NVIDIA Provides RandR Patch For Border Property

[phoronix]

Phoronix: NVIDIA Provides RandR Patch For Border Property

Following a proposal earlier this summer by NVIDIA to extend the RandR protocol, they have now produced a patch for the X.Org Server that adds border property support to the RandR (Resize and Rotate) extension...

http://www.phoronix.com/vr.php?view=OTg2MQ

[timofonic]

Ironically, things change faster with important binary blob drivers than the open source ones.


My nerd writing, "We are waiting"...


We are waiting the efforts in both Nouveau and ATI open source drivers.

We are waiting more drivers for embedded platforms, specially graphics drivers from powerful SoCs and such.

We are waiting a single memory management system flexible for graphics drivers and SoCs, TTM/GEM is not alone but a bunch of own solutions for all kind of embedded platforms out there.

We are waiting for Gallium3D to comply with the promised benefits, having a high level of feature set and efficiency and adopted by the rest of supported graphics drivers.

We are waiting for NVIDIA to stop acting like bastards and support aggresively the Nouveau project.

We are waiting to see ATI focus more stonger on open source graphics drivers.

We are waiting to stop the promotion of binary blobs and make policy of distributions stronger, and of course Force companies to providing source code by putting more stones in the path.



The open source ecosystem is losing power on stupid battles against desktop environments, distributions or forks. The real issue is the hardware drivers and attached frameworks like OpenGL (supporting latest OpenGL MUST be a top level priority just after hardware drivers and productivity/industrial software).

What's happening to Linux? Something is wrong in the community, this needs to change...

We need proper open source drivers for PowerVR, NVIDIA, ATI, Intel, Samsung, ARM... products.

Do we need another X fork to change this? I think an AGPLv3 fork would be very positive in mid term. It would make binary blob drivers to use unsupported old Xorg versions and eventually think about open sourcing drivers.

[efikkan]

I'm waiting for people to stop nagging nVidia and realize they are doing us a great favor by offering enterprise quality drivers, even though some features are missing.

[TheCycoONE]

And I've been waiting for overscan support in xrandr for awhile now. Though I think there was already an announcement about this or something very similar on Phoronix before.

Edit: Yep - http://phoronix.com/forums/showthrea...xtending-RandR

[FireBurn]

I use Xrandr to switch off overscan on the xf86-video-ati driver when using HDMI

[RealNC]

What's happening to Linux? Something is wrong in the community, this needs to change...
We need proper open source drivers for PowerVR, NVIDIA, ATI, Intel, Samsung, ARM... products.

There's nothing happening to Linux. It always was like this. No driver ABI means high maintenance costs for drivers. Waiting for open source drivers means in turn that good working drivers will mostly only come for old hardware most of the time.

On the other hand, putting a driver ABI in the kernel would mean high maintenance for kernel devs, so they won't do it. In other words, don't expect results any time soon. Linux will be the "old hardware OS" for a bit longer yet.

Do we need another X fork to change this? I think an AGPLv3 fork would be very positive in mid term. It would make binary blob drivers to use unsupported old Xorg versions and eventually think about open sourcing drivers.

First, you cannot relicense other people's work as you see fit. It's illegal. The XFree86 fork did not relicense anything; they used a version of XFree86 that was using the older license.

Second, even if it were possible, forks happen because developers are unhappy about a project and want changes. Developers are not unhappy about X.Org.

Third, even if it were possible to relicense with a fork and a group of devs would even do it, it would not help at all. There is only one probable outcome: Distros would not use the new version since binary drivers wouldn't work with it, meaning the fork is doomed to obscurity.

Drivers should not be required to be open source for them to work. Linux's inability to provide a working binary driver infrastructure is not a strength, it's a weakness. Even though he means well, don't let Greg Kroah Hartman's opinions get to you. That model only gets so far on the Desktop. And without cooperation from vendors, you're stuck with writing your own drivers. For most hardware, it's doable. For GPUs without released specs, it's very difficult. For GPUs with released specs, it's still a hard job. And even then, look at the performance and missing features. And then there are the patents too.

