On Mon, 23 Feb 2004, Benjamin Herrenschmidt wrote:

quoted

ment ;-)

I plan to submit the latets mach64 driver from BK to linus as soon as the
first part of the cursor patch goes in. It gives enough of the new
cursor api to do a dump of all the drivers in the 2.5 fbdev tree. Could you
make your patch against the fbdev-2.5 bk tree? Try my patch

http://phoenix.infradead.org/~jsimmons/fbdev.diff.gz

If it doesn't work I will put a new one you there.

Note that I'm having some troubles with James current mach64 bk on a
wallstreet PowerBook (mach64 LT-G). There is an endian bug in the
imageblt implementation, I don't know if the chip can be instructed
to do the endian flip, so I modified the function storing to the
fifo, that appear to work, at least with 8 bits mode ;)

But I'm having lockups, very regular. I haven't been able to figure
out why though. Adding a wait_for_idle() at the end of both fillrect
and copyarea() seem to make them disappear (well, almost, I had _one_
still once). It's weird as those shouldn't be necessary.

Maybe we should add a timeout on the wait_* functions with an engine
reset, that would be sub-optimal, but at least we wouldn't lockup the
box (if that is possible of course).

IIRC, there's a big difference between the number of commands you can queue up
in the RAGE PRO and earlier chip. Perhaps the queue check always assumes a RAGE
PRO?

Hope this helps... (No, I didn't look at the code)

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds


-------------------------------------------------------
SF.Net is sponsored by: Speed Start Your Linux Apps Now.
Build and deploy apps & Web services for Linux with
a free DVD software kit from IBM. Click Now!
http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click