Kernel support for graphics cards

[olafBuddenhagen at gmx.net]

Hi,

> Well your not going to see hardware acceleration and things in the
> kernel, but they're are a number of us that believe we need to
> rearchitect things a lot [...]

What exactly do you mean by "hardware acceleration" here?

Actually feeding the accelerator pipelines etc. is hardware access, and
needs to be multiplexed in the kernel. Generating the accelerator
commands necessary to perform the desired graphics operations on the
other hand is pure logic, and can be safely done in a userspace
component.

Luckily, all the complexity of graphics drivers lies in the latter; thus
it's possible to build a robust kernel-supported graphics architecture
without bloating the kernel.

> The work needed is fairly major, but just needs some focused effort
> most likely from a paid developer to actually get it done.. (it's not
> the most glamourous work and it's most definitely not suitable for a
> newbie programmer...)

I disagree.

KGI/GGI has had such an architecture working in principle for years now.
I guess we agree there are serious organizational and probably also
technical issues with KGI as it is; yet it shows prettly clearly that
there is enough potential for doing this work in a volunteer project.
All we need is a more successful effort to channel the existing
potential in the right direction...

-antrik-