NVIDIA Still Working On A Generic Allocator - Has Working Open-Source Implementation

[phoronix]

Phoronix: NVIDIA Still Working On A Generic Allocator - Has Working Open-Source Implementation

For those wondering, NVIDIA is still pursuing a generic allocator / Unix device memory allocator that has been talked about for years and a potential successor to the likes of the Generic Buffer Manager (GBM). They now have an implementation of their proposed allocator working for the open-source NVIDIA (Nouveau) driver though there still is a lot of work ahead...

http://www.phoronix.com/scan.php?page=news_item&px=NVIDIA-Generic-Allocator-2019

[boxie]

oh wow, there's code! about time!

[johanb]

After being quiet for so long I was losing hope for this allocator to ever become reality, I'm happy to hear that I'm wrong.
Well done nvidia!

[polarathene]

Awesome All the best James and anyone else involved towards getting this adopted! Will it also resolve the lack of support with XWayland? From what I've heard the current Wayland support(eglstreams) is unable to use the dGPU if something runs via XWayland.

oh wow, there's code! about time!

There has been for some time now. It was living on Github earlier(stiill technically is): https://github.com/cubanismo/allocator/issues/17

I link to that issue, because James(same maintainer with username cubanismo), just responds by closing it immediately. The repo was not seeing active development there, and it was a legit question for those who had been watching it for some time, kinda rude towards supporters :\

This new Gitlab repo seems to have only been put up in the past week. So I guess development has continued in private. The github repo had development from Sep 2016 to Nov 2017 roughly. Linked issue asking about the project status was Jan 2019. James should probably update that Github README to notify users that the repo is deprecated and work continues on the new Gitlab one?

[Danniello]

Great. So Wayland (with XWayland) with good performance on nVidia card when? 2030?

Jokes aside - no hurry. By the time I will not have nVidia card anyway

[ms178]

I hope it won't take another couple of years until we all can enjoy the fruits of their labour.

[aufkrawall]

I hope it won't take another couple of years until we all can enjoy the fruits of their labour.

Why should "we all" be enabled to "enjoy" anything by their work?
Xwayland Vulkan runs with full performance with GBM as of today.

[geearf]

Michael

The kernel-side bits for Nouveau's implementation can be found via this GitHub tree

The link does not point to GitHub

[ms178]

Why should "we all" be enabled to "enjoy" anything by their work?
Xwayland Vulkan runs with full performance with GBM as of today.

I don't advocate for Nvidia usually, but what is wrong here? Maybe I don't see the full picture, but it might run even better on that new universial allocator as they claim to fix some limitations of GBM (at least on their hardware) and having a well-performing vendor agnostic API might be better for the ecosystem in the long run (easier maintenance).

[aufkrawall]

I don't advocate for Nvidia usually, but what is wrong here? Maybe I don't see the full picture, but it might run even better on that new universial allocator as they claim to fix some limitations of GBM (at least on their hardware) and having a well-performing vendor agnostic API might be better for the ecosystem in the long run (easier maintenance).

GBM is vendor agnostic inside Mesa and the alleged performance limitations are likely hardly more than a foul excuse. We are talking of slow integrated Intel GPUs which apparently can handle it without any issues...