It shows OoO execution and/or write buffers and/or cache weak consistency

turns.c

#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/types.h>

#define NUM_THREADS  4
#define MAX_COUNT 10000000

struct tdata {
    int tid;
    int *counter;
};

int turns[NUM_THREADS];

void *count(void *ptr) {
    long i, max = MAX_COUNT/NUM_THREADS;
    int tid = ((struct tdata *) ptr)->tid;
    int *counter = ((struct tdata *) ptr)->counter;
    int number = 0, next = (tid+1)%NUM_THREADS;

    for (i=0; i < max; i++) {
        while(turns[tid] != number);
        number++;
        *counter+= 1; // The global variable, i.e. the critical section
        turns[next]++;
    }

    printf("End %d counter: %d\n", tid, *counter);
    pthread_exit(NULL);
}

int main (int argc, char *argv[]) {
    pthread_t threads[NUM_THREADS];
    int rc, i;
    struct tdata id[NUM_THREADS];
    int counter = 0;

    turns[0] = 0;
    for(i=1; i<NUM_THREADS; i++){
        turns[i] = -1;
    }

    for(i=0; i<NUM_THREADS; i++){
        id[i].counter = &counter;
        id[i].tid = i;
        rc = pthread_create(&threads[i], NULL, count, (void *) &id[i]);
        if (rc){
            printf("ERROR; return code from pthread_create() is %d\n", rc);
            exit(-1);
        }
    }

    for(i=0; i<NUM_THREADS; i++){
        pthread_join(threads[i], NULL);
    }

    printf("Counter value: %d Expected: %d\n", counter, MAX_COUNT);
    return 0;
}