Quote:

People also said that ppc port of amigaos was impossible, firefox port was impossible

This is different. It's been 15 years since people are telling "it can be done"... still nothing after 15 years. And furthermore, the problem is technical.

You're free to wait another 15 years... I won't.

_________________
http://www.warpdesign.fr/

@pixie

Quote:

pixie wrote: @HenryCase Quote: So as you can see, AROS doesn't need to be rewritten from the ground up to support those features. It really is a question of priorities right now, MP and SMP are not high priorities as there are more important features to work on, though you're welcome to start coding if you believe they are. While there might be more important things my only wish is that these priorities takes SMP into considerations so that they don't have to be rebuilt nor hinders AROS in any way. Multiprocessing is here to stay and there's no turning back.

I agree about multiprocessing and all that, BUT in the case of AROS it would be overkill to introduce MP now ? I mean we don't even have the software that would need the extra resources. Yes, it's a chicken egg problem but this time I really think apps that could make serious use of it would have to emerge first.

_________________
AmigaOS or MorphOS on x86 would sell orders of magnitude more than the current,
hardware-intensive solutions. And they'd go faster.-- D.Haynie

@radical

Quote:

But when things like changing the mouse pointer and dragging and scrolling screens couldn't possibly work, as they did on any real Amiga, it wasn't worth pursuing seriously.

Mouse pointer and scrolling screens is done and parially done atm.

_________________
AmigaOS or MorphOS on x86 would sell orders of magnitude more than the current,
hardware-intensive solutions. And they'd go faster.-- D.Haynie

Quote:

I agree about multiprocessing and all that, BUT in the case of AROS it would be overkill to introduce MP now ?

Nobody's talking about implementing it right now... But rather having new APIs that are safe for SMP/ressource tracking/..., so that at least new applications can use it and be ready (or only need minor changes) for SMP.

You can't switch right now... But if you want to switch, you have to prepare the transition, cause as it is going, we'll never be ready, since the problem will always be the same: "it would be overkill to introduce it right now"...

Last edited by Leo on 27-Mar-2010 at 09:45 PM.
Last edited by Leo on 27-Mar-2010 at 09:45 PM.

_________________
http://www.warpdesign.fr/

@Leo

and it will happen when the time is right......

_________________
AmigaOS or MorphOS on x86 would sell orders of magnitude more than the current,
hardware-intensive solutions. And they'd go faster.-- D.Haynie

@Leo and others

Nobody will try to underestimate the need for SMP, nor will I. People that follows the AROS Developers mailing list probably will remember some messages of mine, asking for SMP support when the very first Athlon64 X2 came on the market. I'm just testing right now a Core i7 985X with its whopping 6x2 cores and gosh... it's clearly THE future... no, it's the CURRENT DAY. However, being more pragmatic, I don't consider SMP a priority for AROS anymore, and that's why:

1. project goals. I'm quite sick to repeat AROS goal is source compatibility with AmigaOS 3.1 API (and binary, on 68K). There are people - like Damocles - that progressively got sick with this, but that's it. AROS goal is to preserve this API, be it right or wrong, is something current AROS developers at least agree and gives a clear direction (at least one!!!) to the whole project. Who didn't agree anymore started Anubis. But "started" is a big-size word, since I didn't see much more than discussions, and not so many, on the ARIX maliling list. Introducing SMP would mean going far from the API which keeps the project unite and living, so it's better for now stick to a single core and having AROS completed, before extending it.

2. Typical computer for AROS won't be a pricey Core i7 900, nor a full-fledged but still economic Phenom II 900. It will be a secondary PC with a older processor, with 2 core max. Thanks to AROS little footprint, there are enough hopes it will perform very well and won't make you feel the loss of the second core.

3. The best platform you may use AROS on - and Steve will agree on this - is the single core Atom netbook, and you just won't believe how good this computer might be, until you install Icaros Desktop onto it (and you learn that the OS can be loaded in 10 seconds instead of 60, and that you can be on the web in less than 20 seconds, instead than waiting 2 minutes for the OS and the browser to load).

