What is the future of AMD/Nvidia drivers?

Here you can discuss ReactOS related topics.

Moderator: Moderator Team

mrugiero
Posts: 482
Joined: Sun Feb 14, 2010 9:12 am

Re: What is the future of AMD/Nvidia drivers?

Post by mrugiero »

Z98 wrote:Because projects that are Linux centric have a history of pretending nothing else exists even when they offer "cross platform" support. They very often try to pretend that everything is the Linux variant of Unix, regardless of whether the platform they're running on is Windows or BSD or Linux. The BSDs have been dealing with that kind of crap for years now when proposed APIs and extensions are made that rely on Linux specific behavior and functionality even if they're ostensibly to allow for portable management of agnostic Unix systems.
I just don't think that's the case for mesa. At the very least, Haiku is able to use it. Also, the abstraction is not only useful to support several OSes, but it's useful for them to be able to change the driver model from time to time, which they do, and if it's true that vmware bases its Windows driver on it (I can't really tell, it's just something I read on Phoronix forums once, and as the source code is not in the main tree, I can't really prove or rebute that), then I can see why they'd care on keeping the abstraction in shape.
Though this is all a moot point since there's no technical reason for the project to invest the resources to try porting Gallium.
I never proposed that idea, and I actually discouraged it. The fact I think it's difficulty is partly alleviated thanks to Gallium, I think porting it to Windows (and ReactOS) would be a waste of time, except for some individual interested in supporting an older piece of hardware that's been dropped from the binary drivers or for which drivers for newer Windows don't exist, and even in those cases the advantage is not really that clear.

erkinalp
Posts: 858
Joined: Sat Dec 20, 2008 5:55 pm

Re: What is the future of AMD/Nvidia drivers?

Post by erkinalp »

There is no legal obstacle on ReactOS' side to distribute closed source drivers. We'd just need to talk to the companies in question and I doubt they'd object since we're not modifying any drivers, just bundling them. As such, there's not any specific need to require that all drivers that are bundled bye open source.
Then the "fully reproducible builds" advocates would come into scene and say that ReactOS is not free even in the sense of the beer.
-uses Ubuntu+GNOME 3 GNU/Linux
-likes Free (as in freedom) and Open Source Detergents
-favors open source of Windows 10 under GPL2

User avatar
Black_Fox
Posts: 1584
Joined: Fri Feb 15, 2008 9:44 pm
Location: Czechia

Re: What is the future of AMD/Nvidia drivers?

Post by Black_Fox »

erkinalp wrote:Then the "fully reproducible builds" advocates would come into scene and say that ReactOS is not free even in the sense of the beer.
Well, the drivers are free as in beer and so would be also ROS if bundled with them. Anyway, they can be added just as another item of ReactOS Application Manager, so the fully reproducible builds can be kept.

fred02
Posts: 551
Joined: Thu Nov 22, 2007 5:54 pm

Re: What is the future of AMD/Nvidia drivers?

Post by fred02 »

Sorry, but I don't understand the point of this whole discussion. :?
As I see it, there is no need to include video (or other drivers, aside a bare minimum) with ROS sources nor make a "distribution" that bundles them, since NT architecture has an API and a DDK/WDK for writing drivers. So if someone want to try himself on the task, he can start right now and test it with w2k, XP and w2k3. Once ROS has the necessary parts correctly implemented it will just work with it too. Furthermore, drivers can be distributed under any licence, be it GPL, BSD, public domain or proprietary, since it is not part of the OS.

mrugiero
Posts: 482
Joined: Sun Feb 14, 2010 9:12 am

Re: What is the future of AMD/Nvidia drivers?

