@persia

programming a factory robot to paint the word Picasso on a car on the assembly line can be done on any x86 embedded system using C, or even visual basic.
When the Honda Asimo robot reaches a flight of stairs it "decides" to either climb, decend, or turn round and walk away. Similarly the mars rover robots cannot wait for a radio signal from home every time they have a problem. The future of robotics is in AI or decision making networks. the best use for XMOS in robotics is not in moving the robot but deciding what move to make, then passing move instructions to the robot controller.

_________________
When in trouble, fear or doubt, run in circles, scream and shout.

@Nimrod

But the question is can't XMOS do this itself without the X1000 dongle?

@persia

No.

@Mechanic

Care to elaborate what the XMOS can do when soldered directly to the board that a pci or usb attached XMOS cannot? On usb there is the rather slow (software) interrupt handling (IIRC 1 ms, but usually doesn't matter it's only for interrupts), but on pci you can trigger interrupts quite fast.

_________________
My programs: via.bckrs.de
MorphOS user since V0.4 (2001)

@Zylesea

Get to the CPU GPIO pins.

Actually I was just trying to get yet another negative snide remark. His innuendo needs
constant practice or it gets dull. So far in this thread the remarks have shown little improvement
over the usual ranting.

a.) X1000 dongle
b.) rather than buy an X1000 which lacks most of the tools necessary to actually use XMOS.
c.) 250 X1000s are not a game changer
d.) you wouldn't need to try to supplement it with XMOS.
e.) an OS that can't even currently use the 2nd core
f.) of it's low end processor
g.) to use parallel XMOS in any efficient way

(Yawn)
And now, I suppose, yet another commercial break.


(Sorry people, I must be getting bored.)

@persia

Quote:

persia wrote: @Nimrod But the question is can't XMOS do this itself without the X1000 dongle?

Anything XMOS can do, it can do regardless of the presence or absence of the X1000.
the result of a decision making net can be a simple boolean switch telling the device to stop or go. it can also be a datastream transmitted one instruction at a time (move to here, do this, move to there, do that). So far so good.
The trick with the X1000 is that it removes the bottleneck between the XMOS and the main processor and improves data transfer in the same way as a serial bus is an improvement over an on/off switch.

_________________
When in trouble, fear or doubt, run in circles, scream and shout.

@Nimrod

Quote:

The trick with the X1000 is that it removes the bottleneck between the XMOS and the main processor and improves data transfer

Any hard evidence to back up those claims ? We don't know yet how the XMOS is connected to the CPU, whats it's performance, how it effects other components connected to the CPU etc etc.

And I would be really impressed if you could show one reallife XMOS application where the bottleneck is a USB connection to the outside world and not something inside/between the XMOS-chip(s).

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@persia

Quote:

One could simply say if you put a more powerful CPU in the X1000 you wouldn't need to try to supplement it with XMOS. You can plunk a lot of XMOS chips into the X1000 and still not have the grunt of the 12 core intel behemoths that are rolling off production lines today.

Yes, that's true. Ignoring the general lack of affordable and powerful Power Architecture CPUs for a moment, one could do that. But it wouldn't offer the customisable I/O or the XMOS-style parallel processing. I'll take the Core i7 processor listed on Wikipedia's MIPS chart as the comparison point. Six core with two threads per core, offering 147600 MIPS. That's 12 threads, which is impressive, but hardly massively parallel in the way it would need to be if it is to be directly comparable with the XMOS architecture.

For an equivalent MIPS number with XMOS, you need about 300 of the Xena chips. Which is a lot, yes. With the quad-core variants, a 256-core expansion box that would get you into the same neighbourhood as the Core i7 MIPS-wise is possible and not horribly expensive or huge. But the thing to note is each XMOS core has eight threads. So scaling up to a similar raw performance as an i7 gives you something on the order of 2000 threads, all able to communicate near-instantaneously.

