@m3x

Thank you for your answer.

I understand that things might become different in a close future. But I must say that I will believe this when I see it, not before. I would also like to stress that this should be a top priority, seing how several other entities (all of them even?) has failed on this point. This should be number one on your to-do list, designing a motherboard should be number two.

Also, at this point in time (for the sake of discussion), Samantha is not "Amiga Hardware" more than neither the Efika nor the Pegasos 2. The reason I point this out is because some people seems to constantly filter out these two products from any discussions about possible Amiga hardware, despite their price/performance ratio compared to others, thus them being potentially better alternatives.

@takemehomegrandma

But remember that Amiga Inc. did announce Samantha as new OS4 hardware, so even if the contract has not been signed yet, it looks to be a sure deal, even if it's not written in stone yet.

@takemehomegrandma

it's our top priority, but please note, signin a contract is not a matter of minutes, it can take several days to accomplish it.

_________________
Massimiliano Tretene, ACube Systems, Soft3

@abalaban

Quote:

Yes and thus you won't be able to install applications anymore...

You log in with the administrator, and change the rights of the registry entries if the app is badly written and refuses to install if you don't change the rights of the registry entries

Quote:

yes and with this WinXP you are unable to do anything because you have disabled almost any service needed by the key features of WinXP

If you make that changes you won't have a single problem. What are the key features of WinXP for you? the autoupdates? for me it's the ability to run apps.

Quote:

Why not just use Win95 instead, it really less ressources hungry and will really fly on our latest hardware

Win95 kernel is crap. Windows 5.x kernels are much more stable and work more smoothly in the same hardware.

Quote:

And also why use Windows if you have almost as much things to tweak in order to obtain a correct working OS than on Linux (except the fact that on Linux those "options" are not hidden and are documented) ?

If you think that try to change the type of monitor once you have installed ubuntu

You won't be able to do it with the GUI and you'll have to edit text files.

The only reason I would run Windows is because of its apps. Linux doesn't have the same apps as Windows.

_________________
The only spanish amiga news web page/club: CUAZ

@COBRA

Quote:

Being sold as a realtime OS does not mean it's more real time than OS4. Embedded Linux also claims to be realtime enough to be suitable for embedded :)

So what maximum worst case interrupt latency does OS4 guarantee? How about embedded linux? How about scheduling latencies? Context switch times?

See, that's why they are not real time operating systems. RTOS (such as QNX) will have the answers to those questions documented.

@umisef

Can you tell me those values for QNX and Embedded Linux?

@Crumb

Quote:

Win95 kernel is crap. Windows 5.x kernels are much more stable and work more smoothly in the same hardware.

Yes and also needs more ressources that's a fact, just look at the minimum requirement on the box

Quote:

What are the key features of WinXP for you? the autoupdates? for me it's the ability to run apps.

For me running apps is not a key feature nor a normal one : it's a prerequisite for an OS

Quote:

If you think that try to change the type of monitor once you have installed ubuntu You won't be able to do it with the GUI and you'll have to edit text files.

I never said it would be easy with a GUI and all. Just said that what you were suggesting was to fall into the same sort of difficulties as with Linux except that generally under Linux it's not an hidden option and is stated in the doc contrary to what you have under Windows.

Quote:

The only reason I would run Windows is because of its apps. Linux doesn't have the same apps as Windows.

Yes and you were suggesting (for security and safety reasons) to use applications originating from the *ix world


PS: I have nothing against you, neither I do against Windows/MS (even if...) that's just that I couldn't let someone say WinXP is really great and is only bloated, slow and ressource hungry just because of the user in fact for me compared to win2k winXP is a regression with some cosmetics "ameliorations" to make it pass.

_________________
AOS 4.1 : I dream it, Hyperion did it !
Now dreaming AOS 4.2...
Thank you to all devs involved for this great job !

@COBRA

Quote:

Did I say it was the only difference?

It was the only difference you mentioned; And you mentioned it in a way which suggested that IT was what was responsible for the difference in interrupt response time.

If that was not what you intended, please explain how comparing the observed worst case interrupt latency of embedded linux on a 200MHz ARM versus that of CMX on a 10MHz 16 bit CPU was supporting your point (which, you remember, was that OS4's message passing is magnitudes faster than QNX's).

Quote:

Actually QNX is the linux-like OS (unix core with a GUI pulled on top of it), while OS4 is not :)

When it comes to worst case interrupt latencies, whether an OS is RTOS or not is rather more important than how it passes messages.

