Jacopo Mondi <jacopo@xxxxxxxxxx> writes: >> They can. It appears, though, that clusters aren't about controls which >> can't change independently. Both of the two are written to the hardware >> at the same time, which appears to be a valid reason to combine them >> into a cluster. > > Is there a reason to write them to hw at the same time ? :) Sure, they are written in a single merged burst on I2C. > As it doesn't seem to me you access registers during probe (to > identify the sensor in example) there's no need to power up the device > during probe, but should be enough to let runtime_pm do so when > requested. Yes, in fact I never try to read anything from the sensor :-) Unfortunately I'm not sure how to initialize the (possibly nonexistent) PM without powering it up, cleanly, in all kernel configs. I mean, I can check if PM failed and power it up "by hand", but it's a bit messy. Suggestions are welcome, though. -- Krzysztof "Chris" Hałasa Sieć Badawcza Łukasiewicz Przemysłowy Instytut Automatyki i Pomiarów PIAP Al. Jerozolimskie 202, 02-486 Warszawa
