Shared Programming Using Library Routines
The most well-liked of them is the use of combo function called fork() and join(). Fork() function is used to make a new child process. By calling join() function parent process waits the terminations of the child process to obtain the desired result.
Example 11: Consider the following set of statements
Process A Process B
: :
fork B ; :
: :
join B; end B;
In the above set of statements process A makes a child process B by the statement fork B. Then A and B continue their computations independently up to A reaches the join statement, At this stage, if B is already ended, then A continues implementing the next statement otherwise it waits for B to finish.
In the shared memory model, a common trouble is to synchronize the processes. It may be possible that more than one process is trying to concurrently modify the similar variable. To solve this problem many synchronization mechanism like test_and_set, monitors and semaphores have been used. We shall not go into the details of these mechanisms. Quite, we shall represent them by a pair of two processes called lock and unlock. Whenever a process P locks a common variable, then only P can use that variable. Other simultaneous processes have to wait for the common variable until P calls the unlock on that variable. Let us see the effect of locking on the output of a program when we do not use lock and when we use lock.
Example 12
Let us write a pseudocode to find sum of the two functions f(A) + f(B). In the first algorithm we shall not use locking.
Process A Process B
sum = 0 :
: :
fork B sum = sum+ f(B)
: :
sum = sum + f(A) end B
:
join B
:
end A
If process A implements the statement sum = sum + f (A) and writes the results into main memory followed by the computation of sum by process B, then we get the right result. But consider the case when B implements the statement sum = sum + f (B) before process A could write result into the main memory. Then the sum contains only f(B) which is not right. To avoid such inconsistencies, we use locking.
Process A Process B
sum = 0 :
: :
: lock sum
fork B sum = sum + f(B)
: unlock sum
lock sum :
sum = sum + f(A) end B
unlock sum
:
join B
:
end A
In this case whenever a process gets the sum variable, it locks it so that no other process can access that variable which makes sure the consistency in results.