Linux Developers Still Reject NVIDIA Using DMA-BUF

[droidhacker]

WAAAH WAAAH!!!! Linux won't give nvidia special treatment for being a bag of douche WAAAAAAAH!!!!

How about this; instead of being stupid about it, TAKE IT UP WITH NVIDIA.

[droidhacker]

What do you think, can we expect a comment of Torvalds on that matter?

Not sure what he would say. Isn't he more on the liberal side concerning the blobs?

I think we can predict what his position will be;

[pingufunkybeat]

At the end of the day, it appears Optimus will remain a Windows-only feature for the immediate future. (Not sure if it's available under OS X.) Chalk up yet another thing in the long list of items that Windows does better than Linux.

This is sad but true.

The good news is that most major players: Intel, Red Hat, AMD, etc. are working on a solution for this problem.

If I'm not mistaken, Nvidia's only contribution to this joint effort was to try to relicense other people's code. They don't want to take part, just use other people's work. That's what this article is about.

I bought a notebook yesterday. It had an option of having an Nvidia GPU (Optimus). I avoided it in a wide arc and went with Intel graphics.

[ryao]

This is sad but true.

The good news is that most major players: Intel, Red Hat, AMD, etc. are working on a solution for this problem.

If I'm not mistaken, Nvidia's only contribution to this joint effort was to try to relicense other people's code. They don't want to take part, just use other people's work. That's what this article is about.

I bought a notebook yesterday. It had an option of having an Nvidia GPU (Optimus). I avoided it in a wide arc and went with Intel graphics.

When will we be able to see optimus support on other operating systems like Illumos, Solaris, FreeBSD, Haiku, Darwin, etcetera?

[Asariati]

If I'm not mistaken, Nvidia's only contribution to this joint effort was to try to relicense other people's code. They don't want to take part, just use other people's work. That's what this article is about.

No, it is not.
The change is done to make sure the Nvidia driver able to load their kernel module. This has nothing to do with "stealing other people's work".

Anyway, the news is too bad. For Linux users. The kernel devs are just doing bullshit this time.

[dargllun]

Lol I knew someone would say something. I have an AMD APU I have HDMI audio support and video acceleration (whatever xbva can afford, for instance use with VLC).

I've been struggling with AMD Video acceleration and HDMI audio for many many many years (mostly using MythTV, but also with XBMC w/ and w/o xvba. My final verdict is: it's all crap compared to a system with an VDPAU capable nvidia card. As much as I hate me saying that given all the work I put into it. Want some facts? Sure:

Blob:
- I happen to have an onboard RS780 (r600g) chip. The closed driver for this HW has been discontinued, so xvba is no option anymore.
- XvBA only partially worked with special xbmc builds. Even the devs themselves complete to AMD for not releasing what they need to fully implement everything
- non-existent ELD support (what's /proc/asound/card0/eld* on your system)

OSS:
- Xv is the only decent video acceleration. Not so great for h264 content. To be discontinued in MythTV in the not-to-far future.
- OpenGL video? Eats even more CPU than Xv. Impossible to get judder free.
- VDPAU video? Eats even more CPU than Xv. Impossible to get judder free. Only partial VDPAU, eg. no deinterlacing
- partial ELD support, but nowhere near what nvidia has, leading to the fact that you can only either
use stereo PCM _or_ 6ch PCM and needing sw upmixing, cause media software is unable to properly understand the audio HW capabilities. Yes, I tried .asoundrc to fix this but it never really worked.

Now take that aging NV GS8400 I finally bought for like 20EUR: everything just works. No magic needed.

As of today, AMD is nearly irrelevant in any of the major media center/DVR camps (mythtv, vdr). Everybody there will tell you to buy nvidia. I learned the hard way why they are right.

[pingufunkybeat]

No, it is not.
The change is done to make sure the Nvidia driver able to load their kernel module. This has nothing to do with "stealing other people's work".

Anyway, the news is too bad. For Linux users. The kernel devs are just doing bullshit this time.

I didn't say "steal other people's work", I said "use other people's work". DMA-BUF is other people's work.

Nice try.

[blacqwolf]

So how did they do it in Windows? Does the Windows kernel offer the technology required for implementing it or is it all in the drivers?

It is mentioned in nVidia's Optimus Whitepaper PDF that the Optimus drivers in Windows relies on a capability in Windows 7 or later to be able to natively detect and utilize multiple cards of different types that share their buffer. This is also why it's impossible to get Optimus machines utilizing both cards on Vista and XP.

[Asariati]

I didn't say "steal other people's work", I said "use other people's work". DMA-BUF is other people's work.

Nice try.

Ok, right. But what did I try?

[boast]

How do we know and trust that the nvidia driver doesn't do something horribly dangerous?

How about giving people the CHOICE to decide for themselves? Or do you force your views on everyone?