Home Page Forums MAME MAME Emulators inside MAME

We haven't done this in a while, but we support a lot more systems that could do it.

Here's the Apple II emulator STM by the legendary Jim Nitchals (RIP).



And here's some other emulator you may recognize.

Every day we keep getting closer to that one Overclocked comic.

The MAMEception Challenge:

How many different emulators (not just MAME, but MAME is preferred; let's open this up so people can have fun with really weird combinations) can you chain together in order to get back to an old version of MAME (within the first year of its creation) running Pac-Man? Ten thousand geek points for setting a new record!

(This challenge may not be entirely feasible in 2023; getting and KEEPING enough accuracy to be able to eventually run MAME 0.1 might be tougher than it sounds)

First year of MAME's creation cuts off at 0.30 or lower. The MacMAME shown above is 0.28 (September of 1997), so I'm going to call that screenshot as the low end, easy score (modern MAME emulating a Mac Quadra 605 running MacMAME 0.28 emulating Pac-Man).

It's too bad MESS had no shot at running 386+MSDOS until I think almost 2010, because making it the second to last emulator in the chain would be irresistible otherwise.

Shout out to good ol' Brad Oliver!

I am still waiting for a standardized formal language for automatically handling nested emulators (also for hardware drivers), including an AI for recognizing and cross-compiling them transparently in the manner of a JIT. For software archeology and longterm preservation such a tool will eventually become essential, when knowledge gets lost and people those wrote the only existing earlier emulators have died decades ago.

Although it currently sounds out of reach by appearing far too complex and inefficient to be automatized, remember that not long ago ChatGPT was considered sci-fi either. I think of something as strictly formalized as a Turing machine (or at least as the LaTeX scripting language), but to make it efficient enough would be undoubtedly hard. (I am not sure if Moore's law is still valid and future quantum computers may mitigate this.) A first step would be a standardized API to control emulators from outside. An algorithm (e.g. as part of the OS) should check if it can run a software natively. If not, it needs to delegate that to its "universal" emulator, which should recursively try the same by delegating it to lower layers (i.e. emulators running on emulated CPUs) until something compatible is found. In an ideal world those lower layers have to follow the same API (if not, a translation layer is needed). A major problem here may be even the 1st check for native executability, because CPU generations can differ in such subtle ways that it is hard to verify, since old optimized machine code can e.g. get messed up by pipelining in ways those crash it only randomly (happens e.g. with Win98SE 32-bit device drivers on 64-bit PC CPUs inside VirtualBox).

Word salad

"AI for recognizing and cross-compiling them transparently in the manner of a JIT. For software archeology and longterm preservation such a tool will eventually become essential, when knowledge gets lost and people those wrote the only existing earlier emulators have died decades ago."

The nutjob in alt.folklore.computers is right. We all need mindforth.

https://mind.sourceforge.net/m4thuser.html

I have to ask, has anyone ever seen =CO=Windler and Arthur T. "Mentifex" Murray in the same room at the same time?

Holy SHIT, that site is HILARIOUS.