Quote:
Originally Posted by Chucky
THEN I would actually use the word "implementing" as when you have the original data and does it. it will behave exactly as the original thing.
|
Actually to mess with all that discussion further.
AMD do not emulate intel. It uses the same ISA. It doesn't use the same insides, nor claims to have something exactly like that. Bah, first AMD x86 were basically clones of intel made through intel own documentation. So different manufacturer of the very same design. After a while they took different routes. One might say AMD 486-Pentium era processors did emulate intel front side bus and that might be a valid argument. But the only x86 processor which actually did emulate intel ISA was transmeta crusoe (which was actually a VLIW and did translate on-the-fly to it's own, internal ISA)
Now then, we already know what emulation is - when one piece of hw or sw tries to act like a different piece of hw or sw. But it ain't really that simple.
There was something like 6502. That's MOS processor which debuted with Commodore PET and was later used (with variation) in C16/116, C+/4, C64, C128 etc.
Now then - if I hook up logic analyzer and grab readings I can describe how external interface works. I can also design logic which will process original 6502 ISA. And that logic is ISA implementation. But external interface and timings are emulation since I don't exactly know how it originally was achieved, only what final result is.
But... there's more to it. There actually is enough documentation to make per-transistor exact copy. And it's done! (but in big scale using discreet components
https://monster6502.com/). So when you know how things are connected inside and you make something based on that design - that's reimplementation (hardware level).
There's also complete verilog netlist obtained through scanning of original chip. Problem with that is when you synthesize that logic - you are not bound to achieve per-transistor level accuracy as it depends heavily of synthesis options and tools versions. But it would still be reimplementation because it's based on original design.
Why Vampire is different? Well afaik apollo team doesn't have original netlist (how logic was implemented in original hardware) nor exact logic design. All they have is how it's supposed to work (some bits and pieces which works... most of the time, but not 100% accurate). That's why it cannot be
reimplementation of AGA. That part of the SAGA which is used to make it work with AGA titles - well that's emulation. Everything beyond, new capabilities, RTG - that's implementation of new features.
AC68080 is also a different matter because it's not based on any other design, it's just ISA compatible - just like any x86 is compatible through ISA with 8086. . So it's implementation of original - apollo team's design, no emulation (apart from 68000/020 bus emulation of V2 necessary to make it work on A500, A600, A1200 mainboard). Minimig, MiST, Vampire - all of those share similarities and indeed use - some - emulation of original hardware. Raspberry or PC use software emulation of ALL original hardware. Neither of those solutions are bad, RPi is way cheaper and achieves better results cpu-wise. So it's up to the user which one prefers.