4. Multicore processor are good, but GPGPU is better. I'd prefer to see OpenCL addressed onto Gallium3D instead of struggling for a way to add SMP to something that wasn't born to support it. Having a reliable GPGPU library would be quite handful for heavier multimedia and rendering jobs. Not counting scientific computations, that I leave out this topic since scientist will hardly use an amigoid operating system at all.

5. AmigaOS 4.x is going SMP as well, and this means, IMHO, that AROS should wait for it BEFORE trying its proper way to do SMP as well. Why? Because AROS has its roots in the 3.1 branch of AmigaOS, and we all agree about this, but it's already much more advanced as well, and it is considered a "modern reimplementation" of Amiga OS, more or less in the same way MorphOS and AmigaOS 4.x are. The need to follow, or at least consider, AmigaOS 4's SMP on the X1000 is - IMHO - crucial for future application developments. Friends at Hyperion are facing exactly the same problem we're facing today when discussing about SMP and AROS, having two completely different ways to handle it would maybe mean issues for coders trying to create applications able to compile on both systems.

6. Before trying to use the other cores of modern processors, we should be able to use correctly the one we already use. This means adding all the missing features AROS lacks of. Printing, accelerated video layers, screen compositing and, last but not least, bug hunt and fixing.

SMP was an example: there are a lot of other missing things that are in the same state... that is to say that would require a new API, and would break compatibility that's being postponed since 10 years because "the time isn't right"... resource-tracking, smp, and so on...

If it was me I'd work on that even before thinking about having some 3D... This is the core OS.

Then 3D stuff, and everything can be built on top of it. Taking advantage of it. Not the other way around (ie: building printing, layers,.. stuff on top of old/hacked kernel)...

Too much importance is being made in compatibility. But it's been 17 (!!) years now since AmigaOS 3.1 has been released! Time for a refresh... No more "when the time is right"...

Last edited by Leo on 27-Mar-2010 at 11:20 PM.
Last edited by Leo on 27-Mar-2010 at 11:18 PM.
Last edited by Leo on 27-Mar-2010 at 11:18 PM.

_________________
http://www.warpdesign.fr/

@Manu

Quote:

I know you've done a lot for Aros way back and I respect that, but I don't see the holy API you talk about.

Go review the threads, especially when AROS64 development was under way. AROS64 was going to be a fork in plain view, MP and SMP was planned to be added but the 3.1 API folks got upset about it and not enough were helping Dr. Schulz to do a proper SMP/MP AROS64 kernel. It all went rapidly downhill from there, my investment (and Dr. Schulz's time) were wasted.

Quote:

From my point of view it was many that supported AROS going in more modern directions AFTER the original goal had been reached.

What a waste of time and effort to complete something so out of date in 2010. I suspect AROS will see 1.0 eventually, in a couple of years, maybe. Then what, 2012 and then you have 1990s OS completely cloned.

Quote:

I would leave AROS too if it would stop at 3.1 "holy" API but I really don't think it will stop there,

If and when it reaches it, I hope it does too, for you sake.

Quote:

Talk is easy, doing is where it's going to cost efforts and sweat from a lot of people.

What really boiled down to is defining who is the actual project leader and looking at their work history. Had I realized who was the real power behind Anubis and what role Dr.Schulz was going to play in it, I don't think I would have been that excited about the project. If Dr.Schulz goes as head of the project, I'll gladly do what I can for him. If he and I can come to an agreement for late Summer or Fall on a new project, I may fund it to some degree.

How does the song go, "What a long strange trip it's been" by the Dead?

_________________
Dammy

@damocles

The problem with Anubis is that it's goal was to reinvent the wheel and take Amiga OS down the same path as OS X. It worked for Apple with hundred of millions of euros to throw at it but it;s hard to see how a tiny niche market machine could achieve it.

Part of it was meant to appeal to Linux folks by saying you want a logical file layout, but to Linux folks they already have a logical file arrangement, it just isn't the Amiga one.

I suppose Anubis can rise from the dead to threaten the chickens once again, but somehow I doubt it.

@persia

Quote:

he problem with Anubis is that it's goal was to reinvent the wheel and take Amiga OS down the same path as OS X. It worked for Apple with hundred of millions of euros to throw at it but it;s hard to see how a tiny niche market machine could achieve it.

