Does Hyperion own the rights to OS 3.1, 3.5, 3.9?

[Plaz]

Coldfire is not compatible with the Amiga.  The End.

And if anyone wants to know why it won't work.... I'm done my homework on the subject. There's a small number of instructions that exist on both 68K and Coldfire, but each executes them differently. It's possible to trap 68K instructions that are invalid for coldfire and "reroute" them. Using this trick make the two about 99.8% compatible. Trouble is these other instructions are not invalid, can't be trapped and will cash Amiga code because of the way they operate. Also there's an issue with I/O differences.

One suggestion I had for the coldfire dev group back in the day was to run dual processors. One to prescreen code, make changes if necessary then feed the corrected code to the second processor. Back then coldfires were not very fast so basically doing JIT compiling would have killed performance.

These days I'm thinking porting one of the Amiga like OS's to cheap dual core 1GHZ ARM processors would be a fun project. Picture this... Core_1 doing JIT recompiling and feeding Core_2 that's running a "softcore 68K" emulation. Piece of cake, I should have all that done in what.... two weeks?  


Plaz

[Plaz]

Coldfire has been demonstrated to work (I remember someone putting a video of the prototype being demoed here on AO).

Oh yes, it will work. But eventually it will run into a bit of 1.x-3.x OS or legacy game code with one of these miss interpeded op codes and crash something. We should I suppose distinguish "working" from "completely stable" then.

Plaz

[Plaz]

I think the Atari Coldfire developers will disagree with you.  

...[And who said 99.8% compatibility isn't "good enough"?  

They don't have to disagree with me. They just need to get a work around for those problem op codes. Perhaps classic Atari avoided them from the begining better than Amiga?

In many cases 99.8 might be good enough, but from my reading, a particular set of these codes are used in classic AOS (kickstart too.. can't recall now). There's probably ways around all this by rewritting bits of important libraries and softbooting to a problem free kickstart. But none so far have been able to work out all the bugs.

I'm hardly a detractor either. Years ago, I was really on the coldfire band wagon. Reseaching the CPU, working up hardware and I/O layouts, looking at ways to trap those trouble codes... but now with other options like soft cores, FPGA tech, and newer processors I don't know if it makes a much sense as it once did.

And why couldn't UAE be ported to run on a Coldfire processor?  That's a rhetorical question......

I think it could since you have alot of control over rewritting all the peices involved unlike constraints that have to be overcome in making a plug in compatible with  classic hardware.

Plaz

Edit: botched quote

[Plaz]

You know, it would be really nice to just clean-room reimplement

That's why minimig and minimi AGA are such cool thing and I hopw they see a lot more development.

Plaz

[Plaz]

by the 040 they were so great that specific libraries had to be included in AmigaOS to deal with the compatibility issues.

If coldfire could have been solved that easily, it already would have been. Unfortunately it ran a bit deeper.

Plaz

[Plaz]

Amigas, we should enjoy using them in the same way people enjoy classic cars.

As one who owns Amigas and is restoring a classic car, I heartily agree. However a similar comparison can be drawn. As I restore my car, I'm not going to be using 40-50 year old technology to rebuild the brakes, fuel system, suspension, electrical, transmission.... most of us in the car hobby are using the latest tech to improve reliability, handling and saftey. What's under the hood won't be bone factory stock, but it will look and feel classic and will be better than original by miles. Classic Amigans hope for the same on their systems

Plaz

[Plaz]

Who said it was easy? Regardless - Coldfire has been shown to work already. See the link I gave. Dragon was actually demonstrated live at that Amiga meeting.

I'd have to guess it was demonstrated on a cafefully selected set of task to avoid known problems. It is too bad that a card wasn't at least released as a dev unit so some of us could have played around and mapped out the good and bad of it all.

Plaz