Coordinating Threads #

In the previous part of this tutorial we have used threads to make use of available processors via parallel execution when rendering images of fractals. We have not used language features for thread coordination. Threads were used only to compute images faster.

However, the FractalApp actually does use a simple form of thread coordination because rendering threads access and manipulate shared mutable state to store computed pixels.

Running computations using threads is also called asynchronous computation because by default we do not know in which order statements of different threads are executed. Using primitives for thread coordination we can establish a known order at least for some of the executed statements of different threads. Coordinating threads in such a way is, therefore, also called synchronization.

In this part we will discuss both simple and more involved forms of thread coordination. After learning about classic synchronization primitives we discuss alternative ways of flexible locking by means of another example application. Finally, we will learn about completable futures that can be used to combine asynchronous computations and discuss how to use scheduled executors for throttling tasks.