Concurrency (L21) Flashcards
Concurrency refers to the concept of separate threads of control/execution within a program.
True.
Parallelism is a synonym for concurrency.
False.
What are parallel programs?
Programs that support simultaneous execution of distinct instruction streams.
By definition, concurrent programs are run on multiple CPUs or cores.
False - parallel programs are.
What are parallel programs typically used for?
Increasing the performance
of a specific algorithm.
What is concurrency typically used for?
For dividing the logic of a complex programs into distinct units.
All concurrent programs provide parallelism, but parallel programs are not necessarily concurrent.
False, all parallel programs provide
concurrency, but concurrent programs are not necessarily parallel.
Controlling access to shared data is necessarily the primary issue when writing concurrent programs.
False - sharing write-able data is a problem, but sharing read-only data is fine.
What is a critical region?
A block of code in which an update to shared data is made.
What is the traditional approach to only allowing one thread to enter/exit the critical region at a time.
Locks, which are locked (1) at the start of the critical region and closed (0) at its end.
Threads that encounter a lock perform another task and circle back later.
False - such threads wait/block until the data is unlocked.
It is not possible for multiple threads to think they have the lock for some data, as setting a lock variable is done in a single line of code.
False - this single high level language instruction is compiled into multiple
machine level instructions.
What is a “test and swap” instruction provided by modern CPUs?
An atomic assignment instruction that
always completes - allows languages to close locks “in a single instruction”.
What C library allows concurrency that follows the lock model?
pthreads
What does Java’s synchronized keyword do?
A synchronized method is tested such that this method can only be used by one thread at a time.
What are two problems encountered when using many threads that share many data structures?
Deadlock & livelock.
What is deadlock?
When separate threads are each waiting for a lock held by the other.
What is livelock?
When threads are actively trying to resolve a dependency or conflict, but their actions prevent any progress from being made, leading to a perpetual “busy” state.
Due to deadlock/livelock, low level thread management doesn’t scale very well.
True.
What concurrency model is provided by Erlang?
Message passing.
Messages are placed in a stack belonging to the receiving thread.
False - they are placed in queues.
A thread always knows where a message comes from.
False - this is true only when messages have a sender ID.
Messages are viewed as synchronous, since it allows threads to synchronize their tasks.
False - they are viewed as asynchronous since sending/receiving threads do not block during transfer.
It is impossible for processes to hang using message passing.
False - a process may wait wait for a message from a task that has crashed.
In message passing, the queue data structures are shared by separate tasks.
False.
