware, Microsoft, Sun, Citrix, Novell, and others claim about adoption rates. Nevermind the fact that their applications already share resources by running on an operating system as opposed to bare metal. ;-)

[Trev]

@Piru

But in terms of performance, not.

Relative to UAE optimized for other contemporary processors, I agree.

If JIT would be implemented it really wouldn't be worth the trouble to use SPEs for the translation. SPEs have akward (read: no) access to the main system memory, and the actual translation takes relatively little CPU resources -> just use the PPE

Awkward in the same way GPUs have awkward access to main system memory. That's not necessarily a problem. Improved performance or not, a dynamic or interpreted emulation that uses the PPE to control an SPE pipeline would be a fun little toy.

Even without a PS3, porting bits of WinUAE to run on a WinFast PxVC1100 or similar add-in card would be fun.

[Trev]

E-UAE on a tight Linux distribution should work, albeit more slowly than on a modern G5.

It's been said by the man himself that the only hurdles to open sourcing Amithlon and his early work on a PPC JIT are serious involvement by a talented programmer and probable drama courtesy of Harald Frank.

Did Petunia ever make it into the wild, or has it been locked into OS4?

Aside: You have to wonder how much of the drama in the Amiga world started out with slight cultural differences, misunderstandings, and misinterpretations. Everything from sarcasm to the definition of intellectual property varies wildly from society to society.

[Trev]

I think legal confusion over who owned what and a lack of consensus on how to move forward killed everything to do with Amithlon/Umilator. No single party appears to have clearly understood the intentions of the others.