OS4 is more linux-like than QNX when looking at realtime capabilities
QNX is more linux-like than OS4 when looking at message passing
OS4 is more linux-like than QNX when looking at kernel architecture (monolithic vs microkernel)
QNX is more linux-like than OS4 when looking at exposed API (POSIX vs proprietary)

And so on. So pointing out anecdotal observations about linux won't tell you much in the OS4 vs QNX debate, because you can't even decide whether linux is OS4-like or QNX-like.


You stated that QNX's message passing was "magnitudes" slower than OS4's. I explained why I consider that a silly statement, complete with explanation involving access-delays and so on. You then simply reiterated your statement (which does not support it :), and related the funny story of embedded linux vs CMX (which lacked any sort of relevance). You have still not provided a single technical argument for why a message round trip under OS4 should be "magnitudes" faster than under QNX; You have not even addressed my reasoning that leads to a factor of merely 2 on moderately powerful hardware, and 3 on really really weak stuff.

@abalaban

Quote:

PS: I have nothing against you, neither I do against Windows/MS (even if...) that's just that I couldn't let someone say WinXP is really great and is only bloated, slow and ressource hungry just because of the user in fact for me compared to win2k winXP is a regression with some cosmetics "ameliorations" to make it pass.

I don't think WinXP is great but I don't agree with those who blindly defend OS4 regardless of what is being discussed.

_________________
The only spanish amiga news web page/club: CUAZ

@umisef

Quote:

If that was not what you intended, please explain how comparing the observed worst case interrupt latency of embedded linux on a 200MHz ARM versus that of CMX on a 10MHz 16 bit CPU was supporting your point (which, you remember, was that OS4's message passing is magnitudes faster than QNX's).

The systems I usually work with have a number of seperated software components, each with one or more tasks/processes. When an external event (interrupt) is triggered, a task wakes up, interprets the event and communicates with the other components whatever is necessary. The overall reaction time of the system will be the time from the hardware interrupt to the task starting to run, plus the time for the necessary inter-task communcation, plus the time for the components to do their necessary processing. Unless you have a very simple system with no tasks, in which case you don't need an OS in the first place.

Quote:

When it comes to worst case interrupt latencies, whether an OS is RTOS or not is rather more important than how it passes messages.

From my experience the speed of message passing is crucial in an efficient embedded software design, at least on sub-400Mhz class embedded CPUs. I have worked with systems which suffered severe problems due to the inefficient inter-task message passing mechanism they used. You may disagree with me, but please don't present your views as facts. I've used like 5 different OS'es in embedded systems development in the last 5 years so you will not convince me otherwise. And note that I'm not saying that OS4 is the best choice for embedded, because it's not, there are many dedicated RTOS'es which are designed for the purpose from the ground up and would do a better job. But I would choose Integrity/VelOSity over QNX any day, because they offer far more efficient inter-task communication, like OS4 does.

Quote:

OS4 is more linux-like than QNX when looking at realtime capabilities

Again, don't present your views as facts, I'm not saying that the above is wrong or right as I don't have the figures. If you're really interested, contact Hyperion and ask for the interrupt latency/etc. timing of their kernel.

Quote:

You have still not provided a single technical argument for why a message round trip under OS4 should be "magnitudes" faster than under QNX

Say you have some data to be processed in an embedded system, being an encapsulated audio+data frame or whatever (very common case in telecomm), streaming from one component to the next for digital processing, channel seperation, decoding, whatever. What takes longer (on any hardware), passing a few pointers, or pysically copying several megabytes of data per second? I think you know the answer.

Now this is getting way off-topic, so I end this discussion here.

Last edited by COBRA on 31-Oct-2006 at 01:35 PM.
Last edited by COBRA on 31-Oct-2006 at 01:31 PM.

@Crumb

Yes I agree with you : in the current state OS4 is unusable in a business environment (except if you rules your owns little business like selling ice-creams for example and use it just to do email and keep track of your stock with a mini-spreadsheet/database)

_________________
AOS 4.1 : I dream it, Hyperion did it !
Now dreaming AOS 4.2...
Thank you to all devs involved for this great job !

@COBRA

On Posix OSes like QNX I guess you could use threads instead of "tasks" so your context switches would become shorter

Last edited by Crumb on 31-Oct-2006 at 01:57 PM.

_________________
The only spanish amiga news web page/club: CUAZ

@COBRA

Quote:

How is it that if you have 256MB of RAM WinXP eats most of it straight after boot, even before any applications have been started? Do you still not see why it's completely useless to make comparison?

