Question

I’m trying to write a script which could output the id of the core that executes the script.

I have already found a system call getCPU which can achieve my goal.
However, I don’t know how to call this system call and get it work in a bash script.

Any help would be grateful!

Dennis

Answer 1

Unless you have taken action to pin a process to a particular core, the scheduler will move it around as it sees fit. So each time the process is scheduled it is likely to change.

On linux with a kernel >= 2.2.8 you can find the core that the process was last run on by looking at the /proc/<PID>/stat file. Eee the PROC(5) man page for more information. A simple awk script can be used to get the relevant information e.g. if your shell was PID 2338 then

awk '{print $39}' /proc/2338/stat

will print the CPU number that your shell last ran on. If you run the awk script several times on a multi-core/cpu system you should see the that the process moves from core to core.

If you want to set the affinity your process to a particular core (or group of cores) then you can use the taskset command.

taskset -p 1 2338

will set the affinity of the process to core 0, -p 3 2338 will set the affinity to cores 0 and 1 etc.