XMOS is currently testing the 'XMP-64', their 64-core development kit, to see what performance their architecture offers when used in large numbers. They get 25000 MIPS out of its 64 cores, with practically no latency between cores, and hence also threads.

I'll also address a point to Kronos here: latency between XMOS chips is 500 nanoseconds in the worst case scenario. Any thread can communicate with any other thread in the cluster in less than half a microsecond. They've tested it, see XMOS website's page on the XMP-64 for details. The short answer is that USB has more latency than XMOS and we can prove it.

@Zylesea

Quote:

Care to elaborate what the XMOS can do when soldered directly to the board that a pci or usb attached XMOS cannot? On usb there is the rather slow (software) interrupt handling (IIRC 1 ms, but usually doesn't matter it's only for interrupts), but on pci you can trigger interrupts quite fast.

XMOS on PCI isn't a good idea. Its better than USB, yes, but you're still waiting for interrupts. Having an event-driven coprocessor waiting around for PCI interrupts defeats the purpose of having one there in the first place. Really, the X1000 project can be seen as research into how a machine can be built with an event-driven coprocessor but without it being wasted waiting for things to happen.

Last edited by T-J on 03-Oct-2010 at 11:31 PM.

@T-J

Put you don't give an example of where USB becomes a bottleneck. There doesn't seem to be such an application currently in progress. The whole thing is currently a solution in search of a problem.

And what precisely is Apple's little cpu chip going to do with 2000 threads thrown at it? Isn't it going to become a bottleneck too? Wouldn't a massively parallel OS running on top of the massively parallel device be more efficient?

@T-J

Quote:

And of course you can connect more than one XMOS easily - that's one of their most salient features.

How I do that?

Quote:

X1000 is a product that can be applied to many uses. It is nicely suited to certain areas of industry and science, especially where there is a need for real-time control of hardware and real-time response to data input.

Do you really think there is a market noone already explorered?

Quote:

It is also interesting to the enthusiast, bringing the cost and skill-base needed to design custom expansion boards down into the range of most computing enthusiasts.

It is only wishful thinking. Could be nice they bought X1K but they wont. There are cheaper ways to have such hobby.

Quote:

As for Beckhoff, they aren't really doing the same thing as A-Eon and Varisys are trying to do. It seems that you would hire Beckhoff to build your factory an automated production line that would then have one function.

You obviously know nothing about Beckhoff then. We are using it in our automation systems and it performs various tasks for us. You can expand your I/O with modules and control various things from your PC (in theory it could be just any computer).

Quote:

Those cameras do all their image processing on-board, so all you're doing is sending the finished product from the camera to the desktop over ethernet. There wouldn't be much point using XMOS there because latency between the camera and the desktop isn't an issue - you're doing everything on the camera, so latency only affects the wait for the finished image.

Forget everything about image processing. You need an I/O to control camera. You might have multiple cameras which all have to be synchronized and preferably synchronized with (LED) lights. Then you might need something for camera's command bus or other logic. Maybe just monitor temperature sensor or air pressure and shut it down if it is overheating.

Quote:

If you need your machine to respond immediately to outside input in real-time though, the event-driven architecture of XMOS is useful. If you need to sample data at extremely low latency, it is interesting. If you want to explore the possibilities of parallel processing, XMOS looks to be the best solution out there right now. If you don't want any of these things, you won't have much use for it.

It can be done and currently is done with FPGAs. But not so many can code successfully in VHDL and XMOS could be more practical solution there. I just cant see X1K or any desktop computer in the picture. If it was smaller, like Acube's SAMs, it could be considered. Too bad we dont use PPC in our systems.

Quote:

You mean, you won't use A-Eon products.

We really couldnt even if I wanted to. For industrial use spare parts must be guaranteed to be available next 5-10 years. If you cant guarantee that you are out.

_________________
Amiga Developer
Amiga 500, Efika, Mac Mini and PowerBook

@amigang

Ok so got a surprise email from myself today, reminding me about this thread. Time to see how close we where and what to predict for next year.

