PVW SKRIPT

Question 1

1) Define Parallel Programming.

2) Use example of summing up an array

Answer 1

1) Parallel programming is about using additional computational resources to solve a problem faster.

2) There is no sequential algorithm that can do better than O(n)
Suppose we had 4 processors, we could produce the result 4 times faster by partitioning the array into 4 segments, having each processor sum one segment and combining the results with an extra 3 additions.

Question 2

Define concurrent programming.

Answer 2

Concurrent Programming is about correctly and efficiently controlling access by multiple threads to shared resources.

Question 3

Define the programming model used in this lecture with key points use the following:

• threads

- shared memory

Answer 3

We assume:

• explicit threads with shared memory
• two or more threads can communicate by writing and reading fields of the same object
• a thread is like a running sequential programm that can create other threads
• threads can see the same object because we assume memory is shared

Question 4

What is scheduling?

Answer 4

There may be more threads than processors, than its up to the Java implementation, with help from the underlying operating system, to find a way to let the threads take turns, this is called scheduling.

Question 5

Why are multithreaded programs harder to test and debug?

Answer 5

Because the are nondeterministic. Rerunning the same multithreaded programm, can give different results.

Question 6

What are the different thread states in Java?

Answer 6

Non-Existing
New
Runnable
Blocked
Waiting
Terminated

Question 7

Define the Non-Existing state.

Answer 7

Non-Existing: Thread hasn’t been created yet

Question 8

Define the New state.

Answer 8

New: Once a Thread object is created, it enters new state

Question 9

Define the Runnable state.

Answer 9

Runnable: Once we call start() on the new thread object, it becomes eligible for execution and the system can start scheduling the thread as it wishes.

Question 10

Define the Blocked state.

Answer 10

Blocked: When a thread attempts to acquire a lock it goes into a blocked state until it has obtained the lock, upon which it returns to a runnable state. In addition, calling the join() method will also transfer a thread into a blocked state.

Question 11

Define the Waiting state.

Answer 11

Waiting: The thread can call wait() to go into a wating state. It will return to a runnable state once another thread calls notify() or notifyAll() and the thread is removed from the waiting queue.

Question 12

Define the Terminated state.

Answer 12

Treminated: We can use interrupt() to signal the thread to stop its execution, it will then transfer to a terminated state. Also exiting the run() method is equicvalent to entering a terminated state.

Question 13

What is a race condition?

Answer 13

A racecondition is a mistake in your program such that whether the program behaves correctly or not depends on the order in which the threads execute.

Race conditions are very common bugs in concurrent programming that, by definition, do not exist in sequential programming.

Question 14

What are the two types of race conditions?

Answer 14

Bad Interleaving

Data Race

Question 15

Define the race condition bad interleaving.

Answer 15

(High Level Race Condition) Erroneous programm behavior caused by an unfavorable execution order of a multithreaded algorithm

Question 16

Define the race condition data race.

Answer 16

Also described as “simultaneous access error”
There are two kinds of data races:

1) When one threads reads an object field at the same moment another thread writes the same field
2) When on thread might write an object field at the same moment that another thread also writes the same field

Question 17

What are 3 alternatives to the programm model with share memory?

Answer 17

1) Message-passing
2) Dataflow
3) Data parallelism

Question 18

How does the distributed memory alternative programm model Message-Passing work?

Answer 18

Message-passing is the natural alternative to shared memory. In this model, explicit threads do not share
objects. For them to communicate, copies of data are exchanged as messages between processes. As
objects are not shared between individual threads, the issue of threads wrongly updating fields does not
occur. One does, however, have to keep track of the different data copies being passed around through
messages.

Question 19

What is the speedup of a program?

Use picture definition

Answer 19

The ratio of the sequential execution time to the time given P processors.

Question 20

What is the intuitive meaning to Amdaahl’s Law?

Answer 20

Speedup of a parallel programm can be greatly decreased due to overhead of caused by several issues intoduced by parallelizing a program.

Especially a sequential part of a program can drastically impact the maximum achievable speedup.

Question 21

What is the formula to Amdahl’s Law?

Answer 21

Question 22

What is the formula for Gustafson’s law?

Answer 22

Question 23

Define a thread

Answer 23

• A set of instructions to be executed one at a time, in a specified order

Question 24

How to create Java Threads using java.lang.Runnable.

Answer 24

Question 24

How to create Java Threads using java.lang.Runnable.

Answer 24