On Mon, 16 Apr 2001, Nicolas Souchu wrote:
> You're kidding... of course not! :)
>
> I don't know. Why do you need this?
A couple of reasons. One is to provide to emulate a linear MMAP
on older graphics hardware with paged and plane-masked display buffers.
That requires some extra special treatment in the page fault handlers.
The other is something that (I think it is James Simmons who has proposed
this) uses the scheduler as a way of virtualizing the graphics hardware
such that different applications of different security clearances can
access the same hardware without exceeding their access privilages.
That is necessary because if you use the page fault handler to do this and
do not control the application's timeslice, you get situations where
interleaved access by the apps does a lot of inefficient reprogramming
of the page tables. (It also probably solves a lot of problems with
soft real time apps.)
--
Brian
