Opinion: PPC, 68080 and other

[Bruce Abbott]

Originally Posted by Promilus

Emulation is much older than software emulators. It doesn't have to be software in the first place. If I take one chip and make it so it works (almost) exactly like other chip it might be an emulation as well.

In the old days there were CPU 'emulators' that plugged into the CPU socket for debugging. Typically they would be the actual CPU along with extra circuitry that allowed single stepping, viewing register contents etc. Another popular device was the EPROM 'emulator', which had static RAM that could be loaded from a host. But that is a different usage of the word that doesn't relate to software emulators.

The A1000 had a system ROM 'emulator', but it was called the Writable Control Store - a more accurate name. I'm betting they had something like it in the prototype machine, and then EPROMs (if fitted instead) were 'emulating' it!

Nobody says a flash ROM 'emulates' an OTP EPROM or mask ROM - it's just a similar device that does the same job. Similarly, the Vampire doesn't 'emulate' a 68k CPU and AGA chipset. It's compatible, in the same way that eg. PC VGA cards were compatible with the original IBM VGA (Video Gate Array), and the Cyrix 486SLC was compatible with Intel's 386 and 486 CPUs.

[pix]

Originally Posted by Bruce Abbott

it's just a similar device that does the same job. Similarly, the Vampire doesn't 'emulate' a 68k CPU and AGA chipset. It's compatible, in the same way that eg. PC VGA cards were compatible with the original IBM VGA (Video Gate Array), and the Cyrix 486SLC was compatible with Intel's 386 and 486 CPUs.

Thanks, I believe yours is the correct view of the term emulator in the context of computing and software. I do appreciate the overall point of the other views though that cloned hardware is rarely the "complete" original.

I reiterate that I absolutely LOVE emulation and I do have PiMiga on my Pi400 - that form factor combined with booting straight into workbench is almost the best you could hope for from an emulator.

However, as you stated, the illusion is easily and frequently broken, often ending up back in the host Linux, or freezing altogether. Emulation can be incredibly stable of course, but the bottom line is you can mess with it. Add to this that there is no illusion whatsoever when you are booting the hosst system and the experience is a firm second best. A compatible device will boot in seconds and provides the exact resources needed for the experience, even if there are some elements of the original missing, or modified, the bare-metal is real and tangible

[pix]

and that's really related to my original point, that with emulation you can boost that emulated performance to level unheard of on the original machine. It would be nice if the new hardware could exceed the original sufficiently to offer the possibility of new applications. I know the PPC does this already, but I think the PPC without the native Amiga hardware is missing a trick.

Kiss and make up. Integrate the 68080 into the new PPC boards, boost the speed as much as feasible and you will have an Amiga that offers everything Amiga ever did and more.

Wishful thinking I guess - but it would surely be better for the Amiga than dividing up an already niche market IMO

[Chucky]

Originally Posted by Bruce Abbott

There is no computer program running inside AGA Amigas that make the chipset look like ECS - it's all actual hardware.

It doesnt have to be a program running to be Emulation.
however in the case of ECS / AGA. they have taken the EXACT gates of the ECS and did addons..

where they in say. SAGA etc have taken assumptions how the gates are configured and added extras to that.

so SAGA is extension on a emulated platform. (Hardware emulation in FPGA)

AGA is an addon on a existing implementation in ASIC.

[Don_Adan]

Originally Posted by Chucky

It doesnt have to be a program running to be Emulation.
however in the case of ECS / AGA. they have taken the EXACT gates of the ECS and did addons..

where they in say. SAGA etc have taken assumptions how the gates are configured and added extras to that.

so SAGA is extension on a emulated platform. (Hardware emulation in FPGA)

AGA is an addon on a existing implementation in ASIC.

But gates not must be set identical. For what?
Same effects can occured if some gates are changed.
Similar for code.
You can create totally different code which make same effect.

[Chucky]

you can make totally different code that have the same effect. however. maybe someone is using that code outside the specs.... and then the "different code with the same effect" suddenlt will not work..

[Promilus]

Originally Posted by Chucky

It doesnt have to be a program running to be Emulation.
however in the case of ECS / AGA. they have taken the EXACT gates of the ECS and did addons..

where they in say. SAGA etc have taken assumptions how the gates are configured and added extras to that.

so SAGA is extension on a emulated platform. (Hardware emulation in FPGA)

AGA is an addon on a existing implementation in ASIC.

They did not. They did take hdl of ECS, written some additional features and then "compiled it" to get totally new and unique configuration of gates. Nobody did take blueprints of ECS and inserted in free space new gates for AGA functionality. That's total BS! Gates - that's hardware implementation of logic. You design logic then implement it in hardware. You do not make additions on hardware level, you add new logic and then do yet another implementation of hardware. The new chip is different from the first one. Can be slightly different, can be entirely different. It doesn't matter. SAGA is compatible with AGA register-level and some signal timings. It does not emulate AGA any more than AGA emulates ECS.

[Chucky]

ok they took tye original data. the original schematics/ideas and did it.

SAGA etc is based on what they GUESS it looks like.

so it mimics the hardware. it is not based on the original idea.. it is based on a guess how it is done. nothing more..

but people have this fear about "emulation" and weird ideas about it.. an emulation can be pretty damn good. but it can also be pretty damn BAD.. doesnt matter if emulation is done in software or hardware. but there are people that thinks that as long it is made in fpga it is a perfect IMPLEMENTATION. while noi. in cases a fpga solution can be worse then the software version. it all depends on how good the the designer can guess the original design,

[Promilus]

@Chucky
AAA was supposed to be register level compatible with ECS while being entirely new chipset and doing things entirely different way. SAGA is register level compatible with AGA doing some things similar way (so maybe the same way, maybe different way but with the same end result) and some other things different way (this apply to both SAGA specific functionality which ain't just few additional config bits - like Maggie - and everything SAGA does not up to AGA specs which results in occasional glitches). In this form SAGA does not emulate AGA, it implements own subsystem which is at some level compatible with ECS and AGA. But doing stuff the other way around doesn't make "emulation" either. Cyrix did make x87 coprocessors own way and it did NOT emulate i387, bah, it was FASTER than i387! But it was coprocessor which was pin compatible with i387, it was also ISA compatible. It was built differently and behave differently, but end result (computation results) were the same. How would you describe that? Emulation? Why? IBM (cyrix actually but manufactured by ibm and licensed as well) 6x86 did use Pentium Socket. It was built different way and executed code different way. Was it Pentium emulation? No. Why would it be? SAGA ain't no different. It does perform similar job, some things work differently, it's built differently. Performs better in some cases and has new features of it's own. Same with AC68080 - all it does emulate (if anything) is 68k physical bus. Exactly the same thing as any 040/060 does on A1200 or 020+ does on Amiga500/600 with the use of "glue logic" which is essentially smaller predecessor of FPGA in form of PAL or GAL PLD chips.