Enthuware

I'm a bit confused by this question.

The question itself states explicitly:

The new value of the element is generated by a utility class's static method, which takes in the existing value as a parameter and returns the new value.

But the answers are RecursiveAction, which does not return a value.
What am I missing here?

Confused.com..

The utility class's method returns the value that you need to update the array element with. But don't have to return that value from RecursiveAction. Please go through the sample implementation provided in the explanation.

HTH,
Paul.

Hi,
Why is option 3 not correct?
"Create a RecursiveAction that subdivides the task and submit all such tasks to a ForkJoinPool. "

The answer states "There is no need to submit the sub tasks to the pool. Once you fork a subtask, it is automatically executed by the pool."
ForkJoinTask has the invokeAll methods:

"public static void invokeAll(ForkJoinTask<?>... tasks) ... Forks the given tasks"
"public static <T extends ForkJoinTask<?>> Collection<T> invokeAll(Collection<T> tasks) ...Forks all tasks in the specified collection"

So by dividing a computation into separate chunks and optionally loading it into a collection you you can use your RA subclass instance to invoke those sub tasks.
How is invoking invokeAll on a RA instance not "submitting the tasks to the ForkJoinPool"? A RA exists in a FJP; when you call invokeAll you're saying, "here's a bunch of tasks I want you to execute, go off and execute them" to the FJP. I.e. the FJP is running all the threads; if you call invokeAll you are forking the subtasks and it's the FJP's job to do the forking, isn't it?

Thanks.

I am not sure exactly what you mean. Can you put it in code?
Once you submit the first/main RA to FJP, there is no need to submit any other task to FJP. The RA will just fork subtasks, which will be executed by the FJP without explicitly submitting the tasks to it.

HTH,
Paul.

Hi! I am really very confused because I cannot understand why is first a fork and then a compute:
newtask2.fork();
newtask1.compute();  
newtask2.join();

Whats the difference between fork and compute methods? Why not invoke two fork() and two join()?

Thanks.
Celeste.-

Please put what you are saying in code, run it, and paste the output so that it will be easier to understand your confusion.

i don't see any necessity to import java.util.ArrayList and java.util.Arrays in the example

Agree with answers 5 and 6. However not agree with possible implementation.
There is question about correct compute method.
methods fork and join should be invoked on left side when compute on right side.
newtask1.fork()
newtask2.compute()
newtask1.join()
or
invokeAll(newtask2, newtask1).
Example will work and compile but this is not a correct implementation in my opinion.

tanzwud wrote:Agree with answers 5 and 6. However not agree with possible implementation.
There is question about correct compute method.
methods fork and join should be invoked on left side when compute on right side.

Why do you think so? I don't see any reason why one is better than the other.
-Paul.

Yes my mistake. Seems like it works fine as soon as fork and join on one side and compute on another with no difference in speed.

I don't see where you guys takes out this is a "easy" question in category;

Option 3 is a bit misleading. For example, I understood it like "Create a RecursiveAction (implementation details) and submit all such tasks to a ForkJoinPool." In other words, it's like "all such tasks" refer to our RecursiveAction implementation instances, not to subtasks. Though, it may be my problem, of course.

In order for d) (fourth option) to be correct, should it say

Subclass RecursiveTask<V>, implement the compute() method,
and return the new value.

Updated.
thank you for your feedback!