Deep Color support

[Pekka Paalanen]

On Mon, 28 Apr 2014 01:42:43 +0200
Wolfgang Draxinger <wdraxinger.maillist at draxit.de> wrote:

> On Sun, 27 Apr 2014 18:25:45 +0300
> Pekka Paalanen <ppaalanen at gmail.com>
> wrote:
> 
> > I guess I just don't understand what "switching X11 servers" means
> > here.
> 
> It means, that if you have multiple X11 servers running on the same
> machine, each configured to use the same GPU, only the OpenGL programs
> using the X11 server/display that currently holds the VT can use the
> GPU.

Oh, that's the DRM set-master related thing. Once we move on to render
nodes, you don't even need a VT to simply work on the GPU, let alone a
display server. IOW, render nodes will just fix that, and you can use
the GPU for computations even from a daemon process if you wish (FOSS
kernel drivers only).

It didn't even occur to me that you might be running several display
servers concurrently.

This is not a X11 or Wayland problem per se, but more like an artifact
of the underlying graphics stack architecture.

> This even stretches to FBOs. Here's a small demonstrator I just put
> together from stuff scattered around my homedirectory (quick and dirty –
> also has a non-growing memory leak, but I don't care).
> http://dl.wolfgang-draxinger.net/test_fbo.cpp
> 
> Start two X11 servers (open a second X11 session or use `startx -- :10`
> `startx -- :11` or similar). On one session run the test_fbo program.
> Then switch to the other X server and repeatedly use
> 
> `display -size 512x512 -depth 8 RGB:/tmp/fbodump.rgb`
> 
> to look at the dumped contents of the FBO. You'll notice that the
> contents of the dump will not change if the X server that test_fbo is
> using is not holding the VT.

Yes, that is by design of the current DRM/DRI, AFAIK, to stop inactive
display server's GPU tasks.

> Then switch to the VT with the X server that test_fbo is using and look
> at the FBO dump again repeatedly. You'll see it changing.


Thanks,
pq