Chuck Swiger wrote:
On Jan 27, 2009, at 4:49 PM, Shawn Badger wrote:
Have you tried reducing HZ to 100 (put kern.hz="100" in /boot/loader.conf and reboot)? Are you running powerd? Look into "sysctl hw.acpi" and "sysctl debug.cpufreq"....

Thanks for the ideas Chuck. I lowered kern.hz to 100 as you suggested (does this affect the kernel's ability to track time in milliseconds? ie. if I want to run a benchmark using the 'time' utility?).

Changing the scheduler quantum won't affect the system clock or the ability to do millisecond-level timing of userland processes. It does affect the granularity of things like ipfw/dummynet if polling is enabled, but shouldn't have any real negative effects otherwise.

For most of Unix history, HZ=100 was a common default, and the reduced context switch frequency should result in a decent improvement to power drain. If you have a concern, consider comparing against HZ=250 and see how the battery life and responsiveness or granularity of network traffic, etc feel....

Thanks for the info. I'll definitely do some tests and find a good balance.
And the output of the two sysctl queries is posted here:

I'm not very familiar with acpi, so if you see anything that could be optimized, I'd appreciate the feedback.

I have limited experience with running FreeBSD on a laptop personally [1], so others will likely have more relevant feedback; I'm just aware of some starting points. :-)

Regards, -- -Chuck

[1]: I've helped a few people run FreeBSD 5.x/6.x on various IBM ThinkPads (circa T.42s) an maybe an HP Pavillion or Dell Latitude, and I've run FreeBSD a bit on a Mac mini and a MacBookPro (2,2), but I don't use FreeBSD on a laptop regularly...I think of it as a server OS. :-)

I too generally think of FreeBSD as a server OS, but I just can't get over how nice the development environment is - hence the laptop.

_______________________________________________ mailing list
To unsubscribe, send any mail to ""

Reply via email to