[PATCH 0/1]: Acceleration threshold configuration

Andy Green andy at openmoko.com
Mon Oct 13 13:54:31 CEST 2008

Hash: SHA1

Simon Kagstrom wrote:

> * I've based this on the bitbang-all-the-way-baby from the Andy branch.
>   My experiments with spi_async() have not been very successful - I'm
>   seeing missed interrupts just like with the "plain" workqueue
>   implementation, but in this case it's actually harder to loop until
>   the status is OK - the completion callback is called from a
>   workqueue, so submitting other stuff (with reg_read()) isn't
>   guaranteed to actually execute.

With this and since I know there is great interest about threshold
stuff, I stuck bitbang-all-the-way and this threshold patch into stable.
 I just built it and confirmed the accels were not broken, I did not
check threshold performance.

> * This patch will break (the anyway non-working) acceleration wakeup
>   implementation. However, I'm thinking that we should just change that
>   to be set when the device is suspended anyway.

Sounds like we should do that.

> * I'd like some more testing of this patch!

Well it should get some now... if it's still shaky I think we need to
look at Sean McNeil's patch for 2.6.26 because with that he got reliable
performance.  But I rather think with bitbang-all-the-way now only
having one bunch of code touching the ports, and interrupt lockout
protecting them, the only thing left is going to be edge interrupt vs
level if there still is trouble.

> * I'm slightly worried about the working hours for Andy, Werner and the
>   others, don't you have 40h weeks at Openmoko? :-)

We're so busy we didn't find time for the heart attack yet.

- -Andy
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org


More information about the openmoko-kernel mailing list