On Jun 15 11:05:39, Ted Unangst wrote: > On Mon, Jun 15, 2009 at 10:29 AM, Jan Stary<h...@stare.cz> wrote: > > What is the best way to learn about the power/frequency/thermal > > control options of my CPU from bsd's point of view (besides > > dmesg and sysctl)? For example, what are the P-states and C-states > > my CPU can enter, and which of those does bsd support? > > you can adjust hw.setperf from 0 to 100. > > given the current level of acpi support, the only state your cpu can > enter is "on". > > > What exactly is the relation of apm, acpi, and hw.setperf? > > apm is what laptops used to go to sleep 10 years ago. acpi is what > laptops today use to annoy kernel developers. hw.setperf is a uniform > userland interface to what may be one of many backend drivers. > > > So, neither apm nor acpi (acpicpu) is needed to use hw.setperf? > > no. > > > Also, the Enhanced SpeedStep support on my CPU reduces to > > > > cpu0: unknown Enhanced SpeedStep CPU, msr 0x061a082006000820 > > cpu0: using only highest and lowest power states > > cpu0: Enhanced SpeedStep 2667 MHz (1212 mV): speeds: 2667, 2000 MHz > > > > - is there something I can do about it? Is there a point in running > > current (as opposed to 4.5-stable) with regard to this? > > current has different acpi code. maybe that works. what difference > acpicpu makes over est.c is probably none. > > on most machines now, setperf works by poking registers in the cpu > telling it to speed up or slow down. est knows about two settings, > fast and slow. acpi may have information about some other settings in > the middle which are unlikely to be of use unless your cpu is > frequently exactly 40% busy.
I just upgraded to -current. That makes it boot GENERIC.MP with ACPI, good. With regard to CPU freq control (now that acpi is in charge of that, and not est), the difference is indeed none. Setting hw.setperf to whatever only makes a difference when crossing setperf=50, which lowers 2667 to 2000 as before.
