Chapter 2 Interprocess Communication & Synchronization

Interprocess Communication is needed when1. Two or more processes want to share/exchange data2. To prevent processes that enter a critical region from interfering with one another3. To ensure process dependencies are correctly sequenced (read then write)

*Two processes share a common fixed length buffer* Producer deposits, Consumer consumesProblem: Producer attempts to deposit but buffer is full / Consumer attempts to withdrawal but buffer is empty

Producer sleeps until awakened by consumer when buffer is NOT full / Consumer sleeps until awakened by Producer when buffer is NOT empty

When two processes attempt to read and/or write shared data and the final result depends on the order in which they are run

Critical Regions to provide Mutual Exclusion

Part of the program where shared data is accessed

1. No two processes simultaneously inside CR2. No assumptions may be made about the speed or number of CPUs3. No process running outside its CR may block another process4. No process should have to wait forever to enter its critical region

Goal: Use a shared variable to keep track of whose turn it is to enter the critical region-Violates condition 3: Process can forced to wait for a process in a non-critical region->Taking turns is not a good solution when onr process is slower than the other

Continuously testing a variable until some value appears (wastes CPU time)

Waiting process spins in a tight loop reading a variable waiting to see if it can proceed

Goal: Use lock/warning variables to remove strict alternationELABORATION:*before entering critical region, each process calls enter_region and waits until it is safe to enter*while inside set flag interested to trueIf another process is not enterested enter_region returns immediately* If not process hangs until interested flag goes false*flag is set to false when a process EXITS the critical region

1. Busy wait w/ Strict Alternation2. Busy wait w/ flags3. Bakery Algorithm (busy wait for multiple processes)*ALL are adequate but waste CPU resources (loops)SOLUTION: use hardware

1. Test & Set - test and set a variable in one instruction2. Atomic swap - switch register & memory in one instruction3. Turn off interrupts - processes won't be switched out unless it asks to be suspended

Two types: Test & set and Swap*single shared variable lock+ uses busy wait, but simpler+ works for any  number of processes-non concurrent code can lead to unbounded waiting-priority inversion

Higher priority process waits for lower priority process