New 68060 accelerator idea

[modrobert]

Originally Posted by alexh

It's funny you said this. One of the MiST developers has created an Archimedes core for the FPGA. He started with the AMBER for the ARM2, an open source re-creation of the ARM2. However compatibility was very bad. So what the guy did was take a cycle accurate software model of the ARM2 from an Archimedes emulator and turn it into a DPI model (that's a model which plugs into a HDL simulator). Controlled random data was loaded into both the DPI model and the HDL and the state of both compared every cycle. It fixed LOTS of bugs and the MiST core is now quite usable. I applaud this developer. Nice work.

Interesting approach, I like the idea.

I was thinking along the lines of contributing by picking any instruction, and just sit down with an oscilloscope and compare core with the original CPU. Tedious work, but if many help out, shouldn't take that long to verify each instruction. The only thing missing is the framework, an initial 68060 design, something to build on, which is a big hurdle no doubt.

Originally Posted by pandy71

Strange - people demands 68060 fastest as possible and at the same time cycle exact - where is point of this?

I agree, that makes no sense, but how about having both options available in separate designs?

For example, an open source 68060 core where the main goal is compatibility (read cycle exact) released under BSD or similar permissive license on GitHub, then you can have open or closed source forks of that core where the goal is performance. In other words, the open source 68060 core can be used as a reference, a common ground for any custom versions.

I think that current teams with their own custom CPU cores would benefit as well, and most likely return the favour by keeping the reference 68060 CPU design up to date with fixes, as it doesn't really compete with their speed oriented design.

[ppcamiga1]

'strapping a jet engine to a vw bug'

Classic amiga is realy nice, but need better procesor than 68k.
c2p takes time, rgb to ham conversion also takes time.
And this procesor need to be Big Endian.
If we will have cheap risc accelerator for classic, classic amiga may be again attractive for developers.

[ptyerman]

There already is one, it's called WinUAE!

[ppcamiga1]

Of course there is no reason to use classic amiga at all for years.
Winuae is always faster.
But if we want made classic amiga again interesting for developers, classic amiga need better procesor than 68k.
68k especialy 68060 is not worth of use underpowered, overpriced crap.

[eXeler0]

Originally Posted by ppcamiga1

Of course there is no reason to use classic amiga at all for years.
Winuae is always faster.
But if we want made classic amiga again interesting for developers, classic amiga need better procesor than 68k.
68k especialy 68060 is not worth of use underpowered, overpriced crap.

I get the feeling you're kind of missing the point here ;-)
Nostalgia is never very logical. I think a lot of devs still like to program for the 68k and those of us who like the classic Amiga use the 68k version of the OS.
There is already ppc for Amiga OS4.
Why do you think those of us who stick with the 68k rather than walk the PPC path would want a new PPC card for the classic Amigas?
For me, the Apollo-core team is doing exactly what I want from classic Amigas. A fast enough 68k CPU to run every piece of 68k software at sufficient speeds + removing some of the more annoying AGA limitations to make it more usable with modern monitors etc.

[ptyerman]

Originally Posted by ppcamiga1

Of course there is no reason to use classic amiga at all for years.
Winuae is always faster.
But if we want made classic amiga again interesting for developers, classic amiga need better procesor than 68k.
68k especialy 68060 is not worth of use underpowered, overpriced crap.

Completely missing the point. The Amiga was and is 68k, take away the 68k and it is not a classic Amiga! You obviously think 68k is crap because you have said so, why post in a thread about a 68k accelerator?
Once the new FPGA cores come to fruition it will give 68k a new lease of life, plus it is the only way forward for 68k now anyway.

[Thorham]

I'm getting a strange Amigaworld.net vibe here

[ptyerman]

Wouldn't know, don't use it. Why is there a Amigaworld vibe?

[Locutus]

^- that reply says it all

[Thorham]

Originally Posted by ptyerman

Wouldn't know, don't use it.

Better keep it that way

[ptyerman]

Why, who the hell are you? Is that why it's got like it is round here? All you Amigaworld users here now?

[Thorham]

Originally Posted by ptyerman

Why, who the hell are you?

Just someone who likes 68K.

[ptyerman]

And that gives you the right to launch personal attacks does it? You didn't even answer the question. I and possibly others too, have no idea what you were referring to about Amigaworld. It was a simple question, it would be courteous to at least instil us with the knowledge of what you were referring to?

[Thorham]

Originally Posted by ptyerman

It was a simple question, it would be courteous to at least instil us with the knowledge of what you were referring to?

I was referring to that nonsense about how 'we' need something better than 68k, etc. Crap about making Amigas attractive to developers again (as if PPC would do that ).

[MrFluffy]

I'd like a fpga replacement cpu, to give me amazing cpu speed to give my classic amiga a kick up the backside. I've got a 2000 with a 68020 in there already as a interim.

