3.2.2.2 Message Passing

The message passing is a mechanism transfering data directly from one process to a another one. We distinguish in the Message Passing ($=$ communication) between synchronous and asynchronous communication :

Figure 3.9: Blocking communication

In the blocking communication, all (often 2) involved processes wait until all processes signaled their readiness for data/message exchange.
The following communication acts for processes $P_1$ and $P_2$ like 2 semaphores :
$\bullet$ V($P_1$) : Ready for SEND
$\bullet$ P($P_1$) : Wait until $P_1$ is ready to receive.
$\bullet$ V($P_2$) : Ready for RECV
$\bullet$ P($P_2$) : Wait until $P_2$ is ready to send.

Figure 3.10: Non-blocking communication

In the non-blocking, all processes send or receive their data totally independent from the status of the remaining processes.

• Similar to the synchronization of characteristics using semaphores.
• Guarantees high efficiency in communication without deadlock (in difference to blocking communication).
• The sending process expects no acknowledgment that the operation was executed successfully. Output parameters of the receiving process must not have an undefined value - otherwise the whole program may crash. This to prevent is a function of the programmer!
• Other operations can be executed between the request for a communication routine and its appropriate (but not necessary) status query.

The semaphore concept is typical for shared memory computers wherein the resource management is usually unnoticed by the programmer. On the other hand, the message passing occurs on distributed memory computers and must be completed at least partially by the programmer.