[PATCH 3/4] modesetting: Implement a double-buffered shadow mode

[Adam Jackson]

On Mon, 2015-07-27 at 13:35 -0700, Eric Anholt wrote:

> I'm not opposed to this, but could the kernel driver let us know that
> the GPU has this hardware VNC feature attached somehow?

We could certainly add a drm feature bit. At the moment, I don't know
of a way to query the vbios or anything else about the platform for the
presence of the feature, so I'm not sure how much good it would do.

On the other hand GPUs like this tend to be PCIE x1 at best, and
250MB/s is a pretty poor framerate. Your RAM and CPU are fast enough
that doing the extra memcmp step would likely decrease wall time and/or
pipe latency just from not saturating the bus, even before we consider
whether we're swamping the vnc controller.

- ajax