多线程

Question

The application I'm currently working on performs some I/O or CPU intensive actions (file compression, file transfers, communicating with third party APIs, etc.) that occur when a user presses a 'Send' button.

I'm currently trying to persuade my employers that we should push these actions out to separate threads inside the main application (we'd need a maximum of two worker threads active at any given time), but my colleague has claimed that:

Any extra processing executed on a low priority thread could affect the usability of the GUI.

My view was that pushing I/O or CPU intensive activity to worker threads, updating the UI with Invoke calls during progress reporting, is pretty standard practise for handling intensive activity.

Am I incorrect? If so, could someone provide an explanation?

EDIT:

Thank you for the answers so far.

I should clarify: the colleague's solution to non-blocking is to spawn a child process containing a timer loop that scans a folder and processes the file compression/transfer activities. (Note that this doesn't cover the calls to third party APIs - I have no idea what his solution there would be.

The main issue with this approach is that the main application loses all scope on the state of whatever activity., leading to, IMHO, further complexity (his solution to progress reporting is to expose the Windows message pump in both processes and send custom messages between the two processes).

Answer 1

You are correct. Background threads are the very essence of keeping the UI active, precisely as you have described, via Invoke operations. Keeping everything on the GUI thread will, eventually clog up the plubming and make the GUI unresponsive.

Answer 2

The definitive answer would be to implement it as a proof-of-concept and then profile the app to see what sort of potential performance hit may or may not exist. Maybe on a

Having said that - it sounds like rubbish to me. In fact, it's quite the opposite - using additional threads are often the best way to keep the UI responsive.

Especially with things like the Task Parallel Library, it's just not very difficult to basic multi-threading.

Answer 3

Yes, you are correct. The general principle is that the thread that's responsible for responding to the user and keeping the user interface up to date (usually referred to as the UI thread) should never be used to perform any lengthy operation.
As a rule of thumb, anything that could take longer than about 30ms is a candidate for removal from the UI thread. This is a little aggressive — 30ms is about the shortest interval that most people will perceive as being anything other than instantaneous and it's actually slightly less than the interval between successive frames shown on a movie screen.