How to change order of creation of processes? - c

Question

I have to make a tree of processes like in this image:
I have managed to create such a tree, but with a minor difference. P4 is actually P3, and P3 is P4. I have created them in order, because i can't think of a way of "attaching" P4 to P1, and P3 to P2. This is what I've written so far:
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/types.h>
#include <sys/wait.h>
#include "a2_helper.h"
int main(int argc, char **argv){
printf("Main pid is %d\n", getpid());
pid_t pid1[2];
pid_t pid2[4];
pid_t pid3a, pid3b;
//FIRST LAYER
for(int i = 0; i < 2; i++){
pid1[i] = fork();
if(pid1[i] == 0){
printf("[son] pid %d from [parent] pid %d\n", getpid(), getppid());
//SECOND LAYER
if ( i == 0){
for(int j = 0; j < 4; j++){
pid2[j] = fork();
if (pid2[j] == 0){
printf("\t[son] pid %d from [parent] pid %d\n", getpid(), getppid());
//THIRD LAYER
if (j == 1){
pid3a = fork();
if (pid3a == 0){
printf("\t\t[son] pid %d from [parent] pid %d\n", getpid(), getppid());
exit(0);
}
waitpid(pid3a, NULL, 0);
}
if (j == 2){
pid3b = fork();
if (pid3b == 0){
printf("\t\t[son] pid %d from [parent] pid %d\n", getpid(), getppid());
exit(0);
}
waitpid(pid3b, NULL, 0);
}
//END THIRD LAYER
exit(0);
}
}
//END SECOND LAYER
for(int j = 0; j < 4; j++)
waitpid(pid2[j], NULL, 0);
}
exit(0);
}
}
for(int i = 0; i < 2; i++)
waitpid(pid1[i], NULL, 0);
//END FIRST LAYER
return 0;
}
What can I change in order to achieve that?

Answer 1

It's better for fork one fewer times than the paths you need. Look at the following code based on the graph (without P3 and P4 switched)
void doP1(){
if(fork())
doP2();
else
doP4();
}
void doP2(){
if(fork())
doP3();
else if(fork())
doP5();
else if(fork())
doP6();
else
doP7();
}
void doP3(){
return;
}
void doP4(){
return;
}
void doP5(){
doP8();
}
void doP6(){
doP9();
}
void doP7(){
return;
}
void doP8(){
return;
}
void doP9(){
return;
}
int main(){
doP1();
}
I separated every node of the tree into its own function which do nothing but fork in the way the graph shows.
When a data-path splits into two, you don't need to create two child threads, you just need to create one. The parent goes one way, and the child goes another.
i.e. P1 should only call fork once, P2 should only call fork three times, P5 and P6 shouldn't call fork at all
I hope that answers your question!