I was considering committing all these patches one by one, but as
discussed with brooks@, there is no need to do this. If we ever
need/want to merge these changes back, it is still possible to do this
per application.
1. Make it more SMP polite. Previous version uses average CPU load that
often leads to load underestimation. It make powerd with default
configuration unusable on systems with more then 2 CPUs. I propose to use
summary load instead of average one. IMO this is the best we can do without
specially tuned scheduler. Also as soon as measuring total load on SMP
systems is more useful then total idle, I have switched to it.
2. Make powerd's operation independent from number and size of frequency
levels. I have added internal frequency counter which translated into real
frequencies only on a last stage and only as good as gone. Some systems may
have only several power levels, while others - many of them, so adaptation
time with previous approach was completely different.
3. As part of previous I have changed adaptive mode to rise frequency on
demand up to 2 times and fall on 1/8 per time internal.
4. For desktop (AC-powered) systems I have added one more mode - "hiadaptive".
It rises frequency twice faster, drops it 4 times slower, prefers twice
lower CPU load and has additional delay before leaving the highest frequency
after the period of maximum load. This mode was specially made to improve
interactivity of the systems where operation capabilities are more
significant then power consumption, but keeping maximum frequency all the
time is not needed.
5. I have reduced default polling interval from 1/2 to 1/4 of second.
It is not so important for algorithm math now, but gives better system
interactivity.
Discussed on: mobile@
the parent's signal mask. Once daemon() forked, signals would be ignored
in the child thread. While I'm here, check the return value of daemon().
This fixes termination in the daemon case (bug introduced in last commit).
Noticed by: Frederik Lindberg
status. Add a thread that waits for events on the named pipe instead of
polling the line status via apm or the ACPI sysctl. Additionally, use
robust error handling in case devd goes away or we temporarily can't set
a frequency (i.e., passive cooling has preempted our setting).
Later, this should be improved so that if adaptive control is not being used,
we will block while waiting for AC line events, saving a little CPU.
Submitted by: Frederik Lindberg <fli+freebsd-current/shapeshifter.se>
with moderate IO going on, system usage can hover around 65-77% even
though it would benefit from higher performance. Typically, only truly
idle systems pass the 90% mark so only demote then.
