NVIDIA Developer Talks Openly About Linux Support

Q: What package (X.Org, kernel, etc) has the most room for improvement to better complement video drivers such as NVIDIA's for 3D acceleration, desktop effects, video playback, etc and in what way?

The NVIDIA Linux graphics driver is very complex and stresses nearly all parts of the overall system it interacts with to a much greater degree than most other drivers. Much like the hardware it drives, it is highly optimized and relies on many features in core system components (such as the Linux kernel) to work flawlessly. Inadequate exercise of some of these paths within the Linux kernel, as well as a disregard for out-of-tree drivers, periodically results in subtle regressions in the Linux kernel and its interaction with the NVIDIA kernel module. Reliability has also suffered as changes to the mainline kernel's development model have placed more of the refinement burden on distributors.

However, in terms of the Linux kernel, we seldom find ourselves at odds with or challenged by fundamental architectural shortcomings. The issues are generally just bugs.

In terms of future growth: I believe the largest opportunities are in the X Window System. This isn't because X is bad (many of the core architectural decisions are quite sound: e.g., separation of policy into window managers and composite managers), but the further evolution of the Linux composited desktop will necessarily occur in the primary desktop components: the X server, the X drivers, the window managers and composite managers.

Thankfully, in my view, the X Window System as implemented using the XFree86/X.Org Loadable Driver Framework, has the flexibility and extensibility to be used to solve the modern Linux desktop challenges.

Thanks again to Andy Ritger for taking the time to answer all of these questions. We also have a ATI Linux Q&A thread on the Phoronix Forums that has accumulated more than 910 responses with some answers, for those interested in more Linux graphics reading material. If you appreciate this kind of content, we ask that you consider subscribing to Phoronix Premium (single-page article viewing, ad-free browsing, etc), making a donation, shop with our Amazon.com affiliate link, and visit our other web properties.

References:

[1] http://www.nvidia.com/object/unix.html
[2] http://www.nvidia.com/object/linux_nforce_1.23.html
[3] http://www.nvidia.com/object/sli_multi_os.html
[4] http://current.com/1gp964c
[5] http://ttimo.vox.com/library/post/id-software-and-linux.html
[6] http://www.opengl.org/registry/specs/ARB/provoking_vertex.txt
[7] http://www.opengl.org/registry/specs/ARB/fragment_coord_conventions.txt
[8] http://www.opengl.org/registry/specs/ARB/vertex_array_bgra.txt
[9] http://www.opengl.org/registry/specs/EXT/direct_state_access.txt
[10] http://www.slideshare.net/Mark_Kilgard/opengl-32-and-more
[11] http://www.alientrap.org/nexuiz/
[12] http://www.xpilot.org/
[13] http://lists.x.org/archives/xorg-devel/2009-September/002194.html
[14] http://www.x.org/wiki/Events/XDC2009/Notes#head-78b998732e57409a5c09dde5b0ac4c39d124cd7b
[15] http://lists.freebsd.org/pipermail/freebsd-hackers/2006-June/016995.html
[16] http://people.redhat.com/drepper/dsohowto.pdf

Discuss this article in our forums, IRC channel, or email the author. You can also follow our content via RSS and on social networks like Facebook, Identi.ca, and Twitter (@Phoronix and @MichaelLarabel). Subscribe to Phoronix Premium to view our content without advertisements, view entire articles on a single page, and experience other benefits.