I don't know whether there is a point in posting news about new PPC hardware on this site anymore, but anyway, here it is:

Link: OSW on sale next week

(well, when it comes to the "next week" part, I must say I'll believe it when I see it )

Why not posting such items? As long as you don't add "and MorphOS will be available for it because I decided it." :-)

But I believe we now learnt being on sale does not always mean being available (or even existing).

Quote:

Good point!

Quote:

Good point again!

This is much more interesting than Efika. Of course, the problem is that MOS can never use dual CPU in a transparent way, without breaking compatibility. So any MOS port would either
a) use very little of the available power
b) have to rely on some crappy "co-processor" api
c) break compatibility (and take ages to develop, probably)

But it will be interesting to see nevertheless.

I think MOS could be adapted to multi-processor systems, without breaking compatibility. For example, in windows, there are ways to tell the system some aplications need to be run in only one of the processors available, so we can do something like that.

|||

[ Edited by Velcro_SP 17.09.2011 - 10:14 ]

i think, both, efika and osw are interesting, and when talking about multiprocessing under morphos... why should a programmer take care of it? imo, os should take care of it... programmer can just use more processes at once, that will make it much faster on multiple processors... but the morphos core must support SMP first.

bye, MarK.

A more recent photo would be welcome

If MorphOS will one day support it, it would be a worthy way forward

Quote:

MorphOS itself supports SMP, however you're probably referring to abox, which puts SMP in the same category as full MP .. it's simply not possible without breaking everything.

IE, you could run MorphOS on an SMP system, but abox would have to be alotted one CPU only...


- CISC

Quote:

let's break everything then.
who's still interested in the ABox nowadays ?

now i'm half kidding

Quote:

Technically, the "A-box" is a Quark application/process, right? It would have been cool if the A-box could be run multithreaded and the threads being spread out on cores by the kernel ? la QNX (that is, with the application not even knowing about this, and even less having to care about it) ...

Quote:

Or 4 simultaneous a-boxes?

Quote:

Uhm, do you even know what a box (p un intended) is? ;)

Quote:

Maybe it's just me, but I totally fail to see the point in that (not to mention the confusion it would lead to). :P


- CISC

Quote:


Well, it may seem akward at first, but - hypothetically speaking - running a Blender render client in each of four a-boxes is surely no more awkward of a concept than running render clients on four different computers which form a small render farm. It definetely saves physical space and energy

If switching between different a-boxes was possible without too much effort by the user, this would be definetely an advantage for various performance-intense applications.

Even on my Intel-based workstation, I often end up running specific tasks on one core and use the other for general purpose tasks. The amount of software which can fully exploit multiple CPU cores is very limited upto this day. The speed difference is quite often rather disappointing.

[ Edited by JoBBo on 2006/10/30 22:15 ]

Actually I agree. People are switching workspaces all the time. Why not switch complete instance of OS? MOS is tight enough that the extra bloat wouldn't make it completely impractical. Optimal, no.. but fun, yes.

Then of course these OS instances need some sort of filesharing.

Well.. this thing has to be available first anyway. :)

We can always put abox in a fixed processor, and use qbox in the others... One processor is enough for abox, I think. I know Qbox is not used yet, but its just dreaming awake, when everything works together.

Quote:

Then roll out 1.5 (matured ABox) and start 2.0 with QBox.

I just hope that you'll do a miracle with this card when it will be avaible

There shouldn't be a big problem in creating an API that allows starting Quark-tasks from ABox. The result would be something like PuP (but hopefully done better) and offloading heavy calculating to another core would be possible.

Quote:
"too much work"...

Leo.

Stop whining!

I cannot foresee a better way of comparing two systems... (namely Linux vs MorphOS) perhaps it could be a good way to MOS to shine...

Quote:
I disagree, such feature would be quite useful. You could have stabile stuff running on one, and betatest your own unstabile software on another. In case of crashes, that would make "total reboot" unnecessary in many cases.

Of course there would probably be problems. For example, what would happen if two (or more) programs from separate environments want to use 3D gfx card simultaneously? Well, I never said it would be easy - only that it might be quite useful, if it was successfully done

Quote:

We need a supervisor program running in ABox.

This sort of "SUPER-SCOUT", or "Super-Snoopium" should control any hardware request by any ABox running into its own Core.

It should also prevent as a sort of "Magic Grim-Reaper of MorphOS" that programs born for 68000 exceed their allocated memory.

Should be like this:

Superscout running into QBOX

Superscout assign Core for each ABOX running

Superscout controls hardware requests by any Abox

Superscout allocates fixed memory for any 68xxx program in any Abox.

------------

And there is another task that SuperScout program could perform.

I ask programmers if this scenario it is possible.

Here it is:

Superscout also allocates a certain quantity of memory as a Memory security sandbox in which detour unwanted memory allocations by 68xxx programs.

If a program 68xxx try to hit allocated memory, then the Superscout divert this request to the free sandbox of allocated special memory.

So the 68xxx program will be free fo run, but without causing problems to already allocated memory, and the system get more and more stable.

Then I ask you all:

Is it possible to run such a Superscout program with all these super powers?

This is getting weird. How about simply using freescale's 8641 CPU, which only has one core? If I remember correctly, there were talks about using this chip even before the Pegasos 2 was born.

Quote:

What you are asking for is running two ABox threads. While interesting it is not realistic.