Question

Wherever other questions on SO and other sites ask “How do I write concurrent code?“, the answers always involves pretty vague explanations such as “check for data dependencies” or “interdependencies” inside the code. I’m wondering what these mysterious dependencies actually look like as Java code!?!

• What’s a concrete example of a section of code that can be easily parallelized because it one part doesn’t have a dependency on another part?
• What’s a concrete example of a section of code that must be serial because of the existence of these dependencies?
• How do the existence of these dependencies weigh-in with the decision as to use a thread pool or not?

I guess I’m just not seeing the “forest through the trees” here. Thanks in advance!

Answer 1

What’s a concrete example of a section of code that can be easily
parallelized because it one part doesn’t have a dependency on another
part?

A concurrent application is build on the execution of tasks. You can identify a task as any discrete unit of work. If you can identify parts of your code that stands out as a discrete unit of work and has explicit task boundaries, you can create Runnables for each task and run it in a separate Thread achieving parallelism. Actually a Runnable represents the Task abstraction.
An example would be a process that needs to load multiple files for processing. The loading of files is a discrete unit of work and can be done in the background by threads without blocking the code flow

What’s a concrete example of a section of code that must be serial
because of the existence of these dependencies?

If Task 2 depends on the calculation/processing of Task 1 then Task 2 must wait for Task 1 to complete, thereby serializing the sequence.
An example is to load a file in the background and then search for a key in the file

How do the existence of these dependencies weigh-in with the decision
as to use a thread pool or not?

They don’t. Creating a Thread is an expensive operation and Thread Pools are a construct design in reusing threads so as to avoid the overhead of creating new ones. In the Thread Pool you just pass Tasks as Runnables and the Thread Pool is responsible to assign the Task to a thread or create a new thread if needed. You could also define policies via thread pools etc. But they are a tool for concurrent programs. You are expected to have already identified the tasks to submit to the thread pool for execution.