Quote:

in a years time I hope to have an OS4 machine weather it be Sam460 or X1000 I dont know yet, I hope that both machine will have been on the market for a while and that they have done well.

Well I still have no OS4 machine, X1000 has only been hands of beta testers for a few weeks (if that) and its shame progress have been a little slow and I think we still have quite a bit of waiting to go before X1000 comes out. Yet again I will say it I really, really hope that this time next year I will have a OS4 machine in front of me.

Quote:

I would like to see Timberwolf get it proper release by then, OO for kids maybe getting its first released.

Well we know Timberwolf is progressing and I hope they show where they are up to at Amiwest. As for Ookids we know the ground work has been done but will it progress any further, my guess is once X1000 and OS4.2 are released depending on how well it does at attracting new interest and new developers to OS4 market will see if Ookids get pickup again, I hope it does, but to me Timberwolf is more important program to have that the Amiga community needs.

Quote:

maybe OS4.2 get announced and the OS4 market is slightly better and bigger with the x1000 out

Well I was close, OS4.2 has been announced, I'm guessing but I think the Sam460 has been reasonably popular maybe not as much as the Sam440, as my guess is a lot of users are waiting for the X1000 to come out or are just happy with their Sam440 for what they do on OS4. I think the Amiga market is stable, ie not really shrinking but not really expanding either, I hope X1000 can renew interest in OS4 as I think it is its last chance to expand the market.

Quote:

I think MorphOS will be on a PowerBook and Power Mac G5, Aros will be a lot more stable and be gaining more momentum

Well MorphOS 3 has been announced, I think am I right that it works on Powerbook G4 now? Again like OS4 I think progress is a little slow but I guess that a sign of the times and like OS4.2 and X1000 I hope expands OS4 users, I hope MorphOS 3 when it come out and support new Mac hardware also expands its market, and I really hope there is more respect and more co-operation between camps (dream mode off). Any way Aros, is moving slowly but surely, I think the Aros Kickstart replacement is the biggest progress we've seen, now people can make all in one bundles with UAE, and Aros Rom which I think is pretty neat. I've tried AspireOS and it deferentially a nice speedy mini OS to try out, need more time to check it out of what it can fully do but it is basicly AmigaOS.

Quote:

CommodoreUSA will be out of the picture by then, Amiga Inc may also be out of the picture by then too.

Well credit to them they are still both here (however only just I feel) Amiga lost the trademarks show just how little money they must have left and deals with Icontain also show they are pretty much just licensing out the brand name to anyone who can afford it. C=USA actually got a product to the market so well done to them but my guess and feeling is they are just trying to get investors to scam I could be wrong but their strategy seem to be all over the place, one minute its target the home, retro market now its target the power platform market, I just dont see it working out for them, I give them 10 to 1 they are still here this time next year.

So that it, I really hope this time next year thing will be different in the Amiga world, we will get the X1000 and thing progress from their. So whats your guess, see ya in a year.

_________________
AmigaNG, YouTube, LeaveReality Studio

@ClusterUk

Quote:

If I comment on this you will know my Aros plans so sorry, all I will say is Aros will be a lot more powerful and professional then, and a new IMICA will blow peoples mind

What was it, is it still being work on? Anyone else know, enfact haven't seen much of ClusterUk, last I saw he moved to the DARK SIDE with Apple Mac!!

@Kronos
Quote:

So yes, 2011 will be all the same as 2010, just with a few people disapering from the community.

Annoyingly your most closest to what happened. Lets hope we can do better for next year!

