On Thu, Jun 12 2003, Soeren Sonnenburg wrote: > On Thu, 2003-06-12 at 09:09, Jens Axboe wrote: > > On Thu, Jun 12 2003, Soeren Sonnenburg wrote: > > > On Thu, 2003-06-12 at 07:52, Jens Axboe wrote: > > > > On Wed, Jun 11 2003, Soeren Sonnenburg wrote: > > > > > [...] > > > > > Any reasonable values ? > > > > > > > > > > For the moment I would allow for hdsleeping for 1hr ... so I am doing > > > > > hdparm -S60 /dev/hda (-S0 respectively) also in the laptop-mode.sh > > > > > script as it makes no sense for me to have the hd beeing powered down > > > > > elsewhen... > > > > > > > > I use -S6 and -S0 on the kernels that don't have the AAM patch. But I > > > > almost always use the laptop-mode patch in conjunction with the AAM > > > > patch, so I don't have a lot of mileage on it alone. > > > > > > Sorry, but what is AAM ? > > > > Automatic acousting management, essentially meant to control the > > acoustic level (and thus power leve) of the drive while being used. It > > will keep the drive in three states: spun down, quiet (low power > > consumption), and fast. > > >From that it is not really clear what that patch does... I mean will > that patch try to keep the drive in the lowest powermode possible all > the time ?
Yes, that's about it. > Hmmhh exactly that would be needed for cpufreq... atm it is either fast > or slow but not automagically detected on cpu load... cpudyn has something like that, but it doesn't work very well. It doesn't take io wait time into account either. But I fully agree with you, cpu throttling should be based on the same criteria. > > It was posted on lkml some time back, I can > > repost an updated version if there's any interest. > > At least I am interested... > > Would be nice if it applies cleanly to the rsync version of > benh-devel... It's based off benh-devel, so yes. I'll post a patch later today, need to be near the powerbook first. -- Jens Axboe