Post by mrugiero »

fred02 wrote:As I see it, there is no need to include video (or other drivers, aside a bare minimum) with ROS sources nor make a "distribution" that bundles them, since NT architecture has an API and a DDK/WDK for writing drivers.
It's not about the need, it's about considering it practical. Some people considers practical to have a working setup (in most cases, I wouldn't consider a system using the VESA driver a working setup) right after installing the OS. Of course, I would update it right away, so it wouldn't make a big difference to me.
So if someone want to try himself on the task, he can start right now and test it with w2k, XP and w2k3.
Errm, I don't know what that has to do with the discussion. The idea is not porting drivers to ReactOS (that's against the policy of aiming to full driver compatibility) but to Windows. Someone suggested maybe having open source drivers is a good idea. I like them and I use them on Linux, but I don't really see the point of such "purism" on ReactOS, since most of the things that could motivate you to use it are closed source, and most of the things using acceleration on Windows use Direct3D, which open source drivers (except if you want to write all of it from scratch or maintain the failed attempts) don't even try to provide.
Once ROS has the necessary parts correctly implemented it will just work with it too. Furthermore, drivers can be distributed under any licence, be it GPL, BSD, public domain or proprietary, since it is not part of the OS.
Not really. You can distribute whatever the copyright holder authorizes you to distribute, so no, not "under any license"; a license with a no-redistribution clause can not be distributed. NOW, it is very unlikely that manufacturers stop you from distributing their drivers, as having their hardware to work shows a good side of their work, regardless of the OS, and in the case of ReactOS should imply no extra work from them, so once ReactOS has everything in place and doesn't make them look bad, a brief talk would probably fill all the requirements for enabling redistribution of their drivers.

Webunny
Posts: 1201
Joined: Sat Apr 28, 2012 1:30 pm

Re: What is the future of AMD/Nvidia drivers?

Post by Webunny »

fred02 wrote:Sorry, but I don't understand the point of this whole discussion. :?
As I see it, there is no need to include video (or other drivers, aside a bare minimum) with ROS sources nor make a "distribution" that bundles them, since NT architecture has an API and a DDK/WDK for writing drivers. So if someone want to try himself on the task, he can start right now and test it with w2k, XP and w2k3. Once ROS has the necessary parts correctly implemented it will just work with it too. Furthermore, drivers can be distributed under any licence, be it GPL, BSD, public domain or proprietary, since it is not part of the OS.

Hmm.Well, only to a certain degree. I'll be frank: the moment that ROS auto-recognised/installed the right driver for my networkcard, I thought was a fantastic boon, and immediately made me appreciate the (appearance of) progress of this build. I don't think I'm the only one. Especially for the lay public, something that 'works out of the box' is deemed FAR better and superior then when it doesn't. Right or wrong, that's the impression and perception people will have.

Now, granted, one shouldn't (and can't) include every last driver for any device ever created, that much is clear. But I would *strongly* suggest one includes a dozen or so of the most widely used ones, especially those that became a de facto standard (especially with integrated stuff, like HD audio). And let's not forget, that Windows itself too, auto-recognise a whole bunch of them (or uses generic ones that work on 80% of it). It would be strange to claim, if one wants to be a 100% clone, that ROS shouldn't have any of that, because one can 'download it' or 'make a distribution'. No ordinary user is going to make a special distribution just for him, including all his prefered drivers. That is just IT'ers thinking, not something the lay public wants to do. And yes, they can download it, but so can you on Windows, yet Windows auto-recognises a bunch of them too, regardless. So should ROS.

Even if one wants to keep the basic build of ROS 'pure' without any 'clutter' of drivers, at least in the Community edition, there should be a bunch of (most widely used) auto-recognised drivers. Otherwise, it will be felt by the ordinary user as a 'fail', it's as simple as that.

justincase
Posts: 441
Joined: Sat Nov 15, 2008 4:13 pm

Re: What is the future of AMD/Nvidia drivers?

Post by justincase »

I personally think that more important than including drivers, is to improve the driver compatibility (I get BSOD when I try to use ac97 in QEMU as described both on the wiki, and in rosapps) and installation (i.e. Found new hardware wizard), then making a (hosted) db of hardware & the associated place to download the manufacturers drivers would fill the 'drivers not included' void quite nicely as long as ethernet can be used, and the input/output devices (keyboard, mouse, video card) have working (generic) drivers (like ReactOS already has, e.g. vesa).
If such a database could be realized, the Found new hardware wizard could recommend downloading the manufacturers own driver directly (with an embedded link?) and/or automate installing it. This would be an improvement even over Windows' own implementation. :ugeek:
Last edited by justincase on Thu Apr 24, 2014 8:14 pm, edited 1 time in total.
I reserve the right to ignore any portion of any post if I deem it not constructive or likely to cause the discussion to degenerate.

Webunny
Posts: 1201
Joined: Sat Apr 28, 2012 1:30 pm

Re: What is the future of AMD/Nvidia drivers?

Post by Webunny »

justincase wrote:I personally think that more important than including drivers, is to improve the driver compatibility (I get BSOD when I try to use ac97 in QEMU as described both on the wiki, and in rosapps) and installation (i.e. Found new hardware wizard), then making a (hosted) db of hardware & the associated place to download the correct drivers would fill the 'drivers not included' void quite nicely as long as ethernet can be used, and the input/output devices (keyboard, mouse, video card) have working (generic) drivers (like ReactOS already has, e.g. vesa).
Do you have a debug log for it?

justincase
Posts: 441
Joined: Sat Nov 15, 2008 4:13 pm

Re: What is the future of AMD/Nvidia drivers?

Post by justincase »

Webunny wrote:
justincase wrote:(I get BSOD when I try to use ac97 in QEMU as described both on the wiki, and in rosapps)
Do you have a debug log for it?
Not right now, I'll have to try it again sometime soon though. (Note: I'm using QEMU for windows v1.6.0-sdl w/o fmod)
I reserve the right to ignore any portion of any post if I deem it not constructive or likely to cause the discussion to degenerate.

erkinalp
Posts: 858
Joined: Sat Dec 20, 2008 5:55 pm

Re: What is the future of AMD/Nvidia drivers?

Post by erkinalp »

If such a database could be realized, the Found new hardware wizard could recommend downloading the manufacturers own driver directly (with an embedded link?) and/or automate installing it. This would be an improvement even over Windows' own implementation. :ugeek:
Even if it is possible, it is too hard to keep up with and fully FSF-compliant distributions will crop that out.
-uses Ubuntu+GNOME 3 GNU/Linux
-likes Free (as in freedom) and Open Source Detergents
-favors open source of Windows 10 under GPL2

justincase
Posts: 441
Joined: Sat Nov 15, 2008 4:13 pm

Re: What is the future of proprietary drivers?

Post by justincase »

Perhaps, but if such a system were to be put into action it would (almost) be a necessity for hardware manufacturers to be given the ability to update it (or submit patches for inclusion in the database ;))

Also such a system could be used by Microsoft Windows users with a standalone app.

I don't know how feasible it is (likely not very) but I figure putting the idea out can't hurt, after all the worst thing that could happen (other than nothing) is that somebody tries it :o and fails :shock: (the best case being that they succeed).

Oh, and people using "FSF-compliant distributions" can do without it ... just like on Windows (or use the standalone app made for Windows users if it ever exists).
I reserve the right to ignore any portion of any post if I deem it not constructive or likely to cause the discussion to degenerate.

fred02
Posts: 551
Joined: Thu Nov 22, 2007 5:54 pm

Re: What is the future of AMD/Nvidia drivers?

Post by fred02 »

mrugiero wrote:You can distribute whatever the copyright holder authorizes you to distribute, so no, not "under any license"; a license with a no-redistribution clause can not be distributed.
I meant that a person who would write a driver could distribute it under any licence he/she sees fit AND it will have no incidence on the GPL status of ROS. As to whether a manufacturer can forbid its redistribution, it is again a good feeding ground for lawyers, but considering the NVidia did nothing to stop the development of nouveau, it would be safe to bet that it would not an issue (IANAL).
mrugiero wrote:
fred02 wrote:So if someone want to try himself on the task, he can start right now and test it with w2k, XP and w2k3.
Errm, I don't know what that has to do with the discussion. The idea is not porting drivers to ReactOS (that's against the policy of aiming to full driver compatibility) but to Windows. Someone suggested maybe having open source drivers is a good idea.
I did not express myself clearly. My point was that there is no need to tie driver development to ROS tree or distribution. This is how it works for UniATA and other drivers.

mrugiero
Posts: 482
Joined: Sun Feb 14, 2010 9:12 am

Re: What is the future of AMD/Nvidia drivers?

