Hey, while setting the ondemand governor on Intel/x86 machines is generally a good idea, this is not true for PowerPC (and prolly some other !Intel arches too). When you try to set ondemand (or conservative), the kernel will barf into the logs and switch to performance, which sucks generally. So I see different possibilities here, not sorted, though: 1. let me disable cpufreq settings in xfpm, and let something like cpufreqd or powernowd handle it, but why running two apps, when one is enough? 2. add an very-advanced tab to the config, where one can explicitelly say which governor should be used on ac and bat (like this was in the 0.6.x versions) 3. use userspace governor and some really intelligent code that will detect the load of the machine and give it more power when needed (that can become very complicated) For now, 2 would be ok for me, as powersave on AC is fine here, YMMV as usual :) PS: what would actually happen, if there would be no ondemand and conservative governors in the kernel?
(In reply to comment #0) > Hey, > > while setting the ondemand governor on Intel/x86 machines is generally a good > idea, this is not true for PowerPC (and prolly some other !Intel arches too). > When you try to set ondemand (or conservative), the kernel will barf into the > logs and switch to performance, which sucks generally. > > So I see different possibilities here, not sorted, though: > > 1. let me disable cpufreq settings in xfpm, and let something like cpufreqd or > powernowd handle it, but why running two apps, when one is enough? > > 2. add an very-advanced tab to the config, where one can explicitelly say which > governor should be used on ac and bat (like this was in the 0.6.x versions) > > 3. use userspace governor and some really intelligent code that will detect the > load of the machine and give it more power when needed (that can become very > complicated) > > For now, 2 would be ok for me, as powersave on AC is fine here, YMMV as usual > :) > > PS: what would actually happen, if there would be no ondemand and conservative > governors in the kernel? if there is no ondemand and consevative governors found then simply xfpm will not be playing the the cpu freq. I will go for the first possibility, since in future versions when we move to device-kit power all those stuff will change and latency will take place. more on this http://hal.freedesktop.org/docs/DeviceKit-power/QoS.html I think it should be okay for you to go for 1 right?
even if I'd prefer 2, 1 is okay too
(In reply to comment #2) > even if I'd prefer 2, 1 is okay too You have in the last revision r7248, please check it out and confirm. Thanks.
Closing old bugs to keep things organized.