How to set CPU affinity for a process from C or C++ in Linux?

36,131

Solution 1

You need to use sched_setaffinity(2).

For example, to run on CPUs 0 and 2 only:

#define _GNU_SOURCE
#include <sched.h>

cpu_set_t  mask;
CPU_ZERO(&mask);
CPU_SET(0, &mask);
CPU_SET(2, &mask);
int result = sched_setaffinity(0, sizeof(mask), &mask);

(0 for the first parameter means the current process, supply a PID if it's some other process you want to control).

See also sched_getcpu(3).

Solution 2

Use sched_setaffinity at the process level, or pthread_attr_setaffinity_np for individual threads.

Solution 3

I have done many effort to realize what is happening so I add this answer for helping people like me(I use gcc compiler in linux mint)

#include <sched.h> 
cpu_set_t  mask;

inline void assignToThisCore(int core_id)
{
    CPU_ZERO(&mask);
    CPU_SET(core_id, &mask);
    sched_setaffinity(0, sizeof(mask), &mask);
}
int main(){
    //cal this:
    assignToThisCore(2);//assign to core 0,1,2,...

    return 0;
}

But don't forget to add this options to the compiler command : -D _GNU_SOURCE Because operating system might assign a process to the particular core, you can add this GRUB_CMDLINE_LINUX_DEFAULT="quiet splash isolcpus=2,3" to the grub file located in /etc/default and the run sudo update-grub in terminal to reserve the cores you want

UPDATE: If you want to assign more cores you can follow this piece of code:

inline void assignToThisCores(int core_id1, int core_id2)
{
    CPU_ZERO(&mask1);
    CPU_SET(core_id1, &mask1);
    CPU_SET(core_id2, &mask1);
    sched_setaffinity(0, sizeof(mask1), &mask1);
    //__asm__ __volatile__ ( "vzeroupper" : : : ); // It is hear because of that bug which dirtied the AVX registers, so, if you rely on AVX uncomment it.
}

Solution 4

sched_setaffinity + sched_getaffinity minimal C runnable example

This example was extracted from my answer at: How to use sched_getaffinity and sched_setaffinity in Linux from C? I believe the questions are not duplicates since that one is a subset of this one, as it asks about sched_getaffinity only, and does not mention C++.

In this example, we get the affinity, modify it, and check if it has taken effect with sched_getcpu().

main.c

#define _GNU_SOURCE
#include <assert.h>
#include <sched.h>
#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>

void print_affinity() {
    cpu_set_t mask;
    long nproc, i;

    if (sched_getaffinity(0, sizeof(cpu_set_t), &mask) == -1) {
        perror("sched_getaffinity");
        assert(false);
    }
    nproc = sysconf(_SC_NPROCESSORS_ONLN);
    printf("sched_getaffinity = ");
    for (i = 0; i < nproc; i++) {
        printf("%d ", CPU_ISSET(i, &mask));
    }
    printf("\n");
}

int main(void) {
    cpu_set_t mask;

    print_affinity();
    printf("sched_getcpu = %d\n", sched_getcpu());
    CPU_ZERO(&mask);
    CPU_SET(0, &mask);
    if (sched_setaffinity(0, sizeof(cpu_set_t), &mask) == -1) {
        perror("sched_setaffinity");
        assert(false);
    }
    print_affinity();
    /* TODO is it guaranteed to have taken effect already? Always worked on my tests. */
    printf("sched_getcpu = %d\n", sched_getcpu());
    return EXIT_SUCCESS;
}

GitHub upstream.

Compile and run:

gcc -ggdb3 -O0 -std=c99 -Wall -Wextra -pedantic -o main.out main.c
./main.out

Sample output:

sched_getaffinity = 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 
sched_getcpu = 9
sched_getaffinity = 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
sched_getcpu = 0

Which means that:

  • initially, all of my 16 cores were enabled, and the process was randomly running on core 9 (the 10th one)
  • after we set the affinity to only the first core, the process was moved necessarily to core 0 (the first one)

It is also fun to run this program through taskset:

taskset -c 1,3 ./a.out

Which gives output of form:

sched_getaffinity = 0 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 
sched_getcpu = 2
sched_getaffinity = 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 
sched_getcpu = 0

and so we see that it limited the affinity from the start.

This works because the affinity is inherited by child processes, which taskset is forking: How to prevent inheriting CPU affinity by child forked process?

Python: os.sched_getaffinity and os.sched_setaffinity

See: How to find out the number of CPUs using python

Tested in Ubuntu 16.04.

Solution 5

In short

unsigned long mask = 7; /* processors 0, 1, and 2 */
unsigned int len = sizeof(mask);
if (sched_setaffinity(0, len, &mask) < 0) {
    perror("sched_setaffinity");
}

Look in CPU Affinity for more details

Share:
36,131
jcodeninja
Author by

jcodeninja

C, C++, perl, java Developer who prefers to develop for the linux platform.

Updated on July 12, 2022

Comments

  • jcodeninja
    jcodeninja almost 2 years

    Is there a programmatic method to set CPU affinity for a process in c/c++ for the Linux operating system?

  • ScrollerBlaster
    ScrollerBlaster over 12 years
    This might compliment the technique: to query the CPU the thread is currently running on use this: <stackoverflow.com/questions/491520/…>
  • Manuel Selva
    Manuel Selva over 10 years
    don't forget to define GNU_SOURCE when compiling
  • Alnitak
    Alnitak about 7 years
    you should use the correct type and macros for this (e.g. cpu_set_t and CPU_SET) - there's no guarantee that the bits in a long will be in the right place.
  • PlsWork
    PlsWork over 4 years
    Error: Use of undeclared identifier 'result'
  • Hossein
    Hossein almost 4 years
    How do you specify more cpu cores? instead of just one core?
  • Hossein
    Hossein almost 4 years
    @Thanks a lot, so basically a for loop would do the trick. really appreciate your help
  • Amiri
    Amiri almost 4 years
    Yeah, but note that process can not use other cores. It's useful for benchmarking.