So hard to choose!
Today with ideal kernel, max frec=950... 2 batteries at work (8 am unplugged from car, 2:30 battery change, 7 pm plugged again)
With Lehto 125-950-500, a day without change of battery.
I can assume than user habits are delimiting battery life, anyway, nv kernel beating xlv... how is it possible? If 500 xlv=125 nv, and Lehto 950 ov> ideal 950 ov... ¿?¿?¿?¿
So again reinstaling 125-950-500... I'm getting crazy!!
So hard to choose!
Today with ideal kernel, max frec=950... 2 batteries at work (8 am unplugged from car, 2:30 battery change, 7 pm plugged again)
With Lehto 125-950-500, a day without change of battery.
I can assume than user habits are delimiting battery life, anyway, nv kernel beating xlv... how is it possible? If 500 xlv=125 nv, and Lehto 950 ov> ideal 950 ov... ¿?¿?¿?¿
So again reinstaling 125-950-500... I'm getting crazy!!
LOL!! just toss coins or roll a dice,
i chose Lehto's because that was the first kernel i got from searching about OC.
Be careful with that. I have erased the "current" link and rebooted without reminding to make it again. Now the device doesn't boot any more
I'm not so sure whether you actually need it for booting.
only some poorly designed parts in Maemo5 use it instead of the more reliable `uname -r` approach.
My current plan is to replace the different kernels with a single kernel that can be configured while running. This would mean less maintenance for me, easier installation (from extras) and no need for reflashing.
In addition to min/max frequency settings it should have an option to select a voltage look up table
(normal, LV, ULV, XLV) and the possibility to disable some frequencies in the table
(e.g. 250 Mhz to get a permanent minimum of 500). I'm currently investigating whether this is possible.
Allowing user space to change the vsel column of the mpu table is trivial. It also sounds better to me than having predefined tables in the kernel. The user space program can have those tables. This patch should do the trick:
Disabling specific frequencies might not be as easy. I think that the easiest way is to add a scaling_allowed_frequencies to generic cpufreq code, rather than hacking ARM specific code.