On Wed, Jan 19, 2005 at 01:03:48AM +0900, Izumi Tsutsui wrote:
> But I don't think it can be handled by dumb wsfb driver.
> Some wscons drivers already have {copy,erase}{cols,rows}()
> functions for wscons/rasopos with hardware support, so
> I just guess they could be easily used with the same API by dumb wsfb.

I'm not sure.  My experience with X is that you have to do a lot
of work to get regions that need a simple copy/fill operation, and
often the regions are small enough that the overhead of calling
functions in the hardware brings marginal gain.  It might make
more sense to expose the hardware mechanism for these directly
to userland and duplicate code from the kernel.  This would not
require a new ioctl(), and would more closely match what is done
for, say, XFree86.

The new WSDISPLAYIO_GINFO should have a way to uniquely identify
the hardware type to userland.

> > If there is some acceleration option, then perhaps there should be
> > a future WSDISPLAYIO_GACCELINFO or similar.
> Ok, I agree to have another ioctl for such functions.

Thank you.  I think this makes the most sense.  That ioctl() might
be a small set of ioctl()s.  Perhaps exposing hardware information
for userland use or perhaps accessing hardware functions.

-allen

-- 
 Allen Briggs                     briggs@wasabisystems.com
 Wasabi Systems, Inc.             http://www.wasabisystems.com/