Hi!
Their are two schools of though which I have encountered. One is have
everything in userland. The second is have everything in the kernel. Well
both are wrong. What is really needed? Proper virtualization of the
graphics engine. This means the graphics hardware state is private to each
process and no other process can effect another process graphics hardware
state. This is all that is needed. DRI attemptes to address this. Will
their be more kernel support in the future. Yes if you every want to port
high end graphic servers. Here you end up dealing with with pipes on
different nodes in NUMA systems. Data is passed from node to node to allow
really fast parallel rendering. Note even in this case you don't have
the hardware programmed in the driver but only management of the pipe
state per process.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/