LocalComputationManager improve design ?

[jonenst]

• Do you want to request a feature or report a bug?
  Bug

• What is the current behavior?
  LocalComputationManager uses 2 kinds of excutors:

• an executor to run threads doing the commands of execute() calls
• an executor to run a thread to submit the command tasks to the previous executor. Also this executor is returned by getExecutor().

So far so good. But I see two problems in the code:

• the first type of executor is a newly created executor at each call to execute().
• the thread submitting the tasks submits them all at once, one task per command (so potentially many tasks). But the first thing that these tasks do is to block if > availableCores tasks have already started. This creates many threads for nothing.

• What is the expected behavior?
  Don't create an executor at each execute call
  Don't create threads that immediately block

• What is the motivation / use case for changing the behavior?
  Reduce memory, improve performance, improve design