Lecture 8

Question 1

What is eager evaluation?

Answer 1

Arguments are evaluated before their procedure is called

Question 2

What is the declarative concurrent model of computation DCKL?

Answer 2

An extension of DSKL

Can be extended using data-driven or demand driven concurrency

Question 3

What is data-driven concurrency?

Answer 3

A thread of computation can execute when it has all its data

Question 4

What is demand-driven concurrency?

Answer 4

A thread of computation can be executed no sooner than when the result of its execution is needed by another thread.

Question 5

What does it mean to consume elements of a list while it is being created?

Answer 5

If we use threads to generate a list, and call browse on the list from another thread, it will display the elements as they are being created.

Browse consumes these elements.

Question 6

What is the syntax for threads?

Answer 6

<statement> ::= thread <s> end
</s></statement>

Question 7

What does a concurrent abstract machine contain?

Answer 7

Semantic stacks for each thread.

One single assignment store, shared across threads.

Question 8

What does a scheduler do?

Answer 8

Chooses what executable semantic stack to run. The top most statement is poped and executed.

A stack is executable when the topmost statement is not frozen.

Question 9

What is a fair scheduler?

Answer 9

Each semantic stack (thread) is eventually chosen.

Stacks are chosen with equal frequency.

Question 10

In a concurrent system, when is computation finished?

Answer 10

When all semantic stacks are removed.

A stack is removed when it is empty.

Question 11

What is the semantics of a thread?

(thread <s> end, E)</s>

Answer 11

Create a new stack
Push (<s>, E) on this stack</s>

Question 12

When is a data-driven model of computation declarative?

Answer 12

One of these are true:

All executions of the program must have the same result (logical equivalent)

All executions must fail

Question 13

What can make concurrent programs non-declarative?

Answer 13

If there is a unification error where two threads binds the same variable to different non-unifiable values. The program will always fail, however, the variable will take one of the bindings. If the variable is part of the result, it is not declarative. If it is not part of the result, it is declarative.

Introdusing exceptions

Question 14

What are the operations of the thread module?

Answer 14

Thread.this

Thread.suspend T (need to be explicitly resumed)

Thread.resume T

Thread.preempt T (can be scheduled again by scheduler)

Thread.terminate T (can no longer be run)

These are not part of the declarative model

Question 15

What is a stream?

Answer 15

Potentially infinite datastructure.

Elements can be added to end, and read from beginning.
These to operations can be performed concurrently.

Producer process: Adds elements
Consumer process: Reads elements

Question 16

How are streams implemented?

Answer 16

Using list-like structures with an unbound end:

declare UnboundEnd
Stream = 1|2|3|UnboundEnd

Question 17

How can streams be read from?

Answer 17

Read by starting from the head.

Stream = Head|RestOfStream

Can now use Head

Question 18

How can we extend the stream?

Answer 18

Binding the tail:

UnboundTail = NewElement|NewUnboundTail

Question 19

Give an example of streams being read concurrently while being produced.

Answer 19

thread Integers = {Enumerate 0 100} end
thread Result = {Sum Integers}

{Browse Result}

Enumerate: Producer
Sum: Consumer

Question 20

What affect the performance of a producer-stream-consumer pipeline?

Answer 20

Producer much faster than consumer. Can run out of memory because the producer allocates much faster than the consumer deallocates.

Producer is slower, then the consumer must wait.

Production and consumption speed can vary.