Lecture 7 - Concurrency and Synchronisation 1

Question 1

What is concurrency?

Answer 1

Multiple or things being executed at the same time

Question 2

What is the difference between real and simulated (apparent) concurrency?

Answer 2

Real concurrency: multiple CPUs genuinely executing things at the same time

Apparent concurrence: one CPU using scheduling to give the illusion of concurrency

Question 3

What is the benefit of concurrency?

Answer 3

Increases performance

Question 4

What is the downside to concurrency?

Answer 4

Not all resources can be accessed by multiple threads at the same time

Question 5

When is mutual exclusion needed?

Answer 5

When some resource can only be used by one thread at a time

E.g. shared memory structures, files, …

Question 6

When must synchronisation be used?

Answer 6

When multiple threads want to access the same object at the same time - critical sections

Question 7

What are the 4 conditions to provide mutual exclusion?

Answer 7

No two processes can simultaneously be in critical section

No assumptions made about speeds or numbers of cpus

No process running outside its critical sections may block another process

No process must wait forever to enter its critical region

Question 8

Only critical regions that access __________ need to synchronize with each other

Answer 8

The same resource

Question 9

Synchronisation for threads has the same pattern. What is it?

Answer 9

Attempt to enter critical section

Do stuff in critical section

Exit critical section

Do all non critical stuff

Question 10

How does strict alternation work to protect critical sections?

Whats the problem with it?

Answer 10

Use a shared variable to indicate which thread/process is currently in critical section

Avoids race conditions but can just get stuck

Question 11

How does the Bakery algorithm work to protect critical sections?

Answer 11

Before entering critical section, process receives a number

Numbering scheme always generates numbers in increasing order

Holder of smallest number enters critical section

If two processes receive same number, lower pid is served first

Question 12

How do locks work?

Answer 12

Create a lock associated with shared resource

Lock can be locked or unlocked:

locked - some thread is accessing the resource

Unlocked - not held by any thread

Question 13

How do locks prevent race conditions?

Answer 13

Allow only one thread at a time to use the shared item

Question 14

Whats the problem with locks?

Answer 14

Can reduce concurrency - may have several threads waiting for the one shared resource to become available

Question 15

What are the expected API operations of a lock?

Answer 15

Create new lock

Lock a lock

Unlock a lock

Declare variable as lock

Possibly:

Test if lock is lockable

Destroy lock to reclaim memory

Question 16

Can critical sections be used for syncing between processes?

Answer 16

no

Question 17

What are the lock operations in the posix api?

Answer 17

pthread_mutex_init

pthread_mutex_lock

pthread_mutex_unlock

pthread_mutex_destroy

Question 18

What is a spinlock?

Answer 18

Thread can lock the lock variable if it is unlocked

If not, sits in an infinite loop testing the variable until it becomes unlocked

Question 19

What is the time of check to time of use problem?

Answer 19

Where locks are interrupted between testing the variable and settting the variable if it was unlocked. I.e. race condition on the lock itself

Question 20

why does a spinlock need hardware support?

How does the hardware support work?

Answer 20

Checking if lock is locked and locking it needs to be atomic

1 = unlocked, 0 = locked

put 0 in a register, then exchange that register with the lock (atomic)

if register now contains 0 it was already locked - keep trying again

if register now contains 1 it was unlocked and is now locked