_________________
AmigaNG, YouTube, LeaveReality Studio

os4ppc market (hw &sw): a niche and such will remain. no prices variation, nor better nor worse. x1000 will make happy their users but won't sell much more than the betatester. an improvement in the os could be only if hyperion pays new developers. acube will continue to sell its own sams at the same prices.

natami: it'll continue to go on but no devices on the market still.

aros: continue flow and development. aeros will stick some linux devs. devbase will be larger. faster development.

morphos: 3.0 on the way, some minor improvements with x86 in the mind.

commodore usa: with christmas 2011 passed, they've made some good sellings with c64 and vic but not so high. "amiga workstation" market doesn't fly. amiga htpc is closed due to high prices. amiga aio is on mart shelves. not so bad pc. commodore os doesn't arrive.

@amigang

kronos was just reasonable about this, i dont understand what is annoying about reason? you can use some logical induction to have predictions based on past experience, there are even sayings like that. it cannot of course guarantee the results, but statistically you can have the closest approximation. based on that it isnt any problem to predict the development in amigaland, as it was proven with the last example "amigaone 500". alas the sceptics are being continously proven right.

Last edited by wawa on 29-Sep-2011 at 12:23 PM.

@Kronos

About the XMOS - CPU connection

"The Xena chip is wired on one Serdes Lane, so this gives it a high-speed connection to the CPU."

@amigang

In general, things are moving slower then we would wan`t to, but they are moving.

Well, Kronos can be well anoying sometimes, but he is smart user and defenetely has a moment of "slapping the truth to the face". However, some users stil remain as of 2011 which is a real miracle for itself too ...

_________________
Future Acube and MOS supporter, fi di good, nothing fi di unprofessionals. Learn it harder way!

@vox

yes it is really a miracle, these users who remain. and people like kronos have their "slapping the truth to the face" (as you call it) again and again, though you can alaways twist it a little around and say, "well, they are right, but... "
however, amigang, you were so smart to send this mail to yourself last year, why not do it another time? lets see what will happen in a year from now. sometimes im puzzled myself, i think its a cool idea!°

Quote:

clusteruk wrote: @amigang ...I will win the lottery, buy out Amiga inc and rule the world of Amiga

well I'm hoping that happens for you Steve because at least we'll know Amiga land will be in good hands

as for me next year, I'm hoping I'll be healthy enough to play with my then 7 and 4 year olds and of course rockin with my AmigaOne X1000

Last edited by klx300r on 29-Sep-2011 at 09:27 PM.

_________________
____________________________
c64-2sids, A1000, A1200T-060@50(finally working!),A4000-CSMKIII
! My Master Miggies- Amiga 1000 & AmigaOne X1000 !
mancave-ramblings
X1000 I BELIEVE

Quote:

WolfToTheMoon wrote: All in all I'm ... expecting ... only further reduction in the number of AmigaOS users in the next year

ya & we will all sell our OS4.x machines to buy PC's with Amiga stickers on them to run almighty Linux shells to mimic AmigaOS

_________________
____________________________
c64-2sids, A1000, A1200T-060@50(finally working!),A4000-CSMKIII
! My Master Miggies- Amiga 1000 & AmigaOne X1000 !
mancave-ramblings
X1000 I BELIEVE

@amigang

Quote:

@Kronos Quote: So yes, 2011 will be all the same as 2010, just with a few people disapering from the community. Annoyingly your most closest to what happened. Lets hope we can do better for next year!

Why should "we" do better ?

OS4 and it's HW is still in the same place were it rises a few eyebrows in the general public, but thats were it ends cos of the actual price and useabilty of those products.

MorphOS will get mobile, but will that create really new users ? Or just the existing ones buying another licence ?

Sure both will get the curious stumpled-in from time to time but those aren't enough slowly fading away from useing their "Amiga" for more than changing the window-deco every 3 months.

AROS will continue just like before and just like desktop-linux it will be "allmost there for 15 years".

Projects like OO.org and Timberwolf will limber on with no real results (how often did we hear the "we changed the rendering-engine and now moving to a new code base" allready ? ).

Sceptics will still be shouted down while people letting their fantasy go rampant will be proclaimed prophets.

/me will still be around next year, so let go all hope

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada