Commands tagged cpu

Similar to `cpulimit`, although `prlimit` can be found shipped with recent util-linux.

Example: limit CPU consumption to 10% for a math problem which ordinarily takes up 100% CPU:

Before:

See the difference `prlimit` makes:

To actually monitor the CPU usage, use `top`, `sar`, etc.. or:

Using the output of 'ps' to determine CPU usage is misleading, as the CPU column in 'ps' shows CPU usage per process over the entire lifetime of the process. In order to get *current* CPU usage (without scraping a top screen) you need to pull some numbers from /proc/stat. Here, we take two readings, once second apart, determine how much IDLE time was spent across all CPUs, divide by the number of CPUs, and then subtract from 100 to get non-idle time.

This will limit the average amount of CPU it consumes.

Watch the temperatures of your CPU cores in real time at the command line. Press CONTROL+C to end.

GORY DETAILS: Your computer needs to support sensors (many laptops, for example, do not). You'll need to install the lm-sensors package if it isn't already installed. And it helps to run the `sensors-detect` command to set up your sensor kernel modules first. At the very end of the sensors-detect interactive shell prompt, answer YES to add the new lines to the list of kernel modules loaded at boot.

For each cpu set mask and then monitor your cpu infos. Temp,load avg. etc.

For example for 2nd cpu or 2nd core

taskset 0x00000002 yes > /dev/null &

For example for 3rd cpu or 3rd core

taskset 0x00000004 yes > /dev/null &

For example for 4th cpu or 4th core

taskset 0x00000008 yes > /dev/null &

Monitor your cpu temp with this command if you want

watch -n1 "acpi -t"

Load avg. from top command

top

kerim@bayner.com

http://www.bayner.com/

No need for a colon, and one less semicolon too. Also untested.

There is no need for variables. I also added sleep to reduce cpu usage, however I didn't test it.

Once you know the available frequencies for your CPU, they can be used to do things like set minimum CPU frequency for powerd so that it doesn't ramp down too slow on a server :

/etc/sysctl.conf or /boot/loader.conf:

debug.cpufreq.lowest=DESIRED FREQ HERE

or at terminal

sysctl debug.cpufreq.lowest=DESIRED FREQ HERE

If shell escaping of the command is problematic, you can write the command to a file first:

Or read it:

Or, if your shell supports it, you can eliminate echo:

("man batch" lists 1.5 for me, but I don't know how widely it differs.)

[ 2000 -ge "$(free -m | awk '/buffers.cache:/ {print $4}')" ] returns true if less than 2000 MB of RAM are available, so adjust this number to your needs.

[ $(echo "$(uptime | awk '{print $10}' | sed -e 's/,$//' -e 's/,/./') >= $(grep -c ^processor /proc/cpuinfo)" | bc) -eq 1 ] returns true if the current machine load is at least equal to the number of CPUs.

If either of the tests returns true we wait 10 seconds and check again. If both tests return false, i.e. 2GB are available and machine load falls below number of CPUs, we start our command and save it's output in a text file.

The ( ( ... ) & ) construct lets the command run in background even if we log out. See http://www.commandlinefu.com/commands/view/3115/ .

I've wanted this for a long time, finally just sat down and came up with it. This shows you the sorted output of ps in a pretty format perfect for cron or startup scripts. You can sort by changing the k -vsz to k -pmem for example to sort by memory instead.

If you want a function, here's one from my http://www.askapache.com/linux-unix/bash_profile-functions-advanced-shell.html

This is useful if you have a program which doesn't work well with multicore CPUs. With taskset you can set its CPU affinity to run on only one core.