Or it's the way OS4 was going to be done around the year 2000 with Gateway. Remember the days when people were excited about Neutrino kernel being the new OS4 kernel? I was happy when it went to Linux kernel, it was still the step in the right direction. 2010 and we are still swimming in pig slop saying we need to go deeper to find Nirvana. C= would have changed the OS to something radical without a second thought to 3.1 and earlier just like they did with AGA over ECS.

We don't need the Red v Blue v Black wars, we need a entirely new modern OS. Time is being wasted on all three camps as they develop more code, express lane to nowhere as the world passes them by with evolution.

_________________
Dammy

@paolone
I agree with you: Its more important to make AROS complete and stable (and create more apps for it), before adding SMP.

_________________
AmigaOne X5000/40, 2.2 Ghz, 4 GB RAM, Radeon R9 280X, M-Audio Revolution 5.1, 240 GB SSD

Quote:

The problem with Anubis is that it's goal was to reinvent the wheel and take Amiga OS down the same path as OS X. It worked for Apple with hundred of millions of euros to throw at it but it;s hard to see how a tiny niche market machine could achieve it.

I don't think you need millions of euros nor hundreds of developers to write an OS. You need talented people with time, that's for sure. And this wouldn't have been as huge as what Apple did with OSX since Linux was supposed to be the foundation.

What I think is that you need someone with a clear vision of what's needed, and also enough technical skills not to go directly into the wall... Plus some dynamic and interest so that the project doesn't die quickly. Seems like something was missing for Anubis. I haven't looked at it that much so I can't say what was the problem.

That being said, I agree with Damocles.

If the three "camps" decided to work together instead of reinventing the wheel, we would have enough resources to build something great. Especially since MorphOS already as a working and stable modern kernel in the form of Quark (which current ABox/QBox uses as a base). So this wouldn't be needed to be written from scratch. The compatiblity box is also already there with ABox.

No need for millions of euros, hundreds of developers, thousand of hours... Just some humility (so that it doesn't end up in personal stuff and stupid wars), and common sense...

This money stuff, "when the time is right",... are just excuses for this situation and behaviour of all camps... But there's no excuse: we could have a cool and modern OS if we wanted to.

_________________
http://www.warpdesign.fr/

@Leo

Quote:

Seems like something was missing for Anubis. I haven't looked at it that much so I can't say what was the problem.

All the people where there, including some old AROS Devs which have since moved on to Linux. Coordination and a strong manager with clear vision of what had to be done was missing. Other key portion that was missing was someone who was experienced in working in small volunteer groups vs large groups or corporate groups to be the project leader. When working with volunteers, one must have a solid plan which can be modified by the group on a need by need basis. Putting all the ideas into a hat and hope for the best while waiting on group decisions is a death blow to any open source project, IMO, especially when the devs are in so many different time zones spanning across the globe. Then other projects started to steal individual dev's time away and it caused delays and that caused more confusion, and just spiralled into a death dive.

_________________
Dammy

@Leo

Quote:

Leo wrote: @HenryCase: have you actually read the BarrelFish pdf ? It's an experiment OS written with a lot of concepts in mind... What makes you think AROS follows that ? As for memory protection, I keep on seeing people believing it can be done. Yet no one ever did it... On the other side, I see a lot of people explaining why it isn't possible... You cannot forever extend a 20 years old OS... even Microsoft changed.. oh, guess what ? Even Apple gave up on improving classic MacOS. I don't think anyone here can do better than Apple + Microsoft's hundreds of engeeners... You're free to think the contrary though. I don't...

The Barrelfish project is an example of a multikernel OS. I pointed it out as I thought you might be unfamiliar with the concept. in any case, there are no issues I know of with getting AROS SMP via the multikernel approach, and this approach will not require throwing away most code as you were implying before, AROS will be ready for it when devs get around to it.

As for memory protection, there is no belief necessary, I already pointed you to the proof in my last post.

I accept that these features will be important at some point (of course it would be nice to have now if we had more devs, even just to stop the whining), but what I take objection to is people who spread myths that AROS will never have these features without massive reorganisation, despite the evidence to the contrary.

AROS is not a 20 year old OS any more than Windows is, AROS is more advanced than AmigaOS 3.1 in many ways, alright so it's not complete yet, but that doesn't mean you can't recognise it's achievements so far.

Quote:

AROS is not a 20 year old OS any more than Windows is, AROS is more advanced than AmigaOS 3.1 in many ways, alright so it's not complete yet, but that doesn't mean you can't recognise it's achievements so far.

AROS is as old as AmigaOS 3.1 is on a lot of points: no SMP, no memory protection, no resource tracking. On these points, it's as advanced as Windows 95 was. Same as MacOS classic.

We'll see if it's possible, and if it will be done. But I'm afraid there won't be much people to see/believe if it ever happens...

_________________
http://www.warpdesign.fr/

@damocles

Quote:

damocles wrote: Go review the threads, especially when AROS64 development was under way. AROS64 was going to be a fork in plain view, MP and SMP was planned to be added but the 3.1 API folks got upset about it and not enough were helping Dr. Schulz to do a proper SMP/MP AROS64 kernel. It all went rapidly downhill from there, my investment (and Dr. Schulz's time) were wasted.

That time wasn't wasted. AROS64 is still the logical place for that fork to take place, when the time is right.

The failure with the plan for AROS64 and with Anubis is for exactly the same reason, nobody thought to design a nextgen Amiga-like API BEFORE starting coding. Nobody from what I read of the Anubis discussions had a clear idea of what they were implementing beyond what code they could reuse from elsewhere.

There are certain features that should be new in Anubis, one of which is the API, working out what developers want from an API and designing a suitable API to support these wishes (where possible/sensible) should be the number 1 priority for Anubis right now, if any Anubis fans read this I recommend that is exactly what you do before wondering about silly little details.

Last edited by HenryCase on 28-Mar-2010 at 03:38 PM.

Interesting, but I feel there is something to note here.

Whenever anybody mentions "VESA" drivers, most people seem to assume all you get is a frame buffer and that everything is done by the CPU. Now, that may be true for (really) old VESA 1.0 devices, but VBE/AF (Accelerated Functions) has been around since 1996 and gives a consistent hardware interface for a range of accelerated 2D operations (blitting, block filling, sprites etc.)

Thus, for basic 2D operations then, a properly implemented VESA mode driver shouldn't be any slower than a "full" one, you just don't get any video decoding or 3D functionality.

So, although very nice to see AROS strutting it's stuff, what this video demonstrates to me is that a 3GHz processor has ample horsepower to decode several video streams concurrently. I hate to say this, but that isn't really something I didn't know already.

_________________
Doing stupid things for fun...

@Leo

Quote:

Leo wrote: Quote: AROS is not a 20 year old OS any more than Windows is, AROS is more advanced than AmigaOS 3.1 in many ways, alright so it's not complete yet, but that doesn't mean you can't recognise it's achievements so far. AROS is as old as AmigaOS 3.1 is on a lot of points: no SMP, no memory protection, no resource tracking. On these points, it's as advanced as Windows 95 was. Same as MacOS classic.

It is also more advanced on a lot of points, full Mesa support, decent USB stack, HD audio support, etc...

Quote:

Leo wrote: We'll see if it's possible, and if it will be done. But I'm afraid there won't be much people to see/believe if it ever happens...

Time is running out blah blah blah, for whom though? People who develop and use AROS or any other alternative operating system do so because they enjoy it, it's possible to enjoy it in its current state and the more mature it becomes the more fans it gains. The way you describe it it's like an OS isn't something anyone should use until it has SMP, resource tracking, etc... to that I say no, AROS will continue to grow in popularity, and by the time those features come along it will already have an even stronger following. I can think of some exciting developments on the horizon that are likely to draw more people towards AROS, if you'd like examples let me know.

@HenryCase

There are several things in AROS favour. It's open source, it doesn't require a hardware purchase, just download and run in Virtual Box, it's retro flavoured.

All in in I'll wager that 15 years from now it will be the sole Amiga survivor and the three people left will be moaning about how Amiga could have survived if not for Hyperion's mistakes. There is no winning strategy, Amiga Inc proved that even not playing you still lose....

@persia

Why all the preaching of open source?????? Even something is open source doesn't automaticaly mean it will be a huge success.

_________________
I'm an antique. Don't light my fuse