Now I'd love to get hands on, because a) I'm not completely unfamiliar with the chipset of a original amiga at low level, and b) I want to see the old girl go fast with real hardware and have it sat on my desk next to me and c) I want to toe dip in the world of fpga and I see combining the two interests (amiga and fpga) as a great starting point for that to happen.
So, the value to me is in the learning, not in having a faster but still outdated Amiga. Running UAE on whatever platform is not going to provide that.

I've been following the apollo-core effort, and I truly believe they have the skill and determination to suceed despite the odds and will deliver, but there's been so much crap thrown at their effort by other vested interests you can't even discuss it elsewhere at times without it falling into name calling, so the odds of it becoming open source is rather less than good. With the greatest of respect, thats going to give me a faster amiga, but its not going to let me get hands on with the innards of the magic happening unless there's a change of heart about releasing *ONCE* its been released as a closed product (not unlikely). I wonder if there's some keeping cards close to chest to avoid a fear of a compettitor scooping up their knowledge and releasing a higher priced alternative and claiming some aspects of the design. I know they're driven by a desire to make the end product super affordable for normal enthusiasts and I applaud that aim strongly. Plus the vampire v1 already got released under the gpl, yes I know it has bugs still but its still a huge thing for someone to release a entire 68000 core design under the gpl.

But... it doesn't serve my desire to learn being a consumer and I'm not sure I can wait for a change of heart/licensing so if there are any OPEN projects or suggestions on where to start, I'm also following with interest.

So put aside any pointless bickering which might derail things, is there any open source efforts we should be looking at?

[alexh]

Originally Posted by MrFluffy

I've been following the apollo-core effort, and I truly believe they have the skill and determination to succeed despite the odds and will deliver

I am sure they will get something eventually which will out perform all but the fastest 060 accelerator.

Originally Posted by MrFluffy

but there's been so much crap thrown at their effort by other vested interests you can't even discuss it elsewhere at times without it falling into name calling, so the odds of it becoming open source is rather less than good.

I think since the demise of NatAmi their aspirations have come down to earth and they seem to be making a concerted effort to deliver something. You cannot help but applaud such an effort.

Originally Posted by MrFluffy

With the greatest of respect, thats going to give me a faster amiga, but its not going to let me get hands on with the innards of the magic happening unless there's a change of heart about releasing *ONCE* its been released as a closed product (not unlikely).

I'm sure they just want what most commercial developers want. A reasonable ROI. They must have spent hundreds if not thousands of man hours thinking, coding and re-coding their work.

Originally Posted by MrFluffy

I wonder if there's some keeping cards close to chest to avoid a fear of a competitor scooping up their knowledge and releasing a higher priced alternative and claiming some aspects of the design.

Or an existing processor developer suing them for patent infringement?

[pandy71]

Originally Posted by modrobert

I agree, that makes no sense, but how about having both options available in separate designs?

For example, an open source 68060 core where the main goal is compatibility (read cycle exact) released under BSD or similar permissive license on GitHub, then you can have open or closed source forks of that core where the goal is performance. In other words, the open source 68060 core can be used as a reference, a common ground for any custom versions.

I think that current teams with their own custom CPU cores would benefit as well, and most likely return the favour by keeping the reference 68060 CPU design up to date with fixes, as it doesn't really compete with their speed oriented design.

68020 should be target as being most complete ISA in 68k family all before and after are reduced ISA'a implementations.
So even very fast 68060 when compared to same speed 68020 will be "worst".
FPGA 680E/LC60 with same or higher speed should be possible and believe Phoenix pursue this possibility.

[MrFluffy]

Originally Posted by alexh

I am sure they will get something eventually which will out perform all but the fastest 060 accelerator.


I think since the demise of NatAmi their aspirations have come down to earth and they seem to be making a concerted effort to deliver something. You cannot help but applaud such an effort.


I'm sure they just want what most commercial developers want. A reasonable ROI. They must have spent hundreds if not thousands of man hours thinking, coding and re-coding their work.


Or an existing processor developer suing them for patent infringement?

Alex, all of your above presume that the apollo-core team are doing it as a business enterprise to make money. Maybe I'm strongly mistaken, but I have the impression that the people undertaking it are doing so as enthusiasts not as a money making venture.
I'll ask direct and see if I'm mistaken in my understanding.

[modrobert]

Originally Posted by pandy71

68020 should be target as being most complete ISA in 68k family all before and after are reduced ISA'a implementations.
So even very fast 68060 when compared to same speed 68020 will be "worst".
FPGA 680E/LC60 with same or higher speed should be possible and believe Phoenix pursue this possibility.

Interesting, I didn't know 68060 moved towards RISC compared to previous iterations, was assuming it was the peak of CISC. Yes, then it makes sense with 68020 as an open source reference design instead.

Good to know, because the logic analyzer probes (adapter boards with sockets) I've been looking for are a lot cheaper for 68020 compared to 68040/68060, and less CPU pins.

[ptyerman]

@Thorham

Originally Posted by Thorham

I was referring to that nonsense about how 'we' need something better than 68k, etc. Crap about making Amigas attractive to developers again (as if PPC would do that ).

Ahh, I'm with you now. Yes I agree with you.