Primitive benchmark to measure process startup time

startup.c

/* a.c, compiled to ./a */
void main() {}

/* b.c, compiled to ./b */
#include <sys/wait.h>
#include <unistd.h>

int main() {
    for (size_t i = 0; i < 100 * 1000; i++) {
        if (fork() == 0) {
            execve("./a", (char*[]){ NULL }, (char*[]){ NULL });
        } else {
            int wstatus;
            wait(&wstatus);
        }
    }
}

/* Then on a Core i7:

   $ gcc a.c -O0 -o ./a
   $ gcc b.c -O3 -o ./b
   $ time ./b
   real	1m3.293s
   user	0m26.689s
   sys	0m34.515s
   
   This implies 63.293 s * 1000 / 100000 =~ 633 μs of wall clock time per execution of ./a.
 */