> On Wed, 28 Jan 1998, Brian Gerst wrote:
>
> > XFree86 cannot use the interrupts, because it is a user-space program.
> > Interrupts are only available within the kernel. GGI however can use
> > the interrupts, since the KGI part of it puts low-level video drivers
> > into the kernel.
>
> This is obviously true. If a userspace program wants to make use of
> interrupts it needs kernel assistance. For tasks that are not critical in
> terms of timing, we could imagine a kernel module which then passes a
> signal or an IPC message to the userspace program.
FWIW the graphics IRQ is an excellent example of timing-sensitive... The
time it takes to run the IRQ through to userspace is longer than the time
the IRQ is valid.... (at least for vertical refresh timing)
> Does GGI really use interrupts generated by graphics devices and take any
> profit out of this?
On some cards yes... (afaik)
On cards that SHOULD be supported (eg. S3-ViRGE) not yet. GGI is quite
immature though - but IRQ's are a necessity for efficient screen
management.
Otherwise you either get a slower computer (X's style) or display covered
in sparkles with standard VGA's IRQ.
Accelerators provide other info on IRQ as well (finished a series of accel
commands, DMA sequence finished, ...)
On all cards there's other _POLLING_ methods to find out everything
so IRQ's aren't 100% necessary - but the system will run a lot cleaner
(and faster) with them enabled in graphics modes (when the code supports
it).
G'day, eh? :)
- Teunis [I use GGI and am trying to write drivers :]
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu