On Sat, 30 Aug 2003, Thiemo Seufer wrote:
quoted
quoted
quoted
I've looked at the patch and have changed the barriers as I deem
appropriate. But while doing this, I've noticed the code is insonsistent
in a few places -- has it been successfully verified on real hardware?
There are some mb() calls in your patch, I believe a rmb()/wmb() is
enough in those cases.
They are there because you don't really know if the previous access (if
any) to the addressed chip was a read or a write. So you need to use a
barrier that works for both. Please let me know if I got any specific
ones wrong, though.
quoted
quoted
It works on my /260. Which parts look weird to you?
1. bt431_write_reg_inc() uses u16 for value, even though it only uses 8
bits (which of course makes all the calls in bt431_init_cursor() dubious).
2. bt431_load_cursor_sprite() loads control registers instead of the
cursor map.
The hardware cursor is still unused. The code there is very experimental,
and your version is surely an imporvement. :-)
Hmm, but doesn't it lead to a garble where cursor is expected to be?
Maciej
--
+ Maciej W. Rozycki, Technical University of Gdansk, Poland +
+--------------------------------------------------------------+
+ e-mail: macro@ds2.pg.gda.pl, PGP key available +
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf