Dr. Garry Hare Amiwest Speech

[KimmoK]

@KennyR

>Which only feeds the accusations that KMOS is not actually a new company at all.

If I would employ Garda, I would become bplan?

>Same employees,

KMOS bought AmigaInc with AmigaDE. It is logical that they employ some AmigaInc people, unless they are going to waste everything AmigaInc is worth.
Read what Garry said before you post.

>same product, same kind of talk, same aims.

I see some differense in all of those.

>But just no debtors. Hmm.

Why not? I'm sure they are there. It seems KMOS is paying some of the unpaid debt already.

[KimmoK]

"but from what I can tell, there's no indication that there will ever be an OS4.1?"
It's Hyperion's business. Read another interview.



@Colin_Camper

"I thought that, because most of what was going to be 4.1 is going into 4.0, that 4.1 was just going to be a bug fix boing bag."

That has never been the plan. New features have been planned in 4.1. Thing like reworked workbench, Warp3D Nova, etc... not sure about the current plan of the whole content.
(Ben Hermans mentioned some other things in his interview)

"My take on Garry's speech is that he wants to develop AmigaOS into a ultra fast, lean and stable OS."

Yes, something like that.

"The Tao renegotiations speak volumes about the pivotal role intent will have in all this."

AmigaDE is almost 100% unrelated to AOS plan.
I got the feeling that Garry is going to use AmigaDE developers to help to bring AOS forwards (AOS itself, not DE).

"Prepare for an AmigaOS running on VP!"

Never.
It would need to be the other way around.

- VP is not PowerPC, it's incompatible. It's a totally different kind of CPU.
- I'm pretty sure that AOS responsivenes&speed can not be fully reached if the CPU is not used natively. The use of that extreme virtualization would also make it impossible to take full advantage of HW acceleration (like 3D).
- VP/intent does not support memory protection (MP is planned for AOS4.x) it never will.
- I think not even AmigaInc was planning that AOS5 would run on top of VP.

etc, etc...

[KimmoK]

@Glaucus

>Not sure what the state of VP is these days, but creating a complete OS based on the idea is exciting.

There are some realitities that describe it as "not optimum" in various issues.


>>- VP is not PowerPC, it's incompatible. It's a totally different kind of CPU.
>Yes, and what you'd need to do is create a low-level hardware interface that is basically the VP run-time engine.
>The OS would run on top of this.

It is not sensible to run all of the OS on top of VP.

>This way, the OS and all the apps would be translated to native PPC, Intel/AMD, MIPS, etc, on the fly.

But the OS and all of the apps would need to pe ported to another CPU first, the VP CPU.

Ask Hyperion if they are willing to do it...


>>I'm pretty sure that AOS responsivenes&speed can not be fully reached if the CPU is not used natively.
>So what. CPU's these days are ridiculously fast. Do we really need a 3.5Ghz chip to surf the web? With chips like that, even a 50% hit in performance would still provide a very satisfying user experience (and I think VP can achieve far better then 50%).

50% hit in performance would kill AmigaOS and me (because I would be so ashamed and disgusted).
And for house heating I'm going to use other, better means already.

IMO, one of the basic ideas of AOS is to be more efficiant than others.

One of KMOS intentions is to use AOS in some embedded use. 3.5 Ghz CPU does not fit in embedded use.

>>The use of that extreme virtualization would also make it impossible to take full advantage of HW acceleration (like 3D).
>3D graphics is tied to the video card, not the CPU. The VP engine would have nothing to do with this.

If 3D drivers are run on top of VP, it cuts off some of the performance.
If the whole 3D system is virtualized as "GVP" (GraphicsVirtualProcessor?), the efficiency of low-level Warp3D can not be reached. And every thime there appears new Graphic processor, the GVP would need to be rewritten.

Again. IMO: we should outperform Wintel rather than be slower on purpose.

For audio & pleasant sound experience, near "real time OS" performance is required.

etc...

(High level OS things could be run on top VP, but none of the low-lewel OS things. U C, there are reasons why intent requires a HOST OS on almost everywhere.)

>Alti-vec, however, might pose a problem, but there might be ways around this as well.

Right.
IIRC, not even the FPU "problem" is solved yet. Intent FPU is slow.

The way around altivec is that VP does not have altivec, and I'm sure it never will have any good SIMD VP unit.

I think making a SIMD VP would be totally insane attempt. (intent gives the possibility to use native executables amongs VP ones, perhaps that could be used...)

>>VP/intent does not support memory protection (MP is planned for AOS4.x) it never will.
>Perhaps we need to write a specialized VP system for the OS.

It has taken about 13 years for TAO to bring the VP to the current level. I doubt it would be sensible to try top outperform them in a few years time...

>Or perhaps, add it to the hardware interface and spawn a new VP instance for each application and keep them all in their own sandbox.

Hmmm... I wonder where the OS would reside then.

>Anyway, I think it's technically possible, might require more resources then Amiga has right now.

In the end....

I think at some point AOS5 would have been a HW banging OS (with MP etc, for limited set of 64bit CPU motherboards, etc.) that seamlessly runs some parts of it and some applications on top of VP(s).

It should have been possible ... the rest is history.

Let's focus on one thing at a time. All energy to get the initial AOS4.0 out first.

>It would be the ideal system and making an OS that is binary compatible with the majority of systems out there would be a great advantage for the Amiga.

IMO: AOS would not perform like AOS should if it is run on top of virtual machine.

Why would we need a OS that is binary compatible with everything?  ((I hope not just to be the first one at any cost?))
What if PPC already (or tomorrow at least) can be put everywhere?

Isn't it enough if we have applications that can be run on any system out there?  (intent and AmigaDE apps)
(And at the same time applications that run best on AOS.)


(I'm sure after the workbench is rewritten for AOS4.1, it will be possible to have identical looking user interface running on AmigaDE.)

[KimmoK]

"Security through Obscurity as Ben put it"

That's what we have today, and VirusZ.  :-P

Even though that "obscurity" security migh outperform any WIntel security solution that exist today, I'm sure KMOS and Hyperion knows that it's only the beginning.

[KimmoK]

"....that AmigaOS will be ported to ARM."

To a yet another little endian CPU?

"We know that OS4 has employed good design practice and has a HAL so it's quite easy to port to other platforms without any big upheavals."

It would break binary compatibility, unless that's why KMOS now sees intent to be important alongside AOS.

"The questions that remain therefore, is how do the teams at Hyperion and KMOS fit together? Who will do the ARM port? Who will develop the higher levels of OS4 further? Will priorities mean that OS4 development slows or loses out in favour of development that will make the system more scalable?"

Reality is that the desktop niche most likely would not be enough anyway. So AOS needs to scale.

"Also unrelated but interesting. Will the ARM port gain greater focus than the PPC version?"

I think there will not be ARM port of AOS in near future.
ARM is not needed. ARM or Xscale is not scaleable and powerfull enough. Today PowerPC fits in "bigger than PDA" sized products (mobile even) and PPC cores are not that much harder to embed. Tomorrow... we'll see.
And I'm pretty sure that KMOS is not stupid enough to go against Symbian&friends with their own cell phone OS, only M$ is.

[KimmoK]

>>It would break binary compatibility, unless that's why KMOS now sees intent to be important alongside AOS.
>If they want to run on mobile devices (as Garry has plainly said) then the only other option would be to port AmigaOS to Intent.

"mobile devices" covers huge set of devices, not just cell phones as we know them.

And I think KMOS is not stupid enough to try to push AmigaOS to current cell phones with current processors.

Especially, there's absolutely no room for AOS on symbian devices (no physically or otherwise).


After a few more thoughts:   ARM CPU is a minor thing in the whole mobile space.


> But Intent is no operating system, it must be hosted. So running AmigaOS on Intent on the host's OS on the hardware doesn't seem "small and fast"...

Absolutely.


>ARM is more than suitable for mobile devices. That's why Nokia, Ericsson, Panasonic & co are using it right now to run SymbianOS on their devices.
>Symbian has practically a monopoly on the mobile device market - almost all of these devices sold uses ARM processors, there is no reason why this would change in the future.

You are talking about cell phones. There is no room for AmigaOS on any symbian devices.  AmigaOS can not target Symbian devices.

(And companies you mentioned do a lot more than just cell phones, just a side note, but anyway. All of them already use and are familiar with PowerPC.)

ARM cores are integrated in phone media/telecom chips. Also MIPS and PowerPC cores are possible to be integrated if SymbianOS (binaries) is not used. If binaries are not used, there's no absolute must to use ARM either, and especially so with the enterntainment chip of the device. (phone stuff and JAVA stuff run in separate spaces, often on separate CPU/MPU)


> there is no reason why this would change in the future.

I think there is no reason for a cell phone to switch to AmigaOS. It would need to be something new...


>Compaq isn't going to start producing computers with PowerPC because it doesn't run the OS everyone uses (Windows).

Right.

>Just like Nokia isn't going to start producing PowerPC phones, because it doesn't run the OS everyone uses (Symbian).

Right.
Except that Nokia can get Symbian ported to any CPU, most (user) apps run on top of JAVA anyway.
And there is also no reason for Nokia to use AmigaOS.
And there is no absolute reason to use the bulk chip from philips that a lot of smaller companies use.

Asian market and asian cell phone manufacturers might have other interests, though...  (there are those who do not use WinCE or Symbian, tens of millions of phones, etc.)

(I wonder what CPU does MyOrigo use (another phone, intent based, with Finnish origins) ....   perhaps ARM, though, as one can buy cell phone chip with ARM from the local shop, grocery shop even.)

>If you want to make a new OS to run on mobile devices, then it must run on ARM.

If you plan to reflash Nokia (etc) cell phones to use AmigaOS, then yes. In practice it's impossible to make business with it.

Also. Several phones use more than one CPU. There might be one handling the phone side and some other handling PDA/enterntainment side.

>>And I'm pretty sure that KMOS is not stupid enough to go against Symbian&friends with their own cell phone OS, only M$ is.
>From reading the speech and the interview, this seems to be exactly what KMOS is doing! I'm as dubious as you are about how successful they can be....

Hmmm... Garry did mentione phone use several times. I thought it was mainly as an example and (only) as a one possible client/user of their mediasystem.

But ok. I'm not sure any more. It would need to be some new device, some new comer from asia perhaps, most likely not for European(/rest of the world) standards...

The competition is hard. Companies work heavily to try to differentiate from the rest as their advantage, find new sub niches perhaps, without loosing the mainstream.

I can not see how KMOS could gain any signifficant ground there... not anytime soon ... only remotely possible would be companies outside WindowsCE and Symbian users. How many there is? How big are their markets?

Interesting.