On Mon, Apr 25, 2005 at 05:46:12PM +0200, Leonardo Macchia wrote: > On Mon, Apr 25, 2005 at 17:32:36 +0200, Mattia Dongili wrote: > > > ...Anyway :) did you already tried your solution? I'm sorry, I'm pretty > > sceptical because it's battery checking that usually sucks much cpu > > with ACPI because of the kernel doing busy wait, but it's a known issue. > > I have about 100 processes. With that solution (nanosleep of 0,1 s in > any cycle) everything seems to work better... even if sometimes there is
yup! that means that reading 100 processes will take at least 10s ?? > a hit of cpufreqd. > > It's quite hard to track what is consuming more CPU, however in default > configuration it's unusable in my laptop, it's too CPU-comsuming... if you have a modular acpi you could simply remove acpi_ac and/or acpi_battery to test if things are better. > I will try to change default config to check what's really > CPU-consuming... try putting some measurements between suspecious function calls, I'm thinking of gettimeofday around the libsys_get_cpu(), libsys_scan_system_info() and get_running_programs() in main.c around 347->360. Ok, libsys_get_cpu() can't be our problem actually... Thanks for your time :) -- mattia :wq! -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