Also, lets not forget what would happen if NVidia or AMD would tomorrow release the sources of their drivers. They would not be included in the kernel nor X.Org because - with a probability of 99.9% - they won't meet the quality standards and will contain too much Windows-specific infrastructure in them. Most people seem to think that simply releasing code is all that's needed. No, it's not. Look at Google's Android code for a good example of what happens. NVidia and AMD will not be willing to spend tons of resources to develop their drivers in a manner that is suitable for inclusion into the kernel and X.Org. They would practically need to rewrite it. This is a major undertaking with too little gain for them. NVidia and AMD aren't "non-profit organisations". AMD's open source efforts didn't manifest because they're "such nice folks." There was demand by OEMs for this. In other words, hard facts about hardware sells and profits. And when there's facts, this enables them to calculate costs/gains ratios. Those ratios look pretty bad when thinking about rewriting and releasing the sources of their binary drivers.

[marwi509]

I'm waiting for people to stop nagging nVidia and realize they are doing us a great favor by offering enterprise quality drivers, even though some features are missing.

I guess it depends on how you see it. I always buy nVidia cards (i.e. I'm a paying customer) and I will keep buying them because of their general quality and their great linux support. I don't see any "favors" in that. ^^

[monraaf]

There's nothing happening to Linux. It always was like this. No driver ABI means high maintenance costs for drivers. Waiting for open source drivers means in turn that good working drivers will mostly only come for old hardware most of the time.

On the other hand, putting a driver ABI in the kernel would mean high maintenance for kernel devs, so they won't do it. In other words, don't expect results any time soon. Linux will be the "old hardware OS" for a bit longer yet.

Doesn't Intel already provide Linux drivers for their yet to be released next gen hardware? And AMD has also almost caught up. They are just missing HDMI audio support for their newer GPUs, which I understand is upcoming, and they need to do something about hardware accelerated H.264 decoding and some more work on power management, and then the drivers are probably good enough for the majority of Linux users. I'd say we're 80% there.

[Licaon]

mmm, too bad Xorg 1.11 just br0ke some ABI stuff and the nVidia driver is trippy

>glxinfo
name of display: :0
X Error of failed request: BadValue (integer parameter out of range for operation)
Major opcode of failed request: 135 (GLX)
Minor opcode of failed request: 3 (X_GLXCreateContext)
Value in failed request: 0x2d
Serial number of failed request: 27
Current serial number in output stream: 28
[1] 14731 exit 1 glxinfo

AaronP of nVidia:

Unfortunately, there was a change to a data structure without a corresponding change to the extension module ABI version that breaks GLX. It's not exactly clear whether this structure counts as part of the extension ABI, so I need to get to the bottom of that before a driver with official xserver 1.11 support can be released. It may be that we need to skip xserver 1.11.0 if my hunch is correct and this is a bug in the X server.

It's not the ABI change itself that's the problem, it's that combined with the fact that the extension ABI version number (which is the one libglx looks at) was not changed at the same time. I sent out a change to do that, but it's too late for the 1.11.0 release. Fortunately, working around the problem in libglx shouldn't be too hard and official support should appear in a release hopefully fairly soon.
http://lists.x.org/archives/xorg-dev...st/024752.html

All the public released drivers still have that version marked as unofficial. You'll need to wait for a future release.

@

[Yfrwlf]

There's nothing happening to Linux. It always was like this. No driver ABI means high maintenance costs for drivers. Waiting for open source drivers means in turn that good working drivers will mostly only come for old hardware most of the time.

I'll probably never understand this as I can't fathom how it'd ever be impossible to create a standardized language/ABI which is open to future improvements while retaining backwards compatibility. OpenGL 4.0 exists, why can't Linux Graphics Driver ABI 4.0 exist? You make that part of the kernel, and then you can use any driver that supports LGDABI 4.0 or older. :P This would allow drivers to be shipped along with hardware and give users, developers, and hardware vendors more freedom and flexibility.

Standards = freedom = good.