GTA04 PMU / Core arch block diagram v1

Andy Green andy at openmoko.com
Wed Apr 2 21:23:15 CEST 2008


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Somebody in the thread at some point said:

| the MPU-centric solution doesn't help, and you still have to solve
| the general Linux resume problem, or live with the same inconsistent
| behaviour we tried to get rid of in the first place.

Dunno about any "general Linux resume problem" in this area, what I do
know is --->

| Thus, the MPU only needs do things that can't be done sensibly by
| the CPU if we assume that the CPU can respond (= wake up, find out
| what happened, make a decision, act on it - in our scenario, the
| peripherals it needs to talk to quickly are already awake) within,
| say, 100ms.

... today the CPU takes way longer than this on our kernels.  And
tomorrow we inherit a new BSP with drivers we didn't see before with
their own resume behaviours.  I think we shouldn't bet the farm on it
having great resume behaviour but expect it to be cranky -- and for
driver-specific reasons not "general" ones.

Just because MPU actions can't do everything the CPU can doesn't mean
what they can do for us isn't valuable.  According to the current plan,
the actions are programmable from userspace anyway, if resume becomes
lightning fast just turn the MPU actions off and you're sorted.  Then we
get good snappy behaviour either way.

- -Andy
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org

iEYEARECAAYFAkfz3SMACgkQOjLpvpq7dMp7QwCfRf9+ZxqLE5EohEPkQ7a4lTvn
DGMAnj1We/+5gKLQ8xwP7Hf+h26/oW4A
=ScBM
-----END PGP SIGNATURE-----




More information about the openmoko-kernel mailing list