NVIDIA To Meet With Wayland, Linux Kernel Developers To Discuss GBM vs. Streams

[Hi-Angel]

I don't get it. Doesn't this means Mesa devs are giving up to proprietary software? If so, why?

The gbm vs streams discussion is not relevant to proprietary at all. It's about pros and cons of two approaches, and cons of NVidia's one are very bad. They should have been participating in development from the very beginning, but instead they came late to a big party, and wondering why so little peoples and everybody drunk.

[bug77]

I don't get it. Doesn't this means Mesa devs are giving up to proprietary software? If so, why?

What do the Mesa devs have to do with anything being discussed here? This is between Wayland implementors only.

[Awesomeness]

So what are compositor devs to do now, spend significant amounts of time to add EGLStreams support, and then be required to maintain two backends even though one of those backends will only be used by a single closed source driver, just because EGLStreams is a "standard"?

Well, NVidia could get into a contractual obligation with Gnome Foundation and KDE e.V. (because let's face it: Mutter and KWin will be the two Wayland compositors regular people will actually encounter, not Weston) to maintain their implementation.

[asdfblah]

The gbm vs streams discussion is not relevant to proprietary at all. It's about pros and cons of two approaches, and cons of NVidia's one are very bad. They should have been participating in development from the very beginning, but instead they came late to a big party, and wondering why so little peoples and everybody drunk.

What do the Mesa devs have to do with anything being discussed here? This is between Wayland implementors only.

Ahh, I guess I'm confused then, I was thinking about some other feature... Sorry.

[finalzone]

Well, NVidia could get into a contractual obligation with Gnome Foundation and KDE e.V. (because let's face it: Mutter and KWin will be the two Wayland compositors regular people will actually encounter, not Weston) to maintain their implementation.

In detriment on maintenance thus bringing unnecessary stress to the developers because a company avoid using existing implementation. EGLStream is very much like the old xorg-x11-drv-nv driver, obfuscation without access to the source and the inability to fix the real problem. History would repeat again which are the reason neither Gnome nor KDE nor other desktop environment developers want to relive again.

[finalzone]

Well, NVidia could get into a contractual obligation with Gnome Foundation and KDE e.V. (because let's face it: Mutter and KWin will be the two Wayland compositors regular people will actually encounter, not Weston) to maintain their implementation.

That method failed with xorg-x11-drv-nv (remember?) where no one were able to fix issues like the basic 3D support Nvidia wasn't interested to do.
Neither Gnome Foundation nor KDE nor other desktop environment developers want to deal with such problem having already established a standard with GBM. If Intel and AMD can use GBM, so can Nvidia. The ball is on their side.
Frankly, Nvidia as PC GPU leader is really disappointing.

[Khrundel]

The gbm vs streams discussion is not relevant to proprietary at all. It's about pros and cons of two approaches, and cons of NVidia's one are very bad. They should have been participating in development from the very beginning, but instead they came late to a big party, and wondering why so little peoples and everybody drunk.

https://www.x.org/wiki/Events/XDC2014/XDC2014RitgerEGLNonMesa/nvidia-and-compositors.pdf

It's 2014 at least. Maybe earlier.

[Khrundel]

Well to be honest both approaches have merits and demerits.
Merits:
EGL streams:
1.) Khronos kinda standard(again like OpenGL many things are veguely defined and can lead to vendor specific issues all over the place)
2.) would allow nVidia to kinda support wayland(apparently with heavy timing issues or hacks) from the user space driver instead of proper atomic KMS

Not a big list, lets enlarge it.
3.) EGL streams is the way for frame producer and consumer to negotiate image format. Imagine some game working in window on wayland compositor.
GBM approach:
Nvidia OpenGL/Vulkan driver renders scene to some oncard framebuffer, using convinient format with compression or something.
Then driver converts it to hardcoded GBM frame and passes to a wayland compositor. Then compositor in 99% of cases just calls same nvidia openGL driver to place a picture to some part of the screen using some transform (scalling, transparency, etc).
EGLStreams aproach:
Nvidia OpenGL/Vulkan driver negotiate with same Nvidia OpenGL/Vulkan about how to pass rendering result and then compositor gets something it don't care about as an EGLImage and passes it to the driver, telling which transformations should be applied.

[Geopirate]

Well part of it is many Wayland devs were/are Mesa devs so doing some things similarly and in a somewhat compatible format made sense. They also had a vested interest in what would become xWayland, which is why we're even talking about wide adoption in the next couple years.

The main problem is that Nvidia was late to the party as several have said. Part of it is they worked on this for at least a year without telling anyone. The solution they presented is one that required significant work on their part before they could present enough of a solution to make this a debate. So planning and development for that is I think at a minimum 2 years before they announced this to the Wayland folks. So if they had included other people in their planning most likely this debate would have happened before Gnome/KDE/Enlightenment started working seriously on making this happen. Instead Nvidia waited until these people had a 80-90% working implementation and now they want everyone else to start from scratch basically.

Nvidia's main reason for wanting their approach is that it works better with their code for other platforms. Again they should have discussed this much sooner. The cost of paying one intern to research other platforms for an hour a day would have been exponentially less than it will cost them to maintain a GBM workaround for at least the next several years.

[Awesomeness]

neither Gnome nor KDE nor other desktop environment developers want to relive again.

They would not relive anything because they wouldn't touch that code. Only NVidia's engineers would as they'd be obligated to maintain it.