MULTITHREADING AND PARALLEL PROGRAMMING - Introduction to Java Programming: Comprehensive Version - page 1133

46 }
47 else {
48 int mid = (low + high) / 2 ;
49 RecursiveTask<Integer> left = new MaxTask(list, low, mid);
50 RecursiveTask<Integer> right = new MaxTask(list, mid, high);
51
52

split into two parts

right.fork();

fork right
fork left
join tasks

53 left.fork();

54 return new Integer(Math.max(left.join().intValue(),

55 right.join().intValue()));

56 }
57 }
58 }
59 }

The maximal number is 8999999
The number of processors is 2
Time is 44 milliseconds

Since the algorithm returns an integer, we define a task class for fork join by extending
RecursiveTask<Integer> (lines 26-58). The compute method is overridden to return the
max element in a list[low..high] (lines 39-57). If the list is small, it is more efficient to
be solved sequentially (lines 40-46). For a large list, it is split into two halves (lines 48-50).
The tasks left and right find the maximal element in the left half and right half, respec-
tively. Invoking fork() on the task causes the task to be executed (lines 52 and 53). The
join() method awaits for the task to complete and then returns the result (lines 54 and 55).

30.37

✓ ✓

How do you define a ForkJoinTask ? What are the differences between
RecursiveAction and RecursiveTask ?

Check

Point

30.38

How do you tell the system to execute a task?

30.39

What method can you use to test if a task has been completed?

30.40

How do you create a ForkJoinPool ? How do you place a task into a ForkJoinPool ?

K EY T ERMS

condition 1114
deadlock 1126
fail-fast 1128
fairness policy 1113
Fork/Join Framework

multithreading 1098
race condition 1111
semaphore 1124
synchronization wrapper

1127

1128 synchronized block 1112

lock 1112
monitor

thread 1098
thread-safe

1118 1111

C HAPTER S UMMARY
1. Each task is an instance of the Runnable interface. A thread is an object that facilitates
the execution of a task. You can define a task class by implementing the Runnable
interface and create a thread by wrapping a task using a Thread constructor.

2.

After a thread object is created, use the start() method to start a thread, and the
sleep(long) method to put a thread to sleep so that other threads get a chance to run.