UltimatePPC and everything else

[Frags]

True but in this case I think it was simply an erroneous decision that for whatever reason has not been acknowledged or amended (yet).  Hopefuly it will because it would be a shame to see someone with the time and inclination to build this kind of hardware produce something which is effectively useless when it could be killer.

[psxphill]

your comment that I dont need access to the kernel to write a driver, is exactly my issue - being told what I can and can't do. I won't want to buy into software to run as my hobby machine, that is locked down to a single piece of hardware, doesnt give me the freedom to work on parts that I want. Maybe I want to work on the core os. Or maybe I don't but I would like the ability to be there. right now the "morphos" team have the power to pull the plug on the os, or refuse to replacea keyfile if hardware was to die. No thanks.

Buy an UltimatePPC and get AROS running on it then.

[mousehouse]

BTW is Sam Jordan aware that the chosen CPU has an incompatible FPU?

Yes he is. I understand from him that in WarpOS there actually is some abstraction of the FPU code and that we can trap all FPU instructions. It is something we can work on to get this running.

While I understand the skeptical remarks it is the same in all projects. Trade-offs are always part of it - some turn out great, others don't. We believe we can make this work and want to offer a nice package to the Amiga community.

With regards to updates from our side, an often made remark... my current plan is to do a monthly update from our side regarding progress made. I'm really reading all posts on the UltimatePPC thread here, and at AmigaScene.nl and whenever I see something that strikes me as valuable input I discuss it with Gideon.

We showed the board at the last Dutch Amiga meeting in Maarssen on June 17th, and are working very hard to get it to a stage where we can demo the card to an audience. Please keep in mind this is something we do in what used to be the scarce spare time we had!

Next update is planned end of this month...

[Frags]

Wow that`s great!  If WOS runs then that makes a big difference.

People will respect the likes of Harry Sintonen when he says it`s a dead end, that doesn`t mean we want it to fail...

[Piru]

we can trap all FPU instructions

That should work, assuming the other differences can be handled in a similar manner. It remains to be seen what kind of performance impact it might have. WarpUP-WarpOS/Demos/cybermand will be the judge of that I guess.

As I recall it for some reason WOS exposes the supervisor side of the CPU to any apps though (Super() call). Hopefully that isn't used much.

[buzz]

Buy an UltimatePPC and get AROS running on it then.

that sounds like an expensive way to run aros.

you know there is a difference between me having the freedom to do something, and you telling me to do something right ?

I already have plenty of projects anyway, but thanks.

(I may well buy one of these devices, and I certainly wish them the best of luck with the project)

[itix]

Itix' remark hints at yet another conspiracy theory amongst Morphos developers - I assure you I have zero knowledge about OS4's compositing engine and how it came into existance.

Sorry, my fault this time. It was an educated guess based on a hearsay from a beta tester.

[wawrzon]

Yes he is. I understand from him that in WarpOS there actually is some abstraction of the FPU code and that we can trap all FPU instructions. It is something we can work on to get this running.

fine, but the downside of warpos is and ever was that it needs natively compiled binaries to take advantage of it. everything else remains as slow as it was without it. i repeat my proposal ive already made before: 68k emu running transparently on ppc to speed just everything up. the advantage is, you do not create another niche within a niche, that demands out of date, rarely available and practically unsupported compiler targets backends and particular coding. on the contrary, plain 68k code executed sorta atop of abstraction layer (68k emu) allows easily even further migration to future faster extension cards and exchangeability between all amiga-ng solutions.
what do you say?

[itix]

@thread

that sounds like an expensive way to run aros.

That is another problem. It only has averagely performant 68040 to run Workbench but it is relatively expensive toy to run AROS/PPC. AROS/PPC on the other hand is not designed for 68k compatibility and Janus UAE solutions are slower than 68040 on that board.

If it ever runs WarpUp kernel on that board but PPC (PUP+WUP) software library is small. And before someone is going to get his hopes up for new software I should remind that both PUP and WUP kernels implement only bare minimum (Ralph might kill me for that, tho...). There are all those cache flush and context switch issues and developer support ceased to exist aeons ago.

Software support for this board is more than problematic and complex.

[wawrzon]

Buy an UltimatePPC and get AROS running on it then.

as aros is open source everybody is free to do that, but i wouldnt put effort into another fork just to cover a few potential users. aros supports 68k for a reason (except my particular interests) namely it lets control its backwards compatibility. similarly currently there is an effort to get sams supported. but yet another spacial case might become way too much.

@itix: +1000...000!!!

[Tahoe]

(the remaining motherboard connectors are rare)

Sorry, I would have to correct you on this bit... The connectors are still being manufactured by Kel. You can buy them in minimum quantities of 220 pieces. They have already been sourced.

As far as changing the design: so far development seems to be on track, with support for both WarpUp and AmigaOS4 firmly rooted, I personally see no reason why they need to change anything. If somewhere along the line a problem occurs which they cannot overcome they can cross that bridge when they get there I suppose.

If some applications do not run because they rely heavily on the FPU and can't be trapped, maybe the author(s) are willing to adapt the code, or open source it for someone else to help out.

[OlafS3]

Quick correction... Aros 68k is not only needed for testing backward compatibility, it is also used for 68k emulation (in X86/ARM...) and as bridge to the 68k community (the biggest amiga-usergroup)

[wawrzon]

Sorry, I would have to correct you on this bit... The connectors are still being manufactured by Kel. You can buy them in minimum quantities of 220 pieces. They have already been sourced.

didnt know that, this is certainly a good news.

as for warpos, as an owner of csppc i really feel its a dead solution, lets see if i be proven wrong.

but how will os4 work around the fpu incompatibility?

[Frags]

fine, but the downside of warpos is and ever was that it needs natively compiled binaries to take advantage of it. everything else remains as slow as it was without it. i repeat my proposal ive already made before: 68k emu running transparently on ppc to speed just everything up. the advantage is, you do not create another niche within a niche, that demands out of date, rarely available and practically unsupported compiler targets backends and particular coding. on the contrary, plain 68k code executed sorta atop of abstraction layer (68k emu) allows easily even further migration to future faster extension cards and exchangeability between all amiga-ng solutions.
what do you say?

At least we have a few WOS apps though, if it runs.  A CS replacement that goes fast would be saleable I`m sure but given Petro`s offer a Blizzard replacement might now do better.

[wawrzon]

At least we have a few WOS apps though, if it runs.

yes, they are really a few, id say. compare it to a situation when every available application could take advantage of the speedup..

A CS replacement that goes fast would be saleable I`m sure but given Petro`s offer a Blizzard replacement might now do better.

unregarded any offer there is certainly more a1k2 than a4k in use.

[Frags]

Well if we`re talking 68k then I agree that a silly-fast x86 with jit and glue in fpga as an accelerator would be amazing.  In fact it`s a bit of a dream of mine.  I even posited it on natami.org but of course I can offer little to help so it`s just a dream.

edit: well, in ROM or whatever, you know what I mean...