NVIDIA Says It Has No Plans To Support Wayland

[bridgman]

only HD5xxx/6xxx are not supported by open-source drivers now, and that could change for HD5xxx by the end of this calendar year.

HD5xxx (aka Evergreen) is already supported; Alex & Richard are working on HD6xxx now.

[smitty3268]

Finding Intel there is no surprise (largely because of Gallium3D, which has a lot in common with Wayland); it's the presence of AMD that's the Left Field Event.

Part of what is causing nVidia's reticence is that they remain hostile to Gallium3D (a solid support base for Gallium3D is pretty much a necessary for Wayland) - while AMD was late, at least they were (and are) there. Whether or not it will come back to bite them is in the hands of nVidia.

Uh, Intel doesn't support Gallium3D and as far as i know has no plans to ever do so in the future. They seem to like their classic driver, for whatever reason, and have been pretty dismissive whenever someone has mentioned Gallium.

[PGHammer]

Uh, Intel doesn't support Gallium3D and as far as i know has no plans to ever do so in the future. They seem to like their classic driver, for whatever reason, and have been pretty dismissive whenever someone has mentioned Gallium.

Now you *could* have stuck AMD (not Intel) into that sentence as little as a year ago. Do you remember that big dustup over a massive rewrite of Mesa and the compositing layer of X.org (GEM vs. TIM)? That was the dustup that gave birth to Gallium3D (and eventually to Wayland as well); Intel was backing one, while AMD (and nVidia) were backing the other. However, that wasn't even the biggest reason why AMD had been getting whacked - the bigger reason was the lack of open-source support for (at the time) the HD series GPUs - the only solid support required the binary-blob (which wouldn't work on every Linux distribution extant). It is precisely that whacking that has me classifying AMD's about-face (especially in terms of Gallium 3D and Wayland) as a Left Field Event.

Intel likes the classic driver for X (because X.org did *not* adopt the model that Intel wanted in the GEM vs. TIM dustup). Gallium doesn't care about esoterica like that (because it's abstracted) - Wayland doesn't, either. Intel's graphical hardware supported Gallium3D before AMD's hardware did (primarily because the big driver for Gallium3D was portable computers, where, until recently, Intel, not AMD, had the largest GPU presence).

[bridgman]

AFAIK the Gallium3D transition was largely independent of the GEM/TTM discussions... the only real dependency was a decision to only implement the Gallium3D stack over DRI2, which had a dependency on GEM/TTM, which in turn was implemented only for KMS systems.

The "radeon rewrite" initiative was primarily to let one set of userspace code work with both DRI1 and DRI2/GEM/TTM/KMS, by introducing a new abstraction layer for command submission and buffer management.