Post by mrugiero »

fred02 wrote:
mrugiero wrote:You can distribute whatever the copyright holder authorizes you to distribute, so no, not "under any license"; a license with a no-redistribution clause can not be distributed.
I meant that a person who would write a driver could distribute it under any licence he/she sees fit AND it will have no incidence on the GPL status of ROS. As to whether a manufacturer can forbid its redistribution, it is again a good feeding ground for lawyers, but considering the NVidia did nothing to stop the development of nouveau, it would be safe to bet that it would not an issue (IANAL).
You are mixing facts there. NVIDIA can't stop distribution of nouveau, because it is clean room RE, just like ReactOS. At most, they can claim it infringes patents and stop them from distributing binaries of it, but source code is protected by free speech, meaning they can't make them remove or stop distribution of any code NVIDIA doesn't own.
mrugiero wrote:
fred02 wrote:So if someone want to try himself on the task, he can start right now and test it with w2k, XP and w2k3.
Errm, I don't know what that has to do with the discussion. The idea is not porting drivers to ReactOS (that's against the policy of aiming to full driver compatibility) but to Windows. Someone suggested maybe having open source drivers is a good idea.
I did not express myself clearly. My point was that there is no need to tie driver development to ROS tree or distribution. This is how it works for UniATA and other drivers.
There are some drivers that need to be bundled with ReactOS, and having them in the tree (even if it's just a regular sync like they do with WINE) makes it easiear to build the images with it. I mean, UniATA is a driver that is almost surely needed to boot the OS, so you can't expect for the user to install it after boot. It just wouldn't work. However, for most other hardware, I agree with you. Just use generic drivers like VESA by default and put something in place to look for the specific drivers, but there is no need to bloat the images with tons of drivers.

fred02
Posts: 551
Joined: Thu Nov 22, 2007 5:54 pm

Re: What is the future of AMD/Nvidia drivers?

Post by fred02 »

Yes, UniATA is not the best example, but I don't remember what other drivers are bundled with ROS. Some audio driver, I think. :? So I totally agrees with you: there is no point of bundling anything beside bare minimum + network drivers, so that up-to-date versions can be downloaded. Windows is doing it too through Windows Update.

Webunny
Posts: 1201
Joined: Sat Apr 28, 2012 1:30 pm

Re: What is the future of AMD/Nvidia drivers?

Post by Webunny »

fred02 wrote:Yes, UniATA is not the best example, but I don't remember what other drivers are bundled with ROS. Some audio driver, I think. :? So I totally agrees with you: there is no point of bundling anything beside bare minimum + network drivers, so that up-to-date versions can be downloaded. Windows is doing it too through Windows Update.
Not exactly. Windows has a considerably list of drivers it auto-recognises/installs 'out of the box'. Even XP already had that, and if you look at win7, a whole truckload of devices are recognised and the proper drivers installed.

True, they do not provide it for every last NVIDIA or ATI/AMD graph card, and updates are, indeed, often done by downloading it, but acting as if Windows only has a handful of the barest minimum drivers, is simply not correct. If ROS wants to be a clone of winXP and offer the same user-experience, it stands to reason that it should have a similar amount of drivers bundled with it as winXP is/was. I think this is the logical middleground with all the talk about those that want every last driver bundled with it (which is infeasible), and those who would argue it should have next to none.

In reality, it is something in between those two options with all the Windows versions, and I think we should follow that principle.

Post Reply

Who is online

Users browsing this forum: No registered users and 3 guests