[Info] SNES9x Fehlermeldungen

[Andi]

0

Bei Amazon gibt es inzwischen auch wirklich viele negative Kommentare. Na ja, bisher klappt's...

@Glurak:

Eigentlich emuliert Higan den SNES nicht zu 100%. Byuu beschreibt es so:

higan does not have 100% accuracy on the SNES. It has 100% known compatibility with officially licensed software.

To elaborate: compatibility means no bugs that are observable. Anything that can be demonstrated to only happen under emulation, and never on hardware, would count as a bug: distorted audio, input not responding, one bad scanline, a missing effect, etc.

The known qualifier is very important as well. Nobody can play through every possible code path of every single game in the library. We've had a few people now play through the entire library for several minutes per game, but even still, I'd say once a year or so, someone spots a new bug. The way I maintain the 100% compatibility is by fixing new bugs quickly, instead of ignoring them.

[...]

I can name a dozen obscure hardware features/quirks that nobody's been able to reverse engineer yet on the SNES. In fact, I can't name one single processor in all of bsnes that's emulated 100% accurately. Not even the "simple" real-time clocks (I once spent 2-3 weeks meticulously testing and recording every BCD incrementing error on invalid time values for the Epson RTC, but I was only scratching the surface ... it has a whole world of side-effects if you send certain commands without waiting long enough for the chip to process them.)

[...]

Honestly, even with all of the issues listed above, we've only scratched the surface of accurate emulation. Take the case of DICE, the digital integrated circuit emulator. Here is an emulator that works at the transistor level for absolutely perfect recreation of the very first video games ever created. To run Pong at about 5-10fps, DICE requires a 3GHz processor. Yes, you read that right: no computer processor at this time (2011) that can run Pong at the circuit level at full speed. It's not that DICE is a slow program; indeed, it is very well optimized. It's that there is enormous overhead to simulating every last transitor propagation delay.

Dann schreibt er noch etwas, das auch nicht uninteressant ist:

yes, ZSNES is written in x86 assembler. They've ported ~30% to C, but that's mostly the GUI, path handling code, compression libraries, etc. All of the core emulation is in x86, and will be dead as soon as Windows drops their 32-bit backward compatibility (can't happen soon enough. Syswow64 is a total mess. I suspect it'll happen in about ~10 years from now.)
higan's not only portable C++, but the core doesn't even really use the standard library. Even opening files is abstracted through the UI.
Now, it's possible that in 20 years, C++ will be like COBOL or Fortran is today. Nothing I can really do about that. Someone will have to port it, and they'll likely make mistakes in the porting. And those mistakes will probably keep amplifying every time it's ported to a new language, with the original hardware long-gone to compare it to.
All of us now will likely be dead before it happens, but I suspect higan will vanish completely in ~100-200 years or so. Or if another emulator comes along and surpasses it, then probably in ~20-40 years or so.

Quelle: https://www.reddit.com/r/emulation/comme..._accuracy/

Passend dazu: http://arstechnica.com/gaming/2011/08/ac...-emulator/

Alles sehr interessant.