That's what caches give you. Unused memory is wasted memory.

_________________
“It is well enough that people of the nation do not understand our banking and monetary
system, for if they did, I believe there would be a revolution before tomorrow morning.”
~~ Henry Ford

@COBRA

Quote:

I do not agree with your argument about OS efficiency, QNX's message passing will never compete with something like OS4's and their PPC kernel even uses Altivec (if available) to speed up the data copying, but it really shows when you want to exchange large amounts of information.

You always have the choice of using shared memory, and then message passing is as fast as on AmigaOS.

_________________
“It is well enough that people of the nation do not understand our banking and monetary
system, for if they did, I believe there would be a revolution before tomorrow morning.”
~~ Henry Ford

@COBRA

Quote:

Actually QNX is the linux-like OS (unix core with a GUI pulled on top of it), while OS4 is not :)

Didn't you say you worked at QNX? Guess not in the sw department, though...

QNX is not unix, QNX is a microkernel which has support for the POSIX API. BeOS has support for the POSIX API as well, would you say BeOS is unix with a gui on top?

_________________
“It is well enough that people of the nation do not understand our banking and monetary
system, for if they did, I believe there would be a revolution before tomorrow morning.”
~~ Henry Ford

@falemagn

Quote:

Didn't you say you worked at QNX? Guess not in the sw department, though...

Actually it was the sw department :)

Quote:

QNX is not unix

I said unix-like, didn't say IS unix. *sigh*

@COBRA

Quote:

Can you tell me those values for QNX and Embedded Linux?

For embedded linux --- of course not, it's not a RTOS.

For QNX --- it's in the manual; See page 8 (labelled '6') for the info on a P133, for example.

You might also want to look into the evaluation reports available right there on QNX's website (there is other good stuff in the download center, too :)

Last edited by umisef on 31-Oct-2006 at 02:30 PM.

COBRA
"Pegasos2 is discontinued"

Stop spreading your lame FUD!
The Pegasos 2 is not discontinued until Genesi
says so. I told you to listen to what people tell you
since you clearly don't know anything about
the Pegasos 2.

Moderator note: last part removed.

Quote:

Site Netiquette: We share one common ground, we are all Amiga users! Respect other opinions, even though you may not agree. Everyone is entitled to their say, but please do so in a courteous manner. Off-topic posts will be moved to the proper forum, and further action may be taken against repeat offenders. It's easy to flame someone if you disagree, but remember, that type of behaviour is strictly prohibited. We come from different locations, we are a mix of ages, and our backgrounds vary. Take that into consideration before you click the post button. A moment of thought is better than hours of grief. The general rule of thumb is treat others as you would like to be treated.

Last edited by zerohero on 31-Oct-2006 at 02:59 PM.
Last edited by zerohero on 31-Oct-2006 at 02:58 PM.

@COBRA

Quote:

Say you have some data to be processed in an embedded system, being an encapsulated audio+data frame or whatever (very common case in telecomm), streaming from one component to the next for digital processing, channel seperation, decoding, whatever. What takes longer (on any hardware), passing a few pointers, or pysically copying several megabytes of data per second? I think you know the answer.

Once again --- as the message receiver will look at the data (because otherwise passing it is pointless), things won't run "magnitudes" faster, even though the passing of pointers is indeed much faster than copying of data.

In fact, in a non-copying system, the pointer is passed (say for free), and then the receiver will access those megabytes of data per second.
In a copying system, the sender will make the very same accesses (although potentially in a more efficient order), then write out the megabytes per second of the copy (taking roughly the same time on a cacheless system, or very little time on a cache-equipped one), and then the receiver will access megabytes per second of the copy (once again, spending at most as much time as it would in the pointer-passing case)

So, max total time needed to do this in the copying case: Three times as much as in the pointer-passing case. Is three times "magnitudes"? I think you know the answer.

Oh, and

Quote:

please explain how comparing the observed worst case interrupt latency of embedded linux on a 200MHz ARM versus that of CMX on a 10MHz 16 bit CPU was supporting your point (which, you remember, was that OS4's message passing is magnitudes faster than QNX's).

@thread

The TOS are simple really. Here, I'll simplify them.

Be kind and respectful to others, their work, their chosen platforms. We all have opinions and a right to those opinions. Please think before you click on the 'submit' button.

I am sure you are all mature enough to moderate yourselves, for those who are not, thats where we as Moderators come in.

Please play nice, all of you.

Thank you.

_________________
¤¤ Official Hyperion Zealot ¤¤

(No, I didn't type that with a straight face.)