问题
I am trying to:
Run 16 copies concurrently with processor pinning (2 copies per core)
Run 8 copies concurrently with processor pinning (2 copies per core) and flipping processor core to the furthest core after certain function say function 1 finishes.
The problem I am facing is how to select the farthest processor.
Some friends suggested to use sched_getaffinity and sched_setaffinity but I count not find any good examples.
回答1:
To use sched_setaffinity to make the current process run on core 7 you do this:
cpu_set_t my_set; /* Define your cpu_set bit mask. */
CPU_ZERO(&my_set); /* Initialize it all to 0, i.e. no CPUs selected. */
CPU_SET(7, &my_set); /* set the bit that represents core 7. */
sched_setaffinity(0, sizeof(cpu_set_t), &my_set); /* Set affinity of tihs process to */
/* the defined mask, i.e. only 7. */
See http://linux.die.net/man/2/sched_setaffinity & http://www.gnu.org/software/libc/manual/html_node/CPU-Affinity.html for more info.
回答2:
Don't use CPU_SETSIZE as cpusetsize parameter for sched_[set|get]affinity. The names are misleading but this is wrong. The makro CPU_SETSIZE is (quoting man 3 cpu_set) "a value one greater than the maximum CPU number that can be stored in cpu_set_t." You have to use
sched_setaffinity(0, sizeof(cpu_set_t), &my_set);
instead.
回答3:
Minimal runnable example
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?
nproc
respects sched_getaffinity
by default as shown at: How to find out the number of CPUs using python
Tested in Ubuntu 16.04.
来源:https://stackoverflow.com/questions/10490756/how-to-use-sched-getaffinity-and-sched-setaffinity-in-linux-from-c