NVIDIA Wants Feedback On Its Device Memory Allocator Project

After apologizing how they handled the EGLStreams proposal for NVIDIA Wayland support, James Jones of NVIDIA is trying to get the development of their proposed generic device memory allocator library back on track.

Back during XDC2017 was their latest update on their proposed device memory allocation API that they hope will work out for both them and other upstream parties as an alternative to GBM (Generic Buffer Manager) and to be agreed upon universally for use by Wayland compositors. NVIDIA has done most of the prototyping so far and this new library has yet to be implemented by any of the drivers, including their own, and not yet supported by any of the Wayland compositors as it's still in a relatively early state.

With the prototyping nearing completion, NVIDIA is looking to move this forward. Among the paths that could be taken now include moving ideas from this generic allocator back into the design of GBM (basically what could become a "GBM 2.0"), developing a completely new library to replace GBM, or developing a new library that sits besides/on-top-of GBM for in turn doing the low-level graphics buffer allocations.

NVIDIA has also made it clear they will work out an implementation of whatever comes about for Nouveau. James Jones wrote, "I'd like to note that NVIDIA is committed to providing open source driver implementations of these mechanisms for our hardware, in addition to support in our proprietary drivers. In other words, wherever modifications to the nouveau kernel & userspace drivers are needed to implement the improved allocator mechanisms, we'll be contributing patches if no one beats us to it."

More details via this mailing list post.