Tear-Free Acceleration For ATI EXA, Xv

[Dieter]

>> The closed driver was primarily aimed at the workstation market,
>> where most of the users come from the Unix world and so open
>> source is really not a concern.

Sorry, closed drivers *are* a problem, not just in linux, but also
with real Unix.

> NVidia doesn't seem to be having problems with a closed driver;

Nvidia's closed drivers have bugs, including security holes.
Nvidia's closed drivers cause data loss.

Without source code these problems are effectively impossible to fix.

Binary-only drivers are unacceptable. It doesn't matter if
the chip is AMD/ATI or Nvidia or VIA or whoever's. It doesn't
matter if the OS is linux or real Unix or Plan-9 or whatever.

Binary-only drivers are unacceptable. PERIOD. How many times do
we have to say it?

> Stop defending your worthless closed driver, and get your open
> source developers on the same team.

Better yet, stop wasting resources on worthless closed driver(s),
pick one FLOSS driver (*) and have everyone work on that driver.

(*) If it makes sense for, say, R200 and R600 to have different
drivers, that's fine. I mean one driver per generation.

[yotambien]

Sorry, closed drivers *are* a problem, not just in linux, but also with real Unix.

Well, the vast majority of companies out there use Windows and closed drivers and the Earth hasn't apparently exploded yet. Focusing on Nvidia, it can't be that bad when folks are using their tools and hardware to do molecular mechanics and weather simulations among others.

Nvidia's closed drivers have bugs, including security holes.

Of course, because free drivers are also bug free.

Nvidia's closed drivers cause data loss.

This is simply not true. Even if it was, it would have nothing to do with them being closed.

Without source code these problems are effectively impossible to fix.

You mean efectively impossible to fix by somebody outside their party. Nothing impedes them to fix their bugs, and the proof is that they do.

Binary-only drivers are unacceptable. PERIOD. How many times do we have to say it?

It's not how many times and how loud you say it. If you don't back up your position with an argument of some sort it basically goes nowhere. Even if you do it might be possible (marginally) that your interests do not equal, say, AMD's interests.

Stop defending your worthless closed driver, and get your open source developers on the same team.

Their 'worthless closed driver' is the only way many people has to play games at any half decent frame rate, or to run some 3D applications at its full power (*). Is fglrx crap? In my experience, yes; but wiping it out as suggested before is non sense. Fortunately for AMD's stakeholders, their guys know better than following the advice posted at some random forum.

You still haven't got that the open source drivers are not going to catch up with the closed driver in terms of 3D performance. No matter they themselves said it here, the mantra goes on and on. I have the feeling that some of this open source loving (I'm not pointing at you two, guys) comes more from the frustration that fglrx has brought to many linux users than to anything else.


(*) I can't run properly VMD with the open source drivers, and couldn't run Gtkradiant at all until some months ago. Given that I count with three fingers the number of 3D apps I use, the percentage is rather bad (_in my case_).

[rbmorse]

Better yet, stop wasting resources on worthless closed driver(s),
pick one FLOSS driver (*) and have everyone work on that driver.

I don't think that's something ATI controls. They contribute to the radeohHD effort, but it's run by Novell. I'm not sure what ATI's relationship is with "radeon" outside of moral support and possibly some technical assistance.

I'm sure ATI has some influence, but I doubt John or anyone else at ATI is in a position to call one of the open-source driver teams and tell them to cease what they're doing and just work on the other project. ATI might suggest that happen...ATI might have already suggested that happen, but I suspect the personal and theological divides are still too great and ATI cannot force it to happen. And, probably shouldn't, all things considered.

I'm very sure John isn't in a position to make the case to his management that they need to quit working on FGLRX and devote those resources to the open driver. Because like it or not, there's a butt-load of revenue tied to the CAD/Workstation market and almost zero revenue potential in gaming on Linux, which is what you want to see developed.

I'm as sorry as anyone else that Radeon/FGLRX doesn't work any better than it does with games, but I understand why it's not job #1.

FireGL/FGLRX is a pretty good solution for the CAD/Technical workstation segment. That's one area were a lot of people are betting it's going to be possible to make money with Linux-based solutions. It's a (comparatively) huge market and there are a lot of seats at stake. Have you priced a FireGL card, recently?

Look at what happened at SIGGRAPH this year. The Khronos Group threw the gamers under the bus (again) in favor of the development track pushed by the CAD/Workstation faction. When you look at who's funding Khronos that's no surprise, and it's pretty clear as to who's going to control the development of OpenGL for the forseeable future. I don't think they care much about wobbly windows. Or accelerated 3D.

That said, I do think ATI is coming around to the realization that even though there isn't much direct revenue to be had from Desktop Linux it's still something they ought to do better at. Pride if nothing else. And I think they decided to go open-source because it's a perfect approach for what they need to do...it doesn't cost much and what costs there are get split with others. ATI won't incur any support or maintenance liability and that is an important consideration if you don't expect the product to generate any revenue.

The drawbacks are what we see, of course.

[Dieter]

>> Well, the vast majority of companies out there use Windows
>> and closed drivers and the Earth hasn't apparently exploded yet.

Lots of people do lots of stupid things and the Earth hasn't
exploded. But the Earth would be a much better place to live
if people made better choices.

>>> Nvidia's closed drivers have bugs, including security holes.

>> Of course, because free drivers are also bug free.

Compare OpenBSD's security record with ms-virus-server's.

>>> Nvidia's closed drivers cause data loss.

>> This is simply not true.

What, if something hasn't happened to you personally, you think
it hasn't happened to anyone?

>> Even if it was, it would have nothing to
>> do with them being closed.

The inability for users to fix it has everything to do with
not having source.

>> Their 'worthless closed driver' is the only way many people has
>> to play games at any half decent frame rate

If you think that playing games is more important than security and
data integrity you can expect a very bad experience in your future.

>> You still haven't got that the open source drivers are not going
>> to catch up with the closed driver in terms of 3D performance.

Personally, I don't expect to need *any* 3D performance in
the foreseeable future.

> almost zero revenue potential in gaming on Linux, which is what
> you want to see developed

I don't run linux, I run real Unix. I couldn't care less about
silly 3D gaming. What makes you think I care about gaming on linux?

> Have you priced a FireGL card, recently?

Not the GL, but the "FirePro V3700 will cost a mere $99 USD"

http://www.phoronix.com/scan.php?page=news_item&px=NjY0NA

> ATI is coming around to the realization that even though there
> isn't much direct revenue to be had from Desktop Linux it's
> still something they ought to do better at

There is more to revenue than direct revenue. There is more to
FLOSS than just linux. There are more markets than just CAD
workstations and desktop. Big markets.

[yotambien]

>>> Nvidia's closed drivers have bugs, including security holes.

>> Of course, because free drivers are also bug free.

Compare OpenBSD's security record with ms-virus-server's.

That's one example (from which I don't have and you haven't given any numbers, anyway). I'm sure you could come up with others pointing to the opposite direction and they would all still mean nothing, for just examples they are. But if we want to play this game, _my experience_ is that many open source applications are rather buggy. I use and enjoy them, but I'm not blind.

Nvidia's closed drivers cause data loss.

This is simply not true.

What, if something hasn't happened to you personally, you think it hasn't happened to anyone?

Definitely not, for the same reason that you don't think that what's happened to you must have happened to everybody else. To be honest, I get your point, but you'll admit that saying "Nvidia close drivers cause data loss" is a bit categorical and missleading.

>> Even if it was, it would have nothing to
>> do with them being closed.

The inability for users to fix it has everything to do with
not having source.

Sorry, users don't fix anything. Under a positive ligth, they use the software and report problems to developers. Under a negative one, they complain without having a clue. That there are sources out there doesn't mean that somebody is going to hack the hell out of them to bring the ultimate product to the masses; even less in a timely manner.

If you think that playing games is more important than security and
data integrity you can expect a very bad experience in your future. Personally, I don't expect to need *any* 3D performance in
the foreseeable future.

Of course I don't think so, and my 3D needs are very modest too. That's why I run the oss driver most of the time, because it works (*). When time ago Bridgman explained the strategy of having a decent, more or less feature complete oss driver and an optimised, fine tuned closed one I thought "bollocks". My first reaction was the typical defensive position. Then I realised that I was already doing exactly that, picking what best suits me according to my needs.

Oss developers are constantly adding more features and improving their drivers, while AMD's released the docs for their cards. I'm not sure what else is wanted. Sure, if you want it _here_ and you want it _now_ and under the conditions that _you_ happen to prefer, you're gonna get frustrated no matter what.



(*)Being rather busy at the moment I haven't play at all for a good while, meaning no fglrx. My laptop's uptime had never been this high.

[duby229]

I don't think that's something ATI controls. They contribute to the radeohHD effort, but it's run by Novell. I'm not sure what ATI's relationship is with "radeon" outside of moral support and possibly some technical assistance.

I'm sure ATI has some influence, but I doubt John or anyone else at ATI is in a position to call one of the open-source driver teams and tell them to cease what they're doing and just work on the other project. ATI might suggest that happen...ATI might have already suggested that happen, but I suspect the personal and theological divides are still too great and ATI cannot force it to happen. And, probably shouldn't, all things considered.

I'm very sure John isn't in a position to make the case to his management that they need to quit working on FGLRX and devote those resources to the open driver. Because like it or not, there's a butt-load of revenue tied to the CAD/Workstation market and almost zero revenue potential in gaming on Linux, which is what you want to see developed.

I'm as sorry as anyone else that Radeon/FGLRX doesn't work any better than it does with games, but I understand why it's not job #1.

FireGL/FGLRX is a pretty good solution for the CAD/Technical workstation segment. That's one area were a lot of people are betting it's going to be possible to make money with Linux-based solutions. It's a (comparatively) huge market and there are a lot of seats at stake. Have you priced a FireGL card, recently?

Look at what happened at SIGGRAPH this year. The Khronos Group threw the gamers under the bus (again) in favor of the development track pushed by the CAD/Workstation faction. When you look at who's funding Khronos that's no surprise, and it's pretty clear as to who's going to control the development of OpenGL for the forseeable future. I don't think they care much about wobbly windows. Or accelerated 3D.

That said, I do think ATI is coming around to the realization that even though there isn't much direct revenue to be had from Desktop Linux it's still something they ought to do better at. Pride if nothing else. And I think they decided to go open-source because it's a perfect approach for what they need to do...it doesn't cost much and what costs there are get split with others. ATI won't incur any support or maintenance liability and that is an important consideration if you don't expect the product to generate any revenue.

The drawbacks are what we see, of course.

Your position is one that I just cant understand... You seem to think that the closed driver is required for this market but in --reality-- a stable and feature complete open driver would serve far better...

Your reasoning for forcing an unstable, buggy, slow, non-standard closed driver on us in totally unacceptable. It has no weight, and is an elaborate excuse at it's finest.

[mityukov]

Your position is one that I just cant understand... You seem to think that the closed driver is required for this market but in --reality-- a stable and feature complete open driver would serve far better...

I don't think somebody thinks differently. But you should take into account that ATI/AMD has it's own interests.

For instance, they reasonably don't like to open code, used in their proprietary driver, though that could the best option for the driver's development -- far more people could participate in its improving, in this case.

On the other side, there are many "tricks", to which AMD spent lot of money, and every competitor will then be able to use their code for free.

What does it means in business? I'm not economist, but "lower expenses" is a key factor in competing. If AMD allows, let say NVidia, to save some money/time on some research and development, to which ATI already spent -- it's bad for ATI (and for us, actually, as for consumers of production from two hardly competing manufacturers).

I personally think, blaming ATI for that is not rational..

[rbmorse]

Your position is one that I just cant understand... You seem to think that the closed driver is required for this market but in --reality-- a stable and feature complete open driver would serve far better...

Your reasoning for forcing an unstable, buggy, slow, non-standard closed driver on us in totally unacceptable. It has no weight, and is an elaborate excuse at it's finest.

I'm not saying this is what I want, but I do believe it is what is happening. I'm not the one who thinks that a closed driver is required...but ATI and nVidia seem to think so.

I happen to agree that a stable and feature complete open driver is desirable, but I don't know who's going to pay for the full-scale development effort you want. Certainly not the Linux gaming crowd.

Frankly, all I see from you is a bunch of rude sanctimony based upon a profoundly ignorant view of the industry and some bad religious dogma.

My view may in fact be "unacceptable", but let's see a business case that might influence the businesses involved to change their thinking.

[Dieter]

>>>>> Nvidia's closed drivers have bugs, including security holes.
>>>> Of course, because free drivers are also bug free.
>>> Compare OpenBSD's security record with ms-virus-server's.
>> That's one example (from which I don't have and you haven't
>> given any numbers, anyway).

"Only two remote holes in the default install, in more than 10 years!"

http://www.openbsd.org/

I don't have numbers for ms-virus-server, MAXFLOAT is only 3.40282346638528860e+38

>> To be honest, I get your point, but you'll admit that saying "Nvidia close
>> drivers cause data loss" is a bit categorical and missleading.

New version: Nvidia's closed drivers sometimes cause data loss.
Better?

>> Sorry, users don't fix anything.

Some users do. I have fixed bugs in lots of programs.
Users that don't have the skills to fix bugs themselves can
ask for help, or hire someone to fix it for them. None of
those options are available if the user doesn't have source.

------

> I happen to agree that a stable and feature complete open driver is
> desirable, but I don't know who's going to pay for the full-scale
> development effort you want.

Take the resources used for closed driver and apply them to FLOSS
driver.

> My view may in fact be "unacceptable", but let's see a business case
> that might influence the businesses involved to change their thinking.

Good quality FLOSS driver -> more sales.

[stan]

On the other side, there are many "tricks", to which AMD spent lot of money, and every competitor will then be able to use their code for free.

You can still open up your code while ensuring that you're not giving your competitors a free ride: it's called CopyLeft. If AMD wants to prevent NVIDIA from stealing code, AMD should license the driver GPLv3 so that if NVIDIA uses the code they'll have to open up their entire code base. Any improvement that NVIDIA makes would be then available for AMD to incorporate.

it's bad for ATI (and for us, actually, as for consumers of production from two hardly competing manufacturers).

Sorry, but that's baloney! GNOME, KDE, XFCE, Enlightenment etc are all open source, yet they still compete fiercely